Patient-Reported Outcomes as Early Indicators of Disease Progression and Predictors of Survival via Machine Learning in Breast Cancer

Patient-Reported Outcomes as Early Indicators of Disease Progression and Predictors of Survival via Machine Learning in Breast Cancer | Research Square window.SnipcartSettings = { analytics: { enabled: false } }; (function() { var accessVector = localStorage.getItem('access_vector') || ''; window.dataLayer = window.dataLayer || []; if (accessVector) { window.dataLayer.push({ user: { profile: { profileInfo: { snid: accessVector } } } }); } })(); (function(w,d,s,l,i){w[l]=w[l]||[];w[l].push({'gtm.start':new Date().getTime(),event:'gtm.js'});var f=d.getElementsByTagName(s)[0],j=d.createElement(s),dl=l!='dataLayer'?'&l='+l:'';j.async=true;j.src='https://www.googletagmanager.com/gtm.js?id='+i+dl;f.parentNode.insertBefore(j,f);})(window,document,'script','dataLayer','GTM-K279D39R'); Browse Preprints In Review Journals COVID-19 Preprints AJE Video Bytes Research Tools Research Promotion AJE Professional Editing AJE Rubriq About Preprint Platform In Review Editorial Policies Our Team Advisory Board Help Center Sign In Submit a Preprint Cite Share Download PDF Article Patient-Reported Outcomes as Early Indicators of Disease Progression and Predictors of Survival via Machine Learning in Breast Cancer Wanbing Wang, Congyu Zhang, Benyam Muluneh, Quefeng Li, Jim H. Hughes, and 4 more This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-8013336/v1 This work is licensed under a CC BY 4.0 License Status: Posted Version 1 posted You are reading this latest preprint version Abstract Patient-reported outcomes (PROs) offer a non-invasive, low-cost way to capture patients’ experiences of symptoms, functioning, and quality of life. Yet, their potential as early indicators of tumor burden, disease progression, and survival remain unclear. In this study, we retrospectively analyzed 450,147 longitudinal PRO entries from 2,738 breast cancer patients pooled from four clinical trials, including both early- and late-stage disease, covering 15 PRO measures assessing symptoms, functioning, and quality of life. Among patients with radiographically confirmed disease progression, 90.6% experienced at least one deterioration in PROs prior to relapse detection (median PRO deterioration time 85 days vs. relapse time 1,338 days), indicating that PROs often worsen before imaging-confirmed relapse. Using Cox proportional hazards models, PRO deterioration was significantly associated with metastatic sites, tumor burden, and survival. Functional PROs were more affected by metastatic sites, whereas symptomatic PROs showed stronger associations with tumor burden and survival outcomes. Appetite loss had the strongest correlation with tumor burden, while fatigue was the most prognostic symptom for both overall survival (OS) and progression-free survival (PFS). The predictive value of PROs for OS was then evaluated using gradient boosting machine learning models. Models integrating PRO deterioration times of all subscales together with PFS achieved the best predictive performance, correctly identifying patient survival outcomes in over 95% of cases (AUC-ROC = 0.954), outperforming models using PROs (AUC-ROC = 0.838) or PFS alone (AUC-ROC = 0.896). This indicates that integrating PROs with PFS enhances the prediction of OS, providing a more powerful approach than using either measure alone. Together, these findings provide quantitative evidence that PROs can serve as early and complementary predictors of disease progression and survival, supporting their use as patient-centered biomarkers in breast cancer management. Our findings align with FDA and EMA efforts to integrate PROs into oncology endpoints, supporting more patient-centered regulatory evaluation. Health sciences/Oncology/Cancer/Breast cancer Health sciences/Signs and symptoms/Comorbidities Figures Figure 1 Figure 2 Figure 3 Figure 4 Figure 5 Introduction Timely detection and monitoring of disease progression is central to effective cancer management. In breast cancer, radiographic imaging is the current standard for disease monitoring. 1 However, imaging alone does not capture patients’ psychological, functional, or quality-of-life status. Frequent imaging for cancer surveillance also imposes financial and logistical burdens on both patients and healthcare systems 2 , while reducing the frequency risks delayed detection and treatment. These challenges highlight the need for complementary, patient-centered tools that can support efficient and continuous disease monitoring. Patient-reported outcomes (PROs) provide a non-invasive, low-cost means of capturing patients’ perspectives on symptoms, functioning, and quality of life. 3 They can be collected frequently and offer real-time insights into a patient’s evolving health status, potentially revealing early signs of disease progression or treatment failure. 4 , 5 Over the past decade, PROs have gained increasing recognition from patients, clinicians, and regulators as essential components of patient-centered drug development and value-based healthcare. 6 , 7 Despite these advantages, the role of PROs as objective predictors of tumor burden, disease progression, and survival remains poorly understood. Previous studies have yielded inconsistent findings: a meta-analysis of 45 Phase III randomized controlled trials reported no association between PRO improvements and progression-free survival (PFS) benefits, 8 whereas individual participant-level analyses in breast and lung cancer suggested that PROs are prognostic for PFS. 9 , 10 In addition, few studies have quantitatively examined how PROs relate to tumor burden or volume. These gaps make PROs difficult to interpret as objective biomarkers and perpetuate their perception as “soft” measures, limiting their use in clinical decision-making and drug development. 11 , 12 To address these gaps, we retrospectively analyzed 450,147 PRO entries collected using EORTC QLQ-C30 questionnaires from 2,738 breast cancer patients pooled from four clinical trials, encompassing both early- and late-stage disease. The EORTC QLQ-C30 questionnaire assessed patient-reported symptoms, functional status, and quality of life across 15 subscales. We compared the timing of PRO deterioration of all subscales with radiographically confirmed disease progression, and assessed their associations with metastatic sites, tumor burden, and survival outcomes. In addition, we applied machine learning models to evaluate the predictive value of PROs for overall survival (OS) and examined whether integrating PROs with PFS enhances predictive performance. Our findings show that PRO deterioration often precedes radiographic relapse and is strongly associated with disease burden and survival. Functional PROs were more affected by metastatic sites, whereas symptomatic PROs showed stronger associations with tumor burden and survival outcomes. Incorporating PROs with PFS can improve the OS prediction via machine learning models, indicating that PROs can serve as early, quantitative, and patient-centered indicators for disease monitoring and outcome prediction in breast cancer. Results Dataset Integration and Calculation of PRO Deterioration Times Longitudinal PRO data, assessed using the EORTC QLQ-C30 instrument 13 , were pooled together with demographics, clinical characteristics, disease progression, and survival data (PFS and OS) from four breast cancer Phase III clinical trials (BCIRG-001, BCIRG-005, CA012-0, and EFC6089). Longitudinal tumor size and metastatic sites information from the CA012-0 and EFC6089 studies were also collected. Patients without available PRO data were excluded from the analysis. The workflow of data inclusion and exclusion is shown in Supplementary Fig. 1 . A total of 2,738 patients with available PRO data were included in the final analysis cohort. Patient demographics and clinical characteristics across the four studies are summarized in Table 1 . In total, 52,190 PRO questionnaires were collected over time. Some questionnaires contained incomplete responses to the 30 items in EORTC QLQ-C30, resulting in data missingness. Overall, a total of 450,147 available PRO responses were included in our analysis and missingness was less than 5% across the dataset and remained relatively stable over time ( Supplementary Fig. 2 ); missingness in each questionnaire item is summarized in Supplementary Table 1 . Missing data were excluded from the analysis. For each patient, longitudinal responses to all items of the EORTC QLQ-C30 questionnaire were transformed to raw scores at the subscale level to estimate the time to first deterioration within each subscale (hereafter referred to as the PRO deterioration time; see Methods ). Table 1 Study demographics. Study BCIRG-001 (N = 735) BCIRG-005 (N = 1,564) CA012-0 (N = 226) EFC6089 (N = 213) Total (N = 2,738) NCT NCT00688740 NCT00312208 NCT00046527 NCT00081796 Treatment # FAC AC→T Paclitaxel capecitabine Age, years 49.0 (23.0, 70.0) 50.0 (22.0, 74.0) 52.0 (30.0, 83.0) 52.0 (30.0, 76.0) 50.0 (22.0, 83.0) Weight, Kg 68.0 (43.0, 163.6) 68.5 (38.5, 125.0) 68.0 (40.0, 106.2) 66.1 (37.0, 122.9) 68.0 (37.0, 163.6) Stage, N (%) II 377 (51.3%) 1557 (99.6%) 0 0 1934 (70.6%) III 44 (6.0%) 3 (0.2%) 0 6 (2.8%) 53 (1.9%) IV 314 (42.7%) 3 (0.2%) 226 (100.0%) 207 (97.2%) 750 (27.4%) Missing 0 1 (0.1%) 0 0 1 (< 0.1%) Race, N (%) White 642 (87.3%) 0 219 (96.9%) 160 (75.1%) 1021 (37.3%) Black of African American 11 (1.5%) 0 5 (2.2%) 0 16 (0.6%) Asian/Pacific Islander 31 (4.2%) 0 0 0 31 (1.1%) Other/Missing 51 (6.9%) 1564 (100.0%) 2 (0.9%) 53(24.9%) 1670 (61.0%) Menopause, N (%) Postmenopausal 249 (33.9%) 0 37 (16.4%) 0 286 (10.4%) Premenopausal 352 (47.9%) 702 (44.9%) 38 (16.8%) 55 (25.8%) 1,147 (41.9%) Other/Missing 134 (18.2%) 862 (55.1%) 151 (66.8%) 158 (74.2%) 1305 (47.7%) ECOG status, N (%) 0 565 (76.9%) 1529 (97.8%) 84 (37.2%) 102 (47.9%) 2280 (83.3%) 1 170 (23.1%) 35 (2.2%) 137 (60.6%) 89 (41.8%) 431 (15.7%) 2–3 0 0 5 (2.2%) 20 (9.4%) 25 (0.9%) Missing 0 0 0 2 (0.9%) 2 (0.1%) Estrogen receptors Positive 348 (47.3%) 1003 (64.1%) 24 (10.6%) 86 (40.4%) 1461 (53.4%) Negative 254 (34.6%) 437 (27.9%) 51 (22.6%) 100 (46.9%) 842 (30.8%) Unknown/Missing 133 (18.1%) 124 (7.9%) 151 (66.8%) 27 (12.7%) 435 (15.9%) Progesterone receptors Positive 354 (47.5%) 1056 (64.0%) 24 (10.6%) 87 (40.1%) 1521 (55.6%) Negative 257 (34.5%) 467 (28.3%) 52 (22.9%) 101 (46.5%) 877 (32.0%) Unknown/Missing 135 (18.1%) 126 (7.6%) 151 (66.5%) 29 (13.4%) 441 (16.1%) Overall survival, days* 3706 (2333, 3767) 1705 (1010, 2693) 216 (118, 277) 287 (168, 378) 2675 (893, 3735) Progression-free survival, days* 3353 (1136, 3746) 1245 (674, 2052) 84 (42, 155) 99 (43, 173) 1338 (342, 3544) Categorical variables are summarized as number of patients and percentages within the study; continuous variables as medians with ranges (minimum, maximum). # FAC = Fluorouracil, Doxorubicin, and Cyclophosphamide; AC→T = Doxorubicin + Cyclophosphamide followed by a Taxane. *Overall survival and progression-free survival are reported as medians with interquartile ranges (IQR) among patients who experienced the events. We also evaluated the baseline demographics and clinical characteristics that are associated with patient OS and PFS using Least Absolute Shrinkage and Selection Operator (LASSO). For OS, treatment, employment status, baseline number of metastatic lesions, ECOG performance status, and HER2 receptor status are significant covariates. For PFS, disease stage, age, treatment, menopausal status, ECOG performance status, estrogen receptor status, and HER2 receptor status are significant covariates. Forest plots were depicted to compare the contributions of each selected covariate for OS and PFS were in Supplementary Fig. 3a-b . PRO Deterioration Precedes Radiographically Confirmed Breast Cancer Relapse. We examined the temporal relationship between PRO deterioration and radiographically confirmed breast cancer relapse for all four studies, including both early- and late-stage breast cancer patients. Among the 892 patients who experienced relapse during follow-up period, 90.6% (N = 808) reported at least one PRO deterioration before relapse confirmation, indicating that PRO deterioration often precedes imaging-detected disease progression. The median and interquartile range (IQR) of PRO deterioration times across 15 subscales, including multiple symptoms, functioning, and quality of life domains, are summarized in Table 2 , together with the proportion of patients who did not experience PRO deterioration during the study period. The median time to deterioration across all PRO subscales was 85 days, compared with a median time to radiographic confirmed relapse of 1,338 days. The sequence of deterioration across 15 PRO subscales was visualized using a Sankey plot, illustrating the most common temporal order in which symptoms and functional domains worsened across patients. (Fig. 1 ) Appetite loss was among the earliest symptoms to deteriorate in patients who subsequently relapsed. We also analyzed the sequence of PRO deterioration specifically among early-stage breast cancer patients. Of the 684 patients who experienced recurrence in the BCIRG-001 and BCIRG-005 studies, 628 (91.8%) had at least one PRO deterioration prior to recurrence. These findings suggest that longitudinal PRO monitoring can provide early, patient-centered signals of disease progression and may serve as a valuable complementary tool to radiographic surveillance in breast cancer. Table 2 Time to first deterioration and proportion of censored patients across different patient-reported outcomes (PROs). Time to first deterioration BCIRG-001 (N = 735) BCIRG-005 (N = 1,564) CA012-0 (N = 226) EFC6089 (N = 213) Total (N = 2,738) Appetite loss, days 148 (51, 830) 176 (71, 856) 74 (48, 137) 57 (24, 128) 140 (66, 551) Censored, N (%) 391 (53.2%) 806 (51.5%) 146 (64.6%) 120 (56.3%) 1,463 (53.4%) Constipation, days 310 (86, 792) 282 (73, 884) 96 (50, 139) 71 (31, 150) 170 (69, 580) Censored, N (%) 404 (55.0%) 848 (54.2%) 179 (79.2%) 136 (63.8%) 1,567 (57.2%) Diarrhea, days 492 (106, 858) 378 (130, 903) 106 (58, 141) 50 (25, 112) 230 (93, 858) Censored, N (%) 515 (70.1%) 978 (62.5%) 187 (82.7%) 112 (52.6%) 1,792 (65.4%) Dyspnea, days 312 (90, 746) 179 (110, 586) 72 (47, 136) 85 (28, 158) 170 (81, 545) Censored, N (%) 405 (55.1%) 706 (45.1%) 141 (62.4%) 135 (63.4%) 1,387 (50.7%) Fatigue, days 87 (47, 316) 112 (66, 189) 50 (43, 112) 42 (23, 93) 86 (49, 177) Censored, N (%) 223 (30.3%) 392 (25.1%) 95 (42.0%) 81 (38.0%) 791 (28.9%) Financial, days 380 (92, 854) 386 (128, 903) 98 (50, 140) 88 (43, 172) 246 (88, 854) Censored, N (%) 494 (67.3%) 965 (61.7%) 170 (75.2%) 169 (79.3%) 1,798 (65.7%) Nausea/Vomit, days 95 (48, 505) 140 (67, 577) 92 (52, 139) 44 (23, 113) 109 (61, 523) Censored, N (%) 298 (40.6%) 728 (46.5%) 160 (70.8%) 105 (49.3%) 1,291 (47.2%) Pain, days 315 (88, 828) 175 (96, 555) 68 (44, 134) 60 (23, 124) 151 (71, 538) Censored, N (%) 406 (55.3%) 666 (42.6%) 124 (54.9%) 105 (49.3%) 1,301 (47.6%) Insomnia, days 176 (86, 615) 182 (92, 642) 77 (47, 136) 71 (27, 148) 154 (71, 540) Censored, N (%) 367 (50.0%) 717 (45.9%) 149 (65.9%) 139 (65.6%) 1,372 (50.2%) Global Health Status/QOL, days 134 (53, 534) 150 (69, 545) 72 (45, 135) 49 (24, 113) 132 (65, 504) Censored, N (%) 344 (46.9%) 632 (40.5%) 126 (55.8%) 109 (51.2%) 1,211 (44.3%) Physical functioning, days 486 (129, 860) 544 (196, 908) 94 (54, 140) 71 (38, 157) 382 (128, 877) Censored, N (%) 538 (73.3%) 1,196 (76.7%) 183 (81.0%) 152 (71.7%) 2,069 (75.7%) Role functioning, days 298 (53, 712) 358 (85, 891) 73 (46, 135) 71 (27, 143) 182 (69, 570) Censored, N (%) 405 (55.2%) 847 (54.2%) 152 (67.3%) 145 (68.1%) 1,549 (56.6%) Cognitive functioning, days 155 (51, 531) 353 (118, 855) 72 (45, 136) 49 (23, 114) 176 (69, 547) Censored, N (%) 494 (67.3%) 1,065 (68.3%) 162 (71.7%) 138 (65.1%) 1,859 (68%) Emotional functioning, days 420 (98, 852) 535 (145, 906) 76 (46, 136) 52 (24, 134) 349 (87, 862) Censored, N (%) 350 (47.7%) 802 (51.4%) 157 (69.5%) 120 (56.9%) 1,429 (52.3%) Social functioning, days 322 (86, 828) 365 (130, 888) 89 (47, 139) 57 (24, 129) 266 (76, 586) Censored, N (%) 413 (57.2%) 858 (54.9%) 168 (74.7%) 133 (63.0%) 1,572 (57.8%) Time to first deterioration were summarized by median with interquartile range (IQR). In this cohort, 113 patients who showed no deterioration across all PRO subscales during the study were classified as “non-symptomatic” breast cancer patients. Notably, 41.6% (N = 47) of these patients experienced disease relapse, indicating that the absence of PRO deterioration does not necessarily reflect the absence of disease progression. Radiographic assessments therefore remain essential in breast cancer disease monitoring. Metastatic Sites Involvement is Associated with PRO Deterioration. To assess whether the location of metastases is associated with distinct patterns of PRO deterioration, we constructed Cox proportional hazards models for time to first deterioration across each PRO subscale using patients with available metastatic sites data from the CA012-0 and EFC6089 studies (N = 432). The associations between metastatic sites and specific PROs are summarized in Fig. 2 a. Hazard ratios (HRs) and 95% confidence intervals (CIs) for functioning domains are presented in Fig. 2 b-e, and for other PRO subscales in Supplementary Fig. 4a-k . Functional PROs (physical functioning, role functioning, and cognitive functioning) were more affected by metastatic sites compared to symptomatic PROs. Patients with bone metastases tended to show earlier deterioration in physical, role, and emotional functioning compared with those without bone metastases (HR > 1, p < 0.05). Although bone metastases in breast cancer are commonly associated with significant pain 14 , our analysis indicates that pain symptoms may not manifest during the early stages of metastatic involvement. Additionally, patients with lymph node metastases tended to show earlier deterioration in gastrointestinal-related symptoms, including appetite loss and nausea/vomiting. These findings suggest that different metastatic sites are linked to distinct patterns of symptoms and functional deterioration, providing potential clinical insights for disease monitoring and metastasis detection. Tumor Burden is Associated with PRO Appetite Loss. We evaluated the association between tumor size and PRO deterioration using Cox proportional hazards models based on tumor size data from the CA012-0 and EFC6089 studies (N = 359). Larger tumor size was associated with earlier PRO deterioration (HR > 1), with appetite loss showing the strongest association across all subscales (HR = 1.74, 95% CI 1.33–2.26, p < 0.001; Fig. 3 a). Pain was also strongly correlated with tumor size and patients with larger tumors reported earlier onset of pain symptoms (HR = 1.68, 95% CI 1.33–2.13, p < 0.001; Fig. 3 a). To further explore the relationship between tumor burden and the two PROs most strongly associated with it—appetite loss and pain—we compared the tumor regrowth ratio, defined as the amount of tumor growth after its smallest size (nadir) during treatment, between patients who experienced PRO deterioration and those who did not. Patients with appetite deterioration had significantly higher tumor regrowth ratios than those without (p < 0.01; Fig. 3 b). Similarly, patients with pain deterioration had significantly higher tumor regrowth ratios (p < 0.001; Fig. 3 c). Patients whose tumors regrew by ≥ 20% from their nadir (defined as disease progression per RECIST criteria 15 ) experienced significantly earlier worsening of appetite (p < 0.01; Fig. 3 d) and pain (p < 0.0001; Fig. 3 e). These findings suggest that tumor progression contributes to early systemic symptoms such as appetite loss and pain, potentially reflecting metabolic or inflammatory responses to increasing tumor burden. Monitoring these symptoms may provide early signals of tumor relapse in breast cancer. PRO Deterioration is Associated with Patient Survival To further examine the relationship between PRO deterioration and survival outcomes in breast cancer, we constructed time-dependent Cox proportional hazards models for OS and PFS using PRO deterioration times as covariates. 16 To account for potential confounding effects, baseline demographic and clinical variables associated with OS or PFS were identified using the LASSO algorithm and included as covariates in the multivariable models. The HRs and 95% CIs for each PRO subscale are presented for OS in Fig. 4 a and for PFS in Fig. 4 b. Across all subscales, HRs < 1 indicated that patients who experienced earlier PRO deterioration tended to have shorter OS and faster disease progression. Among all domains, fatigue showed the lowest HR in both OS (HR = 0.13, 95% CI 0.11–0.15, p < 0.0001) and PFS (HR = 0.24, 95% CI 0.21–0.27, p < 0.0001) models, suggesting that fatigue-related PRO deterioration was most strongly associated with poorer survival outcomes in breast cancer. Integrating PRO with PFS Enhances OS Prediction Using Machine Learning The prognostic value of PRO data was evaluated using machine learning approaches. We developed gradient-boosting machine learning models to predict OS status based on different feature sets: (1) PRO deterioration times from all subscales, (2) PFS only, and (3) a combination of PRO deterioration times from all subscales and PFS. Among these models, integrating PRO with PFS achieved the best predictive performance for OS, with an area under the receiver operating characteristic curve (AUC-ROC) of 0.954 (Fig. 5 a). The AUC-ROC measures how well a model distinguishes between patients who experienced the event and those who did not. A higher value indicates better discrimination. This combined model outperformed the model using PFS alone (AUC-ROC = 0.896), suggesting that including PRO data meaningfully improves the prediction of OS. Because PFS data may include survival information that becomes available only after the outcome occurs, it can unintentionally include signals related to OS, leading to biased results in the prediction. We built a landmark machine learning model using only PFS data recorded up to one year. This approach helps prevent information leakage and provides a fair estimate of the model’s ability to predict outcomes after day 365. In this analysis, combining PRO with PFS achieved the best prediction of OS (AUC-ROC = 0.876), outperforming models that used PRO alone (AUC-ROC = 0.838) or PFS alone (AUC-ROC = 0.609) (Fig. 5 b). In this landmark setting, PRO alone performed better than PFS, suggesting that PROs may offer earlier and complementary information for predicting long-term survival. Discussion In this study, we retrospectively evaluated the deterioration of 15 patient-reported symptoms, functioning scales, and quality of life using 450,147 PRO data from 2,738 breast cancer patients, including both early- and late-stage disease. Association between PRO deterioration and cancer relapse, tumor burden, and metastatic profiles were identified based on this large PRO dataset. We further observed that different metastatic sites were associated with distinct PRO deterioration patterns, and that larger tumor size correlated with earlier PRO deterioration. Using machine learning approaches, we confirmed that adding PRO to PFS can enhance the OS prediction compared to using PFS alone. Together, these findings provide quantitative evidence that longitudinal PRO monitoring can serve as early, non-invasive indicators of disease progression and survival outcomes in breast cancer. In our analysis, among breast cancer patients who later developed radiographically confirmed relapse, more than 90% exhibited at least one PRO deterioration before relapse detection, indicating that PRO decline often precedes imaging evidence of recurrence. A similar pattern was observed in early-stage breast cancer patients, with approximately 90% showing PRO deterioration prior to recurrence. These findings provide direct evidence that PROs may serve as early indicators of disease progression or recurrence in breast cancer. An interesting finding from our study is that functional PRO subscales were more affected by metastatic sites, whereas symptomatic PROs showed stronger associations with disease burden and survival. This may be because symptomatic PROs more directly reflect tumor-related physiological changes, while functional PROs capture broader aspects of overall health status that often involve multiple organ systems. 17 Among all the PRO subscales, appetite loss demonstrated the strongest correlation with tumor burden. This association may be driven by tumor-induced systemic inflammation and altered metabolic and hormonal signaling in breast cancer. 18 , 19 , 20 Furthermore, fatigue emerged as a key prognostic symptom for survival, showing the lowest HRs for both OS and PFS time-dependent Cox proportional hazards models. This indicates that earlier onset of fatigue is associated with earlier disease progression and worsening fatigue may signify the need to evaluate other indicators of progression. Both appetite loss and fatigue are common and distressing symptoms experienced by breast cancer patients, often resulting from disease progression and treatment-related toxicity. 21 Our study shows that PROs represent critical early signals of disease progression and should be systematically monitored as a part of cancer care. 22 To further explore the prognostic potential of PROs, we developed machine learning models to predict OS using PFS and PROs. Machine learning algorithms were used because PROs and PFS are potentially correlated and their associations with OS are nonlinear and complex, which may not be well captured by traditional linear regression models. 23 , 24 Using a gradient boosting machine learning framework, we found integrating PROs with PFS can substantially enhance OS prediction compared to using PFS alone. These results point to the potential of developing composite surrogate endpoints that integrate PROs and PFS for more accurate OS prediction, consistent with evidence from previous studies. 25 , 26 Despite their clinical relevance, PROs have not yet been fully integrated into standard oncology practice. 27 Our findings provide strong evidence supporting the incorporation of PROs into electronic health record systems to facilitate continuous clinician monitoring and improve patient management. Integration of PROs into routine clinical workflows could enable earlier identification of disease progression and treatment-related toxicity, enhancing timely clinical intervention. 28 , 29 Furthermore, standardized electronic PRO collection could generate real-world data to inform regulatory decision-making and optimize patient-centered care models. 30 While our results demonstrate the potential of PROs to provide early signals of disease progression before radiographic detection, we emphasize the importance of using PRO data in tandem with radiographic tools. In our dataset, we identified 113 “non-symptomatic” breast cancer patients whose all PRO measures showed no deterioration throughout the study. However, 41.6% of these patients still experienced cancer relapse. This finding suggests that while PROs can serve as valuable surveillance tools for a large proportion of breast cancer patients, clinicians should exercise caution when interpreting results for “non-symptomatic” individuals. A lack of “PRO progression” should not be interpreted as support for reducing guideline-recommended clinical assessment and surveillance. Our study has limitations. We evaluated PRO changes using the time to first deterioration approach; however, as demonstrated in our previous analyses, this simplification may not be sensitive enough due to the substantial variability and noise inherent in longitudinal PRO measurements. Future work should leverage more advanced computational methods to capture the clinically meaningful changes in PRO trajectories, such as population modeling approach, which may improve the precision of evaluating how PRO changes are associated with disease progression and survival outcomes. 31 , 32 In conclusion, we analyzed the associations between PRO deterioration and cancer relapse, tumor burden, and survival outcomes in breast cancer. Our findings demonstrate that PROs can serve as complementary, patient-centered indicators to radiographic assessments, enabling non-invasive and more frequent disease monitoring. This work highlights the potential of PROs to evolve from supportive endpoints into reliable clinical biomarkers that enhance disease surveillance in breast cancer care. Our findings align with the ongoing efforts of the FDA and EMA to integrate PROs into oncology trial endpoints and labeling claims, paving the way for more patient-centered regulatory evaluation of cancer therapies. Methods Ethic Statement The data used in this analysis are fully de-identified individual participant data for secondary analysis. The study was granted an Institutional Review Board (IRB) exemption by the University of North Carolina at Chapel Hill. Clinical Study Descriptions Data were pooled from the control arms of four Phase III clinical trials of breast cancer, including female patients with early to late-stage disease. The studies included are summarized below. BCIRG-001 (NCT00688740) This was an open-label, multicenter Phase III trial that compared efficacy and safety between standard adjuvant anthracycline chemotherapy and anthracycline–taxane combination chemotherapy in women with operable, node-positive, early-stage breast cancer. 33 Patients who received fluorouracil, doxorubicin, and cyclophosphamide (FAC) every 3 weeks for six cycles were included in the analysis. The study provided 10-year follow-up data on tumor recurrence. BCIRG-005 (NCT00312208) This was a Phase III, multicenter, prospective trial evaluating efficacy and safety of sequential versus concurrent administration of doxorubicin and cyclophosphamide with docetaxel in women with node-positive, non-metastatic breast cancer. 34 The analysis included patients who received AC→T (doxorubicin and cyclophosphamide every 3 weeks for four cycles, followed by docetaxel every 3 weeks for four cycles). Tumor recurrence data were available with 10-year follow-up. CA012-0 (NCT00046527) This was a randomized, open-label, multicenter Phase III trial comparing ABI-007 (a Cremophor-free, protein-stabilized nanoparticle formulation of paclitaxel) with conventional paclitaxel in patients with metastatic breast cancer. Patients who received conventional paclitaxel were included in the analysis. This study provided longitudinal data on tumor size and metastatic organ involvement for individual patients. EFC6089 (NCT00081796) This was a randomized, open-label Phase III trial comparing larotaxel with capecitabine in patients with metastatic breast cancer progressing after taxane and anthracycline therapy. Patients received capecitabine twice daily for 2 weeks in 3-week cycles. The dataset included longitudinal measurements of tumor size and metastatic organ involvement. PRO Measures The PRO measures used in this study were the EORTC QLQ-C30 questionnaire. 35 The EORTC QLQ-C30 is a 30 item measures that assesses cancer patients disease-related symptoms, functional status, and overall quality of life ( Supplementary Table 1 ). Symptom scales assess common cancer- and treatment-related symptoms such as fatigue, appetite loss, nausea/vomiting, pain, while functional status measure physical, emotional, cognitive, role, and social functioning. The EORTC QLQ-C30, as one of the most commonly used PRO instrument, has been extensively validated across cancer types and languages, demonstrating strong reliability, construct validity, and sensitivity to clinical change. 36 , 37 , 38 PRO Deterioration Time The PRO responses from the EORTC QLQ-C30 were first transformed into subscale scores according to the scoring manual 35 using the following equations: $$\:RS=({I}_{1\:}+{I}_{2\:}+\dots\:+\:{I}_{n})/n$$ Functional scales: \(\:S=\left\{1-\:\frac{(RS-1)}{range}\right\}\bullet\:100\) Symptom scales/QoL scale: \(\:S=\left\{(RS-1)/range\right\}\bullet\:100\) Where RS is the raw score for each PRO subscale calculated from the mean of the corresponding item scores, and S is the transformed subscale score ranging from 0 to 100. The item composition and range for each subscale are provided in Supplementary Table 2 . If one PRO response was missing, the entire score was excluded from the analysis. The time to first PRO deterioration (referred to as PRO deterioration time) was defined as the first time a patient’s transformed subscale score showed a clinically meaningful change of ≥ 10 points from baseline, specifically, an increase of ≥ 10 points for symptom or functioning scales, or a decrease of ≥ 10 points for the global QoL scale. 39 Cox Proportional Hazards Model Cox proportional hazards models were used to evaluate associations between metastatic organs and PRO deterioration, with time to first deterioration as the dependent variable and metastatic organ involvement (pleura, lung, bone, lymph nodes, liver, breast, and other organs) as covariates, based on data from the CA012-0 and EFC6089 studies. Patients without tumor metastatic site information were excluded from this analysis. Separate models were constructed for each of the 15 PRO subscales. In these models, HR greater than 1 indicated that patients with a given metastatic site had a higher risk of earlier PRO deterioration compared to those without metastases at that site. Similarly, Cox proportional hazards models were applied to assess associations between tumor size and PRO deterioration using the same datasets from the CA012-0 and EFC6089 studies. Patients without tumor size data were excluded from this analysis. A total of 15 models were developed for each PRO subscale, with higher HRs indicating a stronger association between larger tumor size and earlier PRO deterioration. Time-Dependent Cox Proportional Hazards Model PRO deterioration time was used as a time-dependent covariate in Cox proportional hazards regression models to examine its associations with OS and PFS using pooled data from four studies (BCIRG-001, BCIRG-005, CA012-0, and EFC6089). Patients without available OS or PFS data were excluded from this analysis. In the time-dependent Cox proportional hazards model, the hazard function for patient \(\:i\) at time \(\:t\) , conditional on the covariate history up to time \(\:t\) , was defined as: $$\:{\lambda\:}_{i}\left(t\:|\:{\mathcal{H}}_{i}\left(t\right)\right)={\lambda\:}_{0}\left(t\right)exp\left\{{\beta\:}_{1}{D}_{i}\left(t\right)+{\theta\:}^{T}{Z}_{i}\right\},\:t\ge\:0$$ where \(\:{\lambda\:}_{0}\left(t\right)\) is an unspecified baseline hazard function, \(\:{D}_{i}\left(t\right)\:\) is a time-dependent indicator variable that equals 1 after the first 10-point deterioration and 0 otherwise, and \(\:{Z}_{i}\) is a vector of baseline fixed (non-time-dependent) covariates. 16 Models were constructed for each PRO subscale, incorporating LASSO-selected baseline covariates to adjust for potential confounders. 40 The most prognostic covariates were selected by the lambda value within one standard error from the minimum (lambda.1se). All variables were standardized to a mean of zero and variance of one for consistency. Forest plots were depicted to compare the contributions of each selected covariate. In these models, HRs below 1 indicated that earlier PRO deterioration was associated with shorter survival or faster disease progression. Population Model Predicting Tumor Size at PRO Deterioration Times Because tumor size obtained via imaging tools (e.g., CT scans) were measured less frequently than PRO assessments, the timing of PRO deterioration often did not align with tumor size measurements. To match the tumor size with PRO deterioration time, we developed a population tumor growth model using a non-linear mixed-effects estimation approach. This method allowed pooling of tumor size data from all patients in the study to simultaneously characterize both the population-level average tumor growth pattern and individual patient trajectories. 41 Using this model, we predicted tumor size at the time of PRO deterioration and examined their associations in the Cox proportional hazards models. This modeling approach has been validated in previous studies for assessing tumor growth heterogeneity across different organs. 42 , 43 Machine Learning Models Predicting OS To evaluate the prognostic value of PRO deterioration times and PFS, we applied XGBoost machine learning algorithms to predict OS using pooled data from four studies (BCIRG-001, BCIRG-005, CA012-0, and EFC6089). Patients without OS status information were excluded from this analysis. Model performance was assessed using the area under the receiver operating characteristic curve (AUC-ROC), which quantifies the model’s ability to distinguish between patients who survived and those who died (AUC = 1.0 indicates perfect prediction; 0.8 indicates good performance; 0.5 represents random chance). We compared AUC-ROC values across three feature sets: (1) PRO deterioration times only; (2) PFS only; (3) A combination of PRO deterioration times and PFS. A landmark XGBoost model was also developed using the same dataset and feature sets. In the landmark analysis, only PFS data available up to day 365 were included to prevent information leakage from future outcomes. To ensure fair comparisons across models, we (1) used identical training and test splits (80% for training data cohort and 20% for testing data cohort), (2) applied consistent preprocessing pipelines (including scaling and feature encoding), and (3) optimized hyperparameters using the same search strategy. Statistical Analysis For group comparisons, Wilcoxon tests were applied to assess differences in continuous variables between two groups. Group differences in Kaplan–Meier curves were evaluated using log-rank tests. Cox proportional hazards models were constructed to estimate HRs with corresponding 95% CIs and p-values. Statistical significance was defined as p < 0.05. For p-value not shown, * p < 0.05; ** p < 0.01; *** p < 0.001; **** p < 0.0001. Data Availability The raw data were accessed through Project Data Sphere under license in accordance with their Data Use Agreement. Project Data Sphere is an open-access platform that aggregates de-identified cancer clinical trial data from biopharmaceutical companies, academic medical centers, and government organizations ( https://www.projectdatasphere.org/ ). The analysis data are available from the corresponding author upon reasonable request. Analysis Codes Availability The statistical analyses were performed using R 4.4.1 and RStudio Version 2022.07.1 + 554. The population tumor growth model was developed using Monolix 2024R1 and model simulations were performed using Simulx 2024R1. Both Monolix and Simulx are available at https://lixoft.com/products/ . The figures were compiled in Adobe Illustrator 2025. The source codes were provided in https://github.com/Whitney0208/PRO-Breast.git . Declarations Conflicts of Interests: B.M may own stock in Novartis. J.H.H is a current employee of Pfizer and may own stock in Pfizer. L.I.W W receives institutional research funding from the National Cancer Institute and previously received personal fees from Celgene/Bristol Myers Squibb as a member of the Scientific Steering Committee for the Connect Multiple Myeloma patient registry. W.A.W. has received institutional research funding from Genentech; receives consulting fees from Teladoc Health, Quantum Health, and Lantern Health; holds equity in Koneksa Health; and has a leadership position in the American Society of Hematology Research Collaborative. E.B. receives institutional research funding from the National Cancer Institute and from the Patient-Centered Outcomes Research Institute, and personal fees for scientific advising from Research Triangle Institute, Thyme Care, N-Power Medicine, Resilience Health, Canopy Care, Savor Health, and Navigating Cancer. Acknowledgements This work was funded by University of North Carolina at Chapel Hill. We disclose the use of AI tools (ChatGPT, Open AI) to assist with grammar and language revisions in the manuscript. References Pulumati A, Pulumati A, Dwarakanath BS, Verma A, Papineni RVL (2023) Technological advancements in cancer diagnostics: Improvements and limitations. Cancer Rep (Hoboken) 6:e1764 Wilson BE, Wright K, Koven R, Booth CM (2024) Surveillance Imaging After Curative-Intent Treatment for Cancer: Benefits, Harms, and Evidence. J Clin Oncol 42:2245–2249 The Lancet N (2019) Patient-reported outcomes in the spotlight. Lancet Neurol 18:981 Di Maio M, Basch E, Bryce J, Perrone F (2016) Patient-reported outcomes in the evaluation of toxicity of anticancer treatments. Nat Rev Clin Oncol 13:319–325 LeBlanc TW, Abernethy AP (2017) Patient-reported outcomes in cancer care - hearing the patient voice at greater volume. Nat Rev Clin Oncol 14:763–772 Basch E (2018) Patient-reported outcomes: an essential component of oncology drug development and regulatory review. Lancet Oncol 19:595–597 Yap C et al (2024) Advancing patient-centric care: integrating patient reported outcomes for tolerability assessment in early phase clinical trials - insights from an expert virtual roundtable. EClinicalMedicine 76:102838 Samuel JN, Booth CM, Eisenhauer E, Brundage M, Berry SR, Gyawali B (2022) Association of Quality-of-Life Outcomes in Cancer Drug Trials With Survival Outcomes and Drug Class. JAMA Oncol 8:879–886 Badaoui S, Kichenadasse G, Rowland A, Sorich MJ, Hopkins AM (2021) Patient-Reported Outcomes Predict Progression-Free Survival of Patients with Advanced Breast Cancer Treated with Abemaciclib. Oncologist 26:562–568 Zhou J, Muluneh B, Wang Z, Yao H, Hughes JH (2025) Leveraging Longitudinal Patient-Reported Outcome Trajectories to Predict Survival in Non-Small Cell Lung Cancer. Clin Cancer Res 31:2685–2694 Frost MH, Sloan JA (2002) Quality of life measurements: a soft outcome–or is it? Am J Manag Care 8:S574–579 Makhni EC, Hennekes ME (2023) The Use of Patient-Reported Outcome Measures in Clinical Practice and Clinical Decision Making. J Am Acad Orthop Surg 31:1059–1066 Kaasa S et al (1995) The EORTC core quality of life questionnaire (QLQ-C30): validity and reliability when analysed with patients treated with palliative radiotherapy. Eur J Cancer 31a:2260–2263 Takei D, Tagami K (2023) Management of cancer pain due to bone metastasis. J Bone Min Metab 41:327–336 Eisenhauer EA et al (2009) New response evaluation criteria in solid tumours: revised RECIST guideline (version 1.1). Eur J Cancer 45:228–247 Fisher LD, Lin DY (1999) Time-dependent covariates in the Cox proportional-hazards regression model. Annu Rev Public Health 20:145–157 Kluetz PG et al (2016) Focusing on Core Patient-Reported Outcomes in Cancer Clinical Trials: Symptomatic Adverse Events, Physical Function, and Disease-Related Symptoms. Clin Cancer Res 22:1553–1558 Koh K et al (2025) Cancer-Associated Cachexia: Bridging Clinical Findings with Mechanistic Insights in Human Studies. Cancer Discov 15:1543–1568 Roeland EJ et al (2020) Management of Cancer Cachexia: ASCO Guideline. J Clin Oncol 38:2438–2453 Yeom E, Yu K (2022) Understanding the molecular basis of anorexia and tissue wasting in cancer cachexia. Exp Mol Med 54:426–432 Abrahams HJG, Gielissen MFM, Schmits IC, Verhagen C, Rovers MM, Knoop H (2016) Risk factors, prevalence, and course of severe fatigue after breast cancer treatment: a meta-analysis involving 12 327 breast cancer survivors. Ann Oncol 27:965–974 Gradishar WJ et al (2024) Breast cancer, version 3.2024, NCCN clinical practice guidelines in oncology. J Natl Compr Canc Netw 22:331–357 Krepper D, Cesari M, Hubel NJ, Zelger P, Sztankay MJ (2024) Machine learning models including patient-reported outcome data in oncology: a systematic literature review and analysis of their reporting quality. J Patient Rep Outcomes 8:126 Siena S et al (2007) Association of progression-free survival with patient-reported outcomes and survival: results from a randomised phase 3 trial of panitumumab. Br J Cancer 97:1469–1474 Kortüm M et al (2025) Symptomatic progression-free survival as an emerging patient-centered endpoint in multiple myeloma: a secondary analysis of MagnetsiMM-3 trial data. BMC Cancer 25:1288 Zhang H et al (2025) Improving Surrogate Endpoints for Survival Prediction Through Integration of Patient-Reported Outcomes. medRxiv , 2010. 2002.25337154 (2025) Colbert JA, Potters L (2025) Overcoming Barriers to Make Patient-Reported Outcome Collection the Standard of Care in Oncology. JAMA Oncol 11:233–234 Mazariego C et al (2022) Priority recommendations for the implementation of patient-reported outcomes in clinical cancer care: a Delphi study. J Cancer Surviv 16:33–43 Stover A et al (2015) Integrating Patient-Reported Outcome Measures into Routine Cancer Care: Cancer Patients' and Clinicians' Perceptions of Acceptability and Value. EGEMS (Wash DC) 3:1169 Bellino S, La Salvia A (2024) The Importance of Patient Reported Outcomes in Oncology Clinical Trials and Clinical Practice to Inform Regulatory and Healthcare Decision-Making. Drugs R D 24:123–127 Zhou J, Muluneh B, Li Q, Hughes JH (2025) Revolutionizing Patient-Reported Outcomes Analysis for Oncology Drug Development Using Population Models. Clin Cancer Res 31:1580–1586 Zou Y, Sun Y, Ravva S, Wagner LI, Zhou J (2025) A Model-Based Meta-Analysis of Pembrolizumab Effects on Patient-Reported Quality of Life: Advancing Patient-Centered Oncology Drug Development. CPT Pharmacometrics Syst Pharmacol Mackey JR et al (2013) Adjuvant docetaxel, doxorubicin, and cyclophosphamide in node-positive breast cancer: 10-year follow-up of the phase 3 randomised BCIRG 001 trial. Lancet Oncol 14:72–80 Mackey JR et al (2016) Long-term outcomes after adjuvant treatment of sequential versus combination docetaxel with doxorubicin and cyclophosphamide in node-positive breast cancer: BCIRG-005 randomized trial. Ann Oncol 27:1041–1047 Fayers P, Aaronson NK, Bjordal K, Grønvold M, Curran D, Bottomley A (2001) EORTC QLQ-C30 scoring manual. European Organisation for research and treatment of cancer Aaronson NK et al (1993) The European Organization for Research and Treatment of Cancer QLQ-C30: a quality-of-life instrument for use in international clinical trials in oncology. J Natl Cancer Inst 85:365–376 Fayers P, Bottomley A (2002) Quality of life research within the EORTC-the EORTC QLQ-C30. European Organisation for Research and Treatment of Cancer. Eur J Cancer 38(Suppl 4):S125–133 Scott NW et al (2006) Comparing translations of the EORTC QLQ-C30 using differential item functioning analyses. Qual Life Res 15, 1103–1115; discussion 1117–1120 Cocks K et al (2024) Time to deterioration of patient-reported outcome endpoints in cancer clinical trials: targeted literature review and best practice recommendations. J Patient Rep Outcomes 8:150 Tibshirani R (1996) Regression shrinkage and selection via the lasso. J Royal Stat Soc Ser B: Stat Methodol 58:267–288 Mould DR, Upton RN (2012) Basic concepts in population modeling, simulation, and model-based drug development. CPT Pharmacometrics Syst Pharmacol 1:e6 Zhou J, Cipriani A, Liu Y, Fang G, Li Q, Cao Y (2023) Mapping lesion-specific response and progression dynamics and inter-organ variability in metastatic colorectal cancer. Nat Commun 14:417 Zhou J, Liu Y, Zhang Y, Li Q, Cao Y (2020) Modeling Tumor Evolutionary Dynamics to Predict Clinical Outcomes for Patients with Metastatic Colorectal Cancer: A Retrospective Analysis. Cancer Res 80:591–601 Additional Declarations Yes there is potential Competing Interest. B.M may own stock in Novartis. J.H.H is a current employee of Pfizer and may own stock in Pfizer. L.I.W W receives institutional research funding from the National Cancer Institute and previously received personal fees from Celgene/Bristol Myers Squibb as a member of the Scientific Steering Committee for the Connect Multiple Myeloma patient registry. W.A.W. has received institutional research funding from Genentech; receives consulting fees from Teladoc Health, Quantum Health, and Lantern Health; holds equity in Koneksa Health; and has a leadership position in the American Society of Hematology Research Collaborative. E.B. receives institutional research funding from the National Cancer Institute and from the Patient-Centered Outcomes Research Institute, and personal fees for scientific advising from Research Triangle Institute, Thyme Care, N-Power Medicine, Resilience Health, Canopy Care, Savor Health, and Navigating Cancer. Supplementary Files SupplementaryFiles11.2.2025.docx Supplementary Files Cite Share Download PDF Status: Posted Version 1 posted You are reading this latest preprint version Research Square lets you share your work early, gain feedback from the community, and start making changes to your manuscript prior to peer review in a journal. As a division of Research Square Company, we’re committed to making research communication faster, fairer, and more useful. We do this by developing innovative software and high quality services for the global research community. Our growing team is made up of researchers and industry professionals working together to solve the most critical problems facing scientific publishing. Also discoverable on Platform About Our Team In Review Editorial Policies Advisory Board Help Center Resources Author Services Accessibility API Access RSS feed Manage Cookie Preferences © Research Square 2026 | ISSN 2693-5015 (online) Privacy Policy Terms of Service Do Not Sell My Personal Information {"props":{"pageProps":{"initialData":{"identity":"rs-8013336","acceptedTermsAndConditions":true,"allowDirectSubmit":true,"archivedVersions":[],"articleType":"Article","associatedPublications":[],"authors":[{"id":539150511,"identity":"fa9ea81b-0c5e-40f9-835d-8ed4080053f1","order_by":0,"name":"Wanbing Wang","email":"","orcid":"","institution":"Division of Pharmacotherapy and Experimental Therapeutics, Eshelman School of Pharmacy, University of North Carolina at Chapel Hill, Chapel Hill, North Carolina, USA; Department of Biostatistics, Gillings School of Global Public Health, University of North Carolina at Chapel Hill, Chapel Hill, North Carolina, USA","correspondingAuthor":false,"prefix":"","firstName":"Wanbing","middleName":"","lastName":"Wang","suffix":""},{"id":539150512,"identity":"d0b8b101-8991-483d-bb06-470667e3f640","order_by":1,"name":"Congyu Zhang","email":"","orcid":"","institution":"Division of Pharmacotherapy and Experimental Therapeutics, Eshelman School of Pharmacy, University of North Carolina at Chapel Hill, Chapel Hill, North Carolina, USA","correspondingAuthor":false,"prefix":"","firstName":"Congyu","middleName":"","lastName":"Zhang","suffix":""},{"id":539150513,"identity":"216fc812-5897-417c-a59e-fc4ee07c3e5b","order_by":2,"name":"Benyam Muluneh","email":"","orcid":"","institution":"Division of Pharmacotherapy and Experimental Therapeutics, Eshelman School of Pharmacy, University of North Carolina at Chapel Hill, Chapel Hill, North Carolina, USA; Lineberger Comprehensive Cancer Center, School of Medicine, University of North Carolina at Chapel Hill, Chapel Hill, North Carolina, USA","correspondingAuthor":false,"prefix":"","firstName":"Benyam","middleName":"","lastName":"Muluneh","suffix":""},{"id":539150514,"identity":"cbb1cccd-b0f7-4479-9161-db74ad30ff90","order_by":3,"name":"Quefeng Li","email":"","orcid":"","institution":"Department of Biostatistics, Gillings School of Global Public Health, University of North Carolina at Chapel Hill, Chapel Hill, North Carolina, USA","correspondingAuthor":false,"prefix":"","firstName":"Quefeng","middleName":"","lastName":"Li","suffix":""},{"id":539150515,"identity":"d5de6573-b18d-4a63-9d54-2739f219b6f2","order_by":4,"name":"Jim H. Hughes","email":"","orcid":"","institution":"Pfizer Inc., USA","correspondingAuthor":false,"prefix":"","firstName":"Jim","middleName":"H.","lastName":"Hughes","suffix":""},{"id":539150516,"identity":"c6a90b45-d7cc-476d-88f3-24e5e05dfd4c","order_by":5,"name":"Lynne I. Wagner","email":"","orcid":"","institution":"Department of Health Policy and Management, Gillings School of Global Public Health, University of North Carolina at Chapel Hill, Chapel Hill, North Carolina, USA","correspondingAuthor":false,"prefix":"","firstName":"Lynne","middleName":"I.","lastName":"Wagner","suffix":""},{"id":539150517,"identity":"630fe4d1-f850-46a4-b072-db7f7ccdfca4","order_by":6,"name":"William A. Wood","email":"","orcid":"","institution":"Lineberger Comprehensive Cancer Center, School of Medicine, University of North Carolina at Chapel Hill, Chapel Hill, North Carolina, USA","correspondingAuthor":false,"prefix":"","firstName":"William","middleName":"A.","lastName":"Wood","suffix":""},{"id":539150518,"identity":"e2a0500b-78ed-4853-98d0-8b97b3f39ba9","order_by":7,"name":"Ethan Basch","email":"","orcid":"","institution":"Lineberger Comprehensive Cancer Center, School of Medicine, University of North Carolina at Chapel Hill, Chapel Hill, North Carolina, USA","correspondingAuthor":false,"prefix":"","firstName":"Ethan","middleName":"","lastName":"Basch","suffix":""},{"id":539150510,"identity":"a789e579-dc25-4001-9345-d411d49f6e2e","order_by":8,"name":"Jiawei Zhou","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAAAqklEQVRIiWNgGAWjYDACCTC2gXB4SNCSRqoWBobDJGiRn918TMKy7XzidvYDjA/ethGhhXHOsTQJybbbiTt7EpgN5xKjhVkix0xCctvtxA03GNikeYnRwgbRcg6khf03UVp4IFoOgG1hJkqLhERasoXkv2TjDWcSmyXnnCNCi/yM5IO3Jc7YyW44fvjghzdlRGgBAWZI3DA2EKkepPYD8WpHwSgYBaNgJAIAfl4yJ9C4AFAAAAAASUVORK5CYII=","orcid":"https://orcid.org/0009-0000-3317-0311","institution":"Division of Pharmacotherapy and Experimental Therapeutics, Eshelman School of Pharmacy, University of North Carolina at Chapel Hill, Chapel Hill, North Carolina, USA; Lineberger Comprehensive Cancer Center, School of Medicine, University of North Carolina at Chapel Hill, Chapel Hill, North Carolina, USA","correspondingAuthor":true,"prefix":"","firstName":"Jiawei","middleName":"","lastName":"Zhou","suffix":""}],"badges":[],"createdAt":"2025-11-02 22:15:11","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-8013336/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-8013336/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":95808537,"identity":"68bb48b0-3461-4174-98a5-c797f4c1e33a","added_by":"auto","created_at":"2025-11-13 08:49:30","extension":"docx","order_by":0,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":124365,"visible":true,"origin":"","legend":"","description":"","filename":"Manuscript11.2.2025.docx","url":"https://assets-eu.researchsquare.com/files/rs-8013336/v1/49281c4ef38a0f66c0342860.docx"},{"id":95808599,"identity":"51681364-9a00-417f-8d08-3801e36d36be","added_by":"auto","created_at":"2025-11-13 08:49:32","extension":"tif","order_by":1,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":5667972,"visible":true,"origin":"","legend":"","description":"","filename":"Figuer2.tif","url":"https://assets-eu.researchsquare.com/files/rs-8013336/v1/f91c6ced7f9c1dc78fa6427b.tif"},{"id":95808438,"identity":"fbf1d651-2f0e-4d86-9641-7df985477855","added_by":"auto","created_at":"2025-11-13 08:49:28","extension":"tif","order_by":2,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":4496116,"visible":true,"origin":"","legend":"","description":"","filename":"Figure1.tif","url":"https://assets-eu.researchsquare.com/files/rs-8013336/v1/43301ef9c7b37b1dde8ef602.tif"},{"id":95808872,"identity":"757f9576-4d7f-4755-9aad-ebfedd274561","added_by":"auto","created_at":"2025-11-13 08:49:41","extension":"docx","order_by":3,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":19853,"visible":true,"origin":"","legend":"","description":"","filename":"Table1.docx","url":"https://assets-eu.researchsquare.com/files/rs-8013336/v1/33c26d88086c26f0c7b261b0.docx"},{"id":95808263,"identity":"65a26ac8-e0b5-4572-b401-ca83e24016bf","added_by":"auto","created_at":"2025-11-13 08:49:24","extension":"tif","order_by":4,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":10545664,"visible":true,"origin":"","legend":"","description":"","filename":"Figure3.tif","url":"https://assets-eu.researchsquare.com/files/rs-8013336/v1/f0a8ef2b0780e766bc1dd07e.tif"},{"id":95808636,"identity":"db230ad3-7a76-4b82-a5f1-41d12361050e","added_by":"auto","created_at":"2025-11-13 08:49:33","extension":"docx","order_by":5,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":18867,"visible":true,"origin":"","legend":"","description":"","filename":"Table2.docx","url":"https://assets-eu.researchsquare.com/files/rs-8013336/v1/0eeeafe0d2406316aaecf354.docx"},{"id":95808543,"identity":"334459bb-0e62-4242-9213-6c82b5c5f200","added_by":"auto","created_at":"2025-11-13 08:49:31","extension":"tif","order_by":6,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":6000256,"visible":true,"origin":"","legend":"","description":"","filename":"Figure4.tif","url":"https://assets-eu.researchsquare.com/files/rs-8013336/v1/918038960f584a88b4cfe0c8.tif"},{"id":95808329,"identity":"f3fd4c57-9a3b-499a-917a-dce0476f63b4","added_by":"auto","created_at":"2025-11-13 08:49:25","extension":"tif","order_by":7,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":1899668,"visible":true,"origin":"","legend":"","description":"","filename":"Figure5.tif","url":"https://assets-eu.researchsquare.com/files/rs-8013336/v1/f7dd1770e26daf9ca9f3b57d.tif"},{"id":95808427,"identity":"dd64132f-e84d-4bdb-b183-53a547000467","added_by":"auto","created_at":"2025-11-13 08:49:27","extension":"json","order_by":8,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":11143,"visible":true,"origin":"","legend":"","description":"","filename":"NCOMMS2588198.json","url":"https://assets-eu.researchsquare.com/files/rs-8013336/v1/0eee491a5cd927a7a181cfde.json"},{"id":95808286,"identity":"5ed69d0b-52d9-4046-ac3c-daee24e9562e","added_by":"auto","created_at":"2025-11-13 08:49:24","extension":"docx","order_by":9,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":5640475,"visible":true,"origin":"","legend":"","description":"","filename":"SupplementaryFiles11.2.2025.docx","url":"https://assets-eu.researchsquare.com/files/rs-8013336/v1/6f92db2f7d439482180c1d10.docx"},{"id":95819053,"identity":"f8958483-c445-4b01-9bfa-070dae4a75fd","added_by":"auto","created_at":"2025-11-13 10:37:46","extension":"xml","order_by":10,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":180225,"visible":true,"origin":"","legend":"","description":"","filename":"NCOMMS25881980enriched.xml","url":"https://assets-eu.researchsquare.com/files/rs-8013336/v1/4ebec7e327f7e01a76f64fba.xml"},{"id":95808485,"identity":"e1b2c232-deb2-407c-86e9-aa5b78e7f387","added_by":"auto","created_at":"2025-11-13 08:49:29","extension":"tif","order_by":11,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":5667972,"visible":true,"origin":"","legend":"","description":"","filename":"Figuer2.tif","url":"https://assets-eu.researchsquare.com/files/rs-8013336/v1/b422f1006984ace0c228a691.tif"},{"id":95808841,"identity":"8d42673b-013d-4055-931b-33a49c7a8355","added_by":"auto","created_at":"2025-11-13 08:49:38","extension":"tif","order_by":12,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":4496116,"visible":true,"origin":"","legend":"","description":"","filename":"Figure1.tif","url":"https://assets-eu.researchsquare.com/files/rs-8013336/v1/6285a66a6ee65772a284a262.tif"},{"id":95808722,"identity":"04e4f37a-1a4a-4bbc-9616-a25b91e4e1fe","added_by":"auto","created_at":"2025-11-13 08:49:34","extension":"tif","order_by":13,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":10545664,"visible":true,"origin":"","legend":"","description":"","filename":"Figure3.tif","url":"https://assets-eu.researchsquare.com/files/rs-8013336/v1/f9d21ff5139f6f6a2f977769.tif"},{"id":95808798,"identity":"aa252d61-1da4-4226-89a6-d04c008a0aaf","added_by":"auto","created_at":"2025-11-13 08:49:37","extension":"tif","order_by":14,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":6000256,"visible":true,"origin":"","legend":"","description":"","filename":"Figure4.tif","url":"https://assets-eu.researchsquare.com/files/rs-8013336/v1/0ab028d6eb64d0b5df03f8b5.tif"},{"id":95808856,"identity":"6d440496-3e09-4b88-92ee-f6a4cc5daf6c","added_by":"auto","created_at":"2025-11-13 08:49:39","extension":"tif","order_by":15,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":1899668,"visible":true,"origin":"","legend":"","description":"","filename":"Figure5.tif","url":"https://assets-eu.researchsquare.com/files/rs-8013336/v1/f452912c95742d2ecd9509de.tif"},{"id":95808633,"identity":"51603217-54db-4161-a557-0e874ed5e249","added_by":"auto","created_at":"2025-11-13 08:49:33","extension":"png","order_by":16,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":453339,"visible":true,"origin":"","legend":"","description":"","filename":"OnlineFiguer2.png","url":"https://assets-eu.researchsquare.com/files/rs-8013336/v1/b211afe5ee2ede7867894bac.png"},{"id":95808601,"identity":"081c426f-b323-4a4c-bc82-6ab1001f8a94","added_by":"auto","created_at":"2025-11-13 08:49:32","extension":"png","order_by":17,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":302064,"visible":true,"origin":"","legend":"","description":"","filename":"OnlineFigure1.png","url":"https://assets-eu.researchsquare.com/files/rs-8013336/v1/f9ecb0956076ca642237ea0c.png"},{"id":95808579,"identity":"51923fe7-4d71-4b4d-9162-447bb7e8e7ee","added_by":"auto","created_at":"2025-11-13 08:49:31","extension":"png","order_by":18,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":693718,"visible":true,"origin":"","legend":"","description":"","filename":"OnlineFigure3.png","url":"https://assets-eu.researchsquare.com/files/rs-8013336/v1/1db15893db765f7eeb65fc3f.png"},{"id":95808496,"identity":"0807c8e8-1624-47b6-81ae-7b4b334576cd","added_by":"auto","created_at":"2025-11-13 08:49:29","extension":"png","order_by":19,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":502618,"visible":true,"origin":"","legend":"","description":"","filename":"OnlineFigure4.png","url":"https://assets-eu.researchsquare.com/files/rs-8013336/v1/508d78fe7832eaf0a1c05570.png"},{"id":95808545,"identity":"effd168e-9620-401d-9476-f6d365d1fa99","added_by":"auto","created_at":"2025-11-13 08:49:31","extension":"png","order_by":20,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":181501,"visible":true,"origin":"","legend":"","description":"","filename":"OnlineFigure5.png","url":"https://assets-eu.researchsquare.com/files/rs-8013336/v1/f28d845edfa8c31c5c34e27b.png"},{"id":95808493,"identity":"38a0c4e0-9bcd-4237-9bc8-a02b350a1020","added_by":"auto","created_at":"2025-11-13 08:49:29","extension":"xml","order_by":21,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":177623,"visible":true,"origin":"","legend":"","description":"","filename":"NCOMMS25881980structuring.xml","url":"https://assets-eu.researchsquare.com/files/rs-8013336/v1/cc19e0e0e9e5fb37dc8d2c57.xml"},{"id":95808614,"identity":"70b3c91c-2b32-412a-898e-05cfed54c5f4","added_by":"auto","created_at":"2025-11-13 08:49:32","extension":"html","order_by":22,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":185815,"visible":true,"origin":"","legend":"","description":"","filename":"earlyproof.html","url":"https://assets-eu.researchsquare.com/files/rs-8013336/v1/ac06d3bbdaa880de88404e00.html"},{"id":95808892,"identity":"2abfe244-5e80-4bf7-94fc-9ee6bfda7420","added_by":"auto","created_at":"2025-11-13 08:49:42","extension":"png","order_by":1,"title":"Figure 1","display":"","copyAsset":false,"role":"figure","size":506109,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003ePatient-reported outcomes (PRO) deterioration precedes radiographically confirmed breast cancer disease progression.\u003c/strong\u003e Sankey plot illustrates the sequence of PRO deterioration and subsequent radiographically confirmed breast cancer relapses\u003cstrong\u003e \u003c/strong\u003ewithin individual patients. PROs positioned on the left represent earlier PRO deterioration. Among 892 patients who experienced disease progression, 808 (90.6%) had at least one PRO deterioration prior to radiographically confirmed cancer relapse.\u003c/p\u003e","description":"","filename":"Figure1.png","url":"https://assets-eu.researchsquare.com/files/rs-8013336/v1/50a16c3fba47e55a7261d6ea.png"},{"id":95808595,"identity":"214327b4-7a17-406e-9016-f9e1a64f3d06","added_by":"auto","created_at":"2025-11-13 08:49:32","extension":"png","order_by":2,"title":"Figure 2","display":"","copyAsset":false,"role":"figure","size":513578,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eMetastatic sites are associated with patient-reported outcome (PRO) deterioration. \u003c/strong\u003e(\u003cstrong\u003ea\u003c/strong\u003e) Heatmap showing the hazard ratios (HRs) from Cox proportional hazards models evaluating the association between metastases in specific organs and time to first deterioration across different PROs. Deeper red tones indicate stronger associations, meaning patients with metastases in that organ were more likely to experience earlier PRO deterioration. (\u003cstrong\u003eb–e\u003c/strong\u003e) Forest plots showing HRs for different metastatic organs across four PRO functioning scales. Circles indicate hazard ratios, with bars representing 95% confidence intervals. HR \u0026gt; 1 indicates that patients with metastases in that organ had a higher risk of earlier PRO deterioration compared to those without. Data in this analysis are patients with available metastatic sites information from the CA012-0 (N = 219) and EFC6089 (N = 213) studies.\u003c/p\u003e","description":"","filename":"Figuer2.png","url":"https://assets-eu.researchsquare.com/files/rs-8013336/v1/0cf3a0f78d64b236ada77ecf.png"},{"id":95808514,"identity":"f062b269-10c6-4f31-98da-10bbb5a9095d","added_by":"auto","created_at":"2025-11-13 08:49:30","extension":"png","order_by":3,"title":"Figure 3","display":"","copyAsset":false,"role":"figure","size":880782,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eLarger tumor size is associated with earlier patient-reported outcome (PRO) appetite loss. \u003c/strong\u003e(\u003cstrong\u003ea\u003c/strong\u003e) Hazard ratios (HRs) with 95% confidence intervals (CIs) from Cox proportional hazards models evaluating how tumor size correlates with PRO deterioration time. Higher HRs indicate a stronger association between larger tumors and earlier PRO deterioration, with appetite loss showing the strongest effect. (\u003cstrong\u003eb–c\u003c/strong\u003e) Boxplots comparing tumor regrowth ratios, defined as the amount of tumor growth after its smallest size (nadir), between patients who experienced PRO deterioration and those censored for appetite loss (\u003cstrong\u003eb\u003c/strong\u003e) or pain (\u003cstrong\u003ec\u003c/strong\u003e). The central line within each box represents the median; box edges are the 25\u003csup\u003eth\u003c/sup\u003e and 75\u003csup\u003eth\u003c/sup\u003e percentiles; and whiskers extend to the most extreme values within 1.5 times the interquartile range (IQR). Points beyond the whiskers indicate outliers. (\u003cstrong\u003ed-e\u003c/strong\u003e) Kaplan–Meier curves for time to PRO deterioration in appetite loss (\u003cstrong\u003ed\u003c/strong\u003e) and pain (\u003cstrong\u003ee\u003c/strong\u003e), stratified tumor regrowth ratio (\u0026lt; 1.2 vs. ≥ 1.2). Shaded areas represent 95% CIs; p-values are derived from Wilcoxon tests (\u003cstrong\u003eb-c\u003c/strong\u003e) or log-rank tests (\u003cstrong\u003ea, d, e\u003c/strong\u003e). Data in this analysis are patients with available tumor size data from the CA012-0 (N = 161) and EFC6089 (N = 198) studies.\u003c/p\u003e","description":"","filename":"Figure3.png","url":"https://assets-eu.researchsquare.com/files/rs-8013336/v1/aece9c848c952d4f9ac0006d.png"},{"id":95808589,"identity":"e1fd0503-d106-4012-af3b-a1195208b30f","added_by":"auto","created_at":"2025-11-13 08:49:31","extension":"png","order_by":4,"title":"Figure 4","display":"","copyAsset":false,"role":"figure","size":523484,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003ePatient-reported outcome (PRO) deterioration is associated with patient survival.\u003c/strong\u003e Hazard ratios (HRs) with 95% confidence intervals (CIs) for PRO deterioration time across 15 subscales, from highest to lowest, based on time-dependent Cox proportional hazards models for overall survival (\u003cstrong\u003ea\u003c/strong\u003e) or progression-free survival (\u003cstrong\u003eb\u003c/strong\u003e). HRs \u0026lt; 1 indicate that patients who experienced earlier PRO deterioration tended to have shorter survival or faster disease progression. Patients from all four studies with available OS data (N = 1,210) and PFS data (N = 1,412) were included in this analysis.\u003c/p\u003e","description":"","filename":"Figure4.png","url":"https://assets-eu.researchsquare.com/files/rs-8013336/v1/718703be56af702c173dfb7e.png"},{"id":95808801,"identity":"ed572d78-7a47-40c6-9e74-ae22d9b79401","added_by":"auto","created_at":"2025-11-13 08:49:37","extension":"png","order_by":5,"title":"Figure 5","display":"","copyAsset":false,"role":"figure","size":202070,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eIntegrating patient-reported outcomes (PROs) with progression-free survival (PFS) improves overall survival (OS) prediction using machine learning. \u003c/strong\u003e(a) Receiver-operating characteristic (ROC) curves for gradient boosting models predicting OS using PRO features only, PFS only, or a combination of PRO and PFS features. The combination features achieved the best predictive performance with the highest area under curve (AUC). (b) ROC curves for gradient boosting models predicting OS after the 365-day landmark. To avoid information leakage, only PFS data available up to day 365 were used, providing an unbiased estimate of post-landmark predictive performance. AUC-ROC quantifies model accuracy in distinguishing between patients who survived and those who died (1.0 = perfect prediction; 0.8 = good performance; 0.5 = random chance, as red dashed lines). Patients from all four studies with available survival status (N = 1,337) were included in the analysis.\u003c/p\u003e","description":"","filename":"Figure5.png","url":"https://assets-eu.researchsquare.com/files/rs-8013336/v1/7328671f49e0bc8789b7b728.png"},{"id":96602825,"identity":"7eeda04a-ac30-4811-8a7f-1826243e627c","added_by":"auto","created_at":"2025-11-24 09:02:32","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":4583649,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-8013336/v1/3d976aa5-6d76-4487-9ec6-5732748963f7.pdf"},{"id":95808869,"identity":"6b027c7f-8614-42ce-93b6-63edadf48ef4","added_by":"auto","created_at":"2025-11-13 08:49:41","extension":"docx","order_by":1,"title":"","display":"","copyAsset":false,"role":"supplement","size":5640475,"visible":true,"origin":"","legend":"\u003cp\u003eSupplementary Files\u003c/p\u003e","description":"","filename":"SupplementaryFiles11.2.2025.docx","url":"https://assets-eu.researchsquare.com/files/rs-8013336/v1/bd0cd28bb11b096f61d7e123.docx"}],"financialInterests":"\u003cb\u003eYes\u003c/b\u003e there is potential Competing Interest.\nB.M may own stock in Novartis. J.H.H is a current employee of Pfizer and may own stock in Pfizer. L.I.W W receives institutional research funding from the National Cancer Institute and previously received personal fees from Celgene/Bristol Myers Squibb as a member of the Scientific Steering Committee for the Connect Multiple Myeloma patient registry. W.A.W. has received institutional research funding from Genentech; receives consulting fees from Teladoc Health, Quantum Health, and Lantern Health; holds equity in Koneksa Health; and has a leadership position in the American Society of Hematology Research Collaborative. E.B. receives institutional research funding from the National Cancer Institute and from the Patient-Centered Outcomes Research Institute, and personal fees for scientific advising from Research Triangle Institute, Thyme Care, N-Power Medicine, Resilience Health, Canopy Care, Savor Health, and Navigating Cancer.","formattedTitle":"Patient-Reported Outcomes as Early Indicators of Disease Progression and Predictors of Survival via Machine Learning in Breast Cancer","fulltext":[{"header":"Introduction","content":"\u003cp\u003eTimely detection and monitoring of disease progression is central to effective cancer management. In breast cancer, radiographic imaging is the current standard for disease monitoring.\u003csup\u003e\u003cspan citationid=\"CR1\" class=\"CitationRef\"\u003e1\u003c/span\u003e\u003c/sup\u003e However, imaging alone does not capture patients\u0026rsquo; psychological, functional, or quality-of-life status. Frequent imaging for cancer surveillance also imposes financial and logistical burdens on both patients and healthcare systems\u003csup\u003e\u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2\u003c/span\u003e\u003c/sup\u003e, while reducing the frequency risks delayed detection and treatment. These challenges highlight the need for complementary, patient-centered tools that can support efficient and continuous disease monitoring.\u003c/p\u003e\u003cp\u003ePatient-reported outcomes (PROs) provide a non-invasive, low-cost means of capturing patients\u0026rsquo; perspectives on symptoms, functioning, and quality of life.\u003csup\u003e\u003cspan citationid=\"CR3\" class=\"CitationRef\"\u003e3\u003c/span\u003e\u003c/sup\u003e They can be collected frequently and offer real-time insights into a patient\u0026rsquo;s evolving health status, potentially revealing early signs of disease progression or treatment failure.\u003csup\u003e\u003cspan citationid=\"CR4\" class=\"CitationRef\"\u003e4\u003c/span\u003e, \u003cspan citationid=\"CR5\" class=\"CitationRef\"\u003e5\u003c/span\u003e\u003c/sup\u003e Over the past decade, PROs have gained increasing recognition from patients, clinicians, and regulators as essential components of patient-centered drug development and value-based healthcare.\u003csup\u003e\u003cspan citationid=\"CR6\" class=\"CitationRef\"\u003e6\u003c/span\u003e, \u003cspan citationid=\"CR7\" class=\"CitationRef\"\u003e7\u003c/span\u003e\u003c/sup\u003e\u003c/p\u003e\u003cp\u003eDespite these advantages, the role of PROs as objective predictors of tumor burden, disease progression, and survival remains poorly understood. Previous studies have yielded inconsistent findings: a meta-analysis of 45 Phase III randomized controlled trials reported no association between PRO improvements and progression-free survival (PFS) benefits,\u003csup\u003e8\u003c/sup\u003e whereas individual participant-level analyses in breast and lung cancer suggested that PROs are prognostic for PFS.\u003csup\u003e\u003cspan citationid=\"CR9\" class=\"CitationRef\"\u003e9\u003c/span\u003e, \u003cspan citationid=\"CR10\" class=\"CitationRef\"\u003e10\u003c/span\u003e\u003c/sup\u003e In addition, few studies have quantitatively examined how PROs relate to tumor burden or volume. These gaps make PROs difficult to interpret as objective biomarkers and perpetuate their perception as \u0026ldquo;soft\u0026rdquo; measures, limiting their use in clinical decision-making and drug development.\u003csup\u003e\u003cspan citationid=\"CR11\" class=\"CitationRef\"\u003e11\u003c/span\u003e, \u003cspan citationid=\"CR12\" class=\"CitationRef\"\u003e12\u003c/span\u003e\u003c/sup\u003e\u003c/p\u003e\u003cp\u003eTo address these gaps, we retrospectively analyzed 450,147 PRO entries collected using EORTC QLQ-C30 questionnaires from 2,738 breast cancer patients pooled from four clinical trials, encompassing both early- and late-stage disease. The EORTC QLQ-C30 questionnaire assessed patient-reported symptoms, functional status, and quality of life across 15 subscales. We compared the timing of PRO deterioration of all subscales with radiographically confirmed disease progression, and assessed their associations with metastatic sites, tumor burden, and survival outcomes. In addition, we applied machine learning models to evaluate the predictive value of PROs for overall survival (OS) and examined whether integrating PROs with PFS enhances predictive performance.\u003c/p\u003e\u003cp\u003eOur findings show that PRO deterioration often precedes radiographic relapse and is strongly associated with disease burden and survival. Functional PROs were more affected by metastatic sites, whereas symptomatic PROs showed stronger associations with tumor burden and survival outcomes. Incorporating PROs with PFS can improve the OS prediction via machine learning models, indicating that PROs can serve as early, quantitative, and patient-centered indicators for disease monitoring and outcome prediction in breast cancer.\u003c/p\u003e"},{"header":"Results","content":"\u003cdiv id=\"Sec3\" class=\"Section2\"\u003e\u003ch2\u003eDataset Integration and Calculation of PRO Deterioration Times\u003c/h2\u003e\u003cp\u003eLongitudinal PRO data, assessed using the EORTC QLQ-C30 instrument\u003csup\u003e\u003cspan citationid=\"CR13\" class=\"CitationRef\"\u003e13\u003c/span\u003e\u003c/sup\u003e, were pooled together with demographics, clinical characteristics, disease progression, and survival data (PFS and OS) from four breast cancer Phase III clinical trials (BCIRG-001, BCIRG-005, CA012-0, and EFC6089). Longitudinal tumor size and metastatic sites information from the CA012-0 and EFC6089 studies were also collected. Patients without available PRO data were excluded from the analysis. The workflow of data inclusion and exclusion is shown in \u003cb\u003eSupplementary Fig.\u0026nbsp;1\u003c/b\u003e.\u003c/p\u003e\u003cp\u003eA total of 2,738 patients with available PRO data were included in the final analysis cohort. Patient demographics and clinical characteristics across the four studies are summarized in Table\u0026nbsp;\u003cspan refid=\"Tab1\" class=\"InternalRef\"\u003e1\u003c/span\u003e. In total, 52,190 PRO questionnaires were collected over time. Some questionnaires contained incomplete responses to the 30 items in EORTC QLQ-C30, resulting in data missingness. Overall, a total of 450,147 available PRO responses were included in our analysis and missingness was less than 5% across the dataset and remained relatively stable over time (\u003cb\u003eSupplementary Fig.\u0026nbsp;2\u003c/b\u003e); missingness in each questionnaire item is summarized in \u003cb\u003eSupplementary Table\u0026nbsp;1\u003c/b\u003e. Missing data were excluded from the analysis. For each patient, longitudinal responses to all items of the EORTC QLQ-C30 questionnaire were transformed to raw scores at the subscale level to estimate the time to first deterioration within each subscale (hereafter referred to as the PRO deterioration time; see \u003cb\u003eMethods\u003c/b\u003e).\u003c/p\u003e\u003cp\u003e\u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab1\" border=\"1\"\u003e\u003ccaption language=\"En\"\u003e\u003cdiv class=\"CaptionNumber\"\u003eTable 1\u003c/div\u003e\u003cdiv class=\"CaptionContent\"\u003e\u003cp\u003eStudy demographics.\u003c/p\u003e\u003c/div\u003e\u003c/caption\u003e\u003ccolgroup cols=\"6\"\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c4\" colnum=\"4\"\u003e\u003c/div\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c5\" colnum=\"5\"\u003e\u003c/div\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c6\" colnum=\"6\"\u003e\u003c/div\u003e\u003cthead\u003e\u003ctr\u003e\u003cth align=\"left\" colname=\"c1\"\u003e\u003cp\u003eStudy\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c2\"\u003e\u003cp\u003eBCIRG-001\u003c/p\u003e\u003cp\u003e(N\u0026thinsp;=\u0026thinsp;735)\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c3\"\u003e\u003cp\u003eBCIRG-005\u003c/p\u003e\u003cp\u003e(N\u0026thinsp;=\u0026thinsp;1,564)\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c4\"\u003e\u003cp\u003eCA012-0\u003c/p\u003e\u003cp\u003e(N\u0026thinsp;=\u0026thinsp;226)\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c5\"\u003e\u003cp\u003eEFC6089\u003c/p\u003e\u003cp\u003e(N\u0026thinsp;=\u0026thinsp;213)\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c6\"\u003e\u003cp\u003eTotal\u003c/p\u003e\u003cp\u003e(N\u0026thinsp;=\u0026thinsp;2,738)\u003c/p\u003e\u003c/th\u003e\u003c/tr\u003e\u003c/thead\u003e\u003ctbody\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003e\u003cb\u003eNCT\u003c/b\u003e\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eNCT00688740\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003eNCT00312208\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003eNCT00046527\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003eNCT00081796\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u0026nbsp;\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003e\u003cb\u003eTreatment\u003c/b\u003e\u003csup\u003e\u003cb\u003e#\u003c/b\u003e\u003c/sup\u003e\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eFAC\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003eAC\u0026rarr;T\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003ePaclitaxel\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003ecapecitabine\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u0026nbsp;\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003e\u003cb\u003eAge, years\u003c/b\u003e\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e49.0 (23.0, 70.0)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e50.0 (22.0, 74.0)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e52.0 (30.0, 83.0)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e52.0 (30.0, 76.0)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e50.0 (22.0, 83.0)\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003e\u003cb\u003eWeight, Kg\u003c/b\u003e\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e68.0 (43.0, 163.6)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e68.5 (38.5, 125.0)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e68.0 (40.0, 106.2)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e66.1 (37.0, 122.9)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e68.0 (37.0, 163.6)\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003e\u003cb\u003eStage, N (%)\u003c/b\u003e\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u0026nbsp;\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eII\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e377 (51.3%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e1557 (99.6%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e0\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e0\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e1934 (70.6%)\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eIII\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e44 (6.0%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e3 (0.2%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e0\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e6 (2.8%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e53 (1.9%)\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eIV\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e314 (42.7%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e3 (0.2%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e226 (100.0%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e207 (97.2%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e750 (27.4%)\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eMissing\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e0\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e1 (0.1%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e0\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e0\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e1 (\u0026lt;\u0026thinsp;0.1%)\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003e\u003cb\u003eRace, N (%)\u003c/b\u003e\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u0026nbsp;\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eWhite\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e642 (87.3%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e0\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e219 (96.9%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e160 (75.1%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e1021 (37.3%)\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eBlack of African American\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e11 (1.5%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e0\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e5 (2.2%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e0\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e16 (0.6%)\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eAsian/Pacific Islander\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e31 (4.2%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e0\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e0\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e0\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e31 (1.1%)\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eOther/Missing\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e51 (6.9%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e1564 (100.0%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e2\u0026nbsp;(0.9%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e53(24.9%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e1670 (61.0%)\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003e\u003cb\u003eMenopause, N (%)\u003c/b\u003e\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u0026nbsp;\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003ePostmenopausal\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e249 (33.9%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e0\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e37 (16.4%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e0\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e286 (10.4%)\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003ePremenopausal\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e352 (47.9%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e702 (44.9%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e38 (16.8%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e55 (25.8%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e1,147 (41.9%)\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eOther/Missing\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e134 (18.2%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e862 (55.1%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e151 (66.8%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e158 (74.2%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e1305 (47.7%)\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003e\u003cb\u003eECOG status, N (%)\u003c/b\u003e\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u0026nbsp;\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003e0\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e565 (76.9%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e1529 (97.8%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e84 (37.2%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e102 (47.9%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e2280 (83.3%)\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003e1\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e170 (23.1%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e35 (2.2%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e137 (60.6%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e89 (41.8%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e431 (15.7%)\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003e2\u0026ndash;3\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e0\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e0\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e5 (2.2%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e20 (9.4%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e25 (0.9%)\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eMissing\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e0\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e0\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e0\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e2 (0.9%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e2 (0.1%)\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003e\u003cb\u003eEstrogen receptors\u003c/b\u003e\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u0026nbsp;\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003ePositive\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e348 (47.3%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e1003 (64.1%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e24 (10.6%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e86 (40.4%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e1461 (53.4%)\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eNegative\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e254 (34.6%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e437 (27.9%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e51 (22.6%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e100 (46.9%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e842 (30.8%)\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eUnknown/Missing\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e133 (18.1%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e124 (7.9%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e151 (66.8%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e27 (12.7%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e435 (15.9%)\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003e\u003cb\u003eProgesterone receptors\u003c/b\u003e\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u0026nbsp;\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003ePositive\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e354\u0026nbsp;(47.5%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e1056\u0026nbsp;(64.0%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e24\u0026nbsp;(10.6%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e87\u0026nbsp;(40.1%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e1521 (55.6%)\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eNegative\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e257\u0026nbsp;(34.5%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e467\u0026nbsp;(28.3%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e52\u0026nbsp;(22.9%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e101\u0026nbsp;(46.5%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e877 (32.0%)\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eUnknown/Missing\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e135\u0026nbsp;(18.1%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e126\u0026nbsp;(7.6%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e151\u0026nbsp;(66.5%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e29\u0026nbsp;(13.4%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e441 (16.1%)\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003e\u003cb\u003eOverall survival, days*\u003c/b\u003e\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e3706 (2333, 3767)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e1705 (1010, 2693)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e216 (118, 277)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e287 (168, 378)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e2675 (893, 3735)\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003e\u003cb\u003eProgression-free survival, days*\u003c/b\u003e\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e3353 (1136, 3746)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e1245 (674, 2052)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e84 (42, 155)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e99 (43, 173)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e1338 (342, 3544)\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003c/tbody\u003e\u003c/colgroup\u003e\u003ctfoot\u003e\u003ctr\u003e\u003ctd colspan=\"6\"\u003eCategorical variables are summarized as number of patients and percentages within the study; continuous variables as medians with ranges (minimum, maximum).\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd colspan=\"6\"\u003e\u003csup\u003e#\u003c/sup\u003eFAC\u0026thinsp;=\u0026thinsp;Fluorouracil, Doxorubicin, and Cyclophosphamide; AC\u0026rarr;T\u0026thinsp;=\u0026thinsp;Doxorubicin\u0026thinsp;+\u0026thinsp;Cyclophosphamide followed by a Taxane.\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd colspan=\"6\"\u003e*Overall survival and progression-free survival are reported as medians with interquartile ranges (IQR) among patients who experienced the events.\u003c/td\u003e\u003c/tr\u003e\u003c/tfoot\u003e\u003c/table\u003e\u003c/div\u003e\u003c/p\u003e\u003cp\u003eWe also evaluated the baseline demographics and clinical characteristics that are associated with patient OS and PFS using Least Absolute Shrinkage and Selection Operator (LASSO). For OS, treatment, employment status, baseline number of metastatic lesions, ECOG performance status, and HER2 receptor status are significant covariates. For PFS, disease stage, age, treatment, menopausal status, ECOG performance status, estrogen receptor status, and HER2 receptor status are significant covariates. Forest plots were depicted to compare the contributions of each selected covariate for OS and PFS were in \u003cb\u003eSupplementary Fig.\u0026nbsp;3a-b\u003c/b\u003e.\u003c/p\u003e\u003cp\u003e\u003cb\u003ePRO Deterioration Precedes Radiographically Confirmed Breast Cancer Relapse.\u003c/b\u003e\u003c/p\u003e\u003cp\u003eWe examined the temporal relationship between PRO deterioration and radiographically confirmed breast cancer relapse for all four studies, including both early- and late-stage breast cancer patients. Among the 892 patients who experienced relapse during follow-up period, 90.6% (N\u0026thinsp;=\u0026thinsp;808) reported at least one PRO deterioration before relapse confirmation, indicating that PRO deterioration often precedes imaging-detected disease progression. The median and interquartile range (IQR) of PRO deterioration times across 15 subscales, including multiple symptoms, functioning, and quality of life domains, are summarized in Table\u0026nbsp;\u003cspan refid=\"Tab2\" class=\"InternalRef\"\u003e2\u003c/span\u003e, together with the proportion of patients who did not experience PRO deterioration during the study period. The median time to deterioration across all PRO subscales was 85 days, compared with a median time to radiographic confirmed relapse of 1,338 days. The sequence of deterioration across 15 PRO subscales was visualized using a Sankey plot, illustrating the most common temporal order in which symptoms and functional domains worsened across patients. (Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003e) Appetite loss was among the earliest symptoms to deteriorate in patients who subsequently relapsed. We also analyzed the sequence of PRO deterioration specifically among early-stage breast cancer patients. Of the 684 patients who experienced recurrence in the BCIRG-001 and BCIRG-005 studies, 628 (91.8%) had at least one PRO deterioration prior to recurrence. These findings suggest that longitudinal PRO monitoring can provide early, patient-centered signals of disease progression and may serve as a valuable complementary tool to radiographic surveillance in breast cancer.\u003c/p\u003e\u003cp\u003e\u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab2\" border=\"1\"\u003e\u003ccaption language=\"En\"\u003e\u003cdiv class=\"CaptionNumber\"\u003eTable 2\u003c/div\u003e\u003cdiv class=\"CaptionContent\"\u003e\u003cp\u003eTime to first deterioration and proportion of censored patients across different patient-reported outcomes (PROs).\u003c/p\u003e\u003c/div\u003e\u003c/caption\u003e\u003ccolgroup cols=\"6\"\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c4\" colnum=\"4\"\u003e\u003c/div\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c5\" colnum=\"5\"\u003e\u003c/div\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c6\" colnum=\"6\"\u003e\u003c/div\u003e\u003cthead\u003e\u003ctr\u003e\u003cth align=\"left\" colname=\"c1\"\u003e\u003cp\u003eTime to first deterioration\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c2\"\u003e\u003cp\u003eBCIRG-001\u003c/p\u003e\u003cp\u003e(N\u0026thinsp;=\u0026thinsp;735)\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c3\"\u003e\u003cp\u003eBCIRG-005\u003c/p\u003e\u003cp\u003e(N\u0026thinsp;=\u0026thinsp;1,564)\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c4\"\u003e\u003cp\u003eCA012-0\u003c/p\u003e\u003cp\u003e(N\u0026thinsp;=\u0026thinsp;226)\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c5\"\u003e\u003cp\u003eEFC6089\u003c/p\u003e\u003cp\u003e(N\u0026thinsp;=\u0026thinsp;213)\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c6\"\u003e\u003cp\u003eTotal\u003c/p\u003e\u003cp\u003e(N\u0026thinsp;=\u0026thinsp;2,738)\u003c/p\u003e\u003c/th\u003e\u003c/tr\u003e\u003c/thead\u003e\u003ctbody\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003e\u003cb\u003eAppetite loss, days\u003c/b\u003e\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e148 (51, 830)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e176 (71, 856)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e74 (48, 137)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e57 (24, 128)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e140 (66, 551)\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eCensored, N (%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e391 (53.2%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e806 (51.5%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e146 (64.6%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e120 (56.3%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e1,463 (53.4%)\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003e\u003cb\u003eConstipation, days\u003c/b\u003e\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e310 (86, 792)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e282 (73, 884)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e96 (50, 139)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e71 (31, 150)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e170 (69, 580)\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eCensored, N (%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e404 (55.0%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e848 (54.2%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e179 (79.2%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e136 (63.8%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e1,567 (57.2%)\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003e\u003cb\u003eDiarrhea, days\u003c/b\u003e\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e492 (106, 858)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e378 (130, 903)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e106 (58, 141)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e50 (25, 112)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e230 (93, 858)\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eCensored, N (%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e515 (70.1%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e978 (62.5%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e187 (82.7%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e112 (52.6%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e1,792 (65.4%)\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003e\u003cb\u003eDyspnea, days\u003c/b\u003e\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e312 (90, 746)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e179 (110, 586)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e72 (47, 136)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e85 (28, 158)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e170 (81, 545)\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eCensored, N (%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e405 (55.1%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e706 (45.1%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e141 (62.4%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e135 (63.4%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e1,387 (50.7%)\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003e\u003cb\u003eFatigue, days\u003c/b\u003e\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e87 (47, 316)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e112 (66, 189)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e50 (43, 112)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e42 (23, 93)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e86 (49, 177)\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eCensored, N (%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e223 (30.3%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e392 (25.1%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e95 (42.0%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e81 (38.0%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e791 (28.9%)\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003e\u003cb\u003eFinancial, days\u003c/b\u003e\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e380 (92, 854)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e386 (128, 903)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e98 (50, 140)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e88 (43, 172)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e246 (88, 854)\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eCensored, N (%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e494 (67.3%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e965 (61.7%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e170 (75.2%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e169 (79.3%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e1,798 (65.7%)\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003e\u003cb\u003eNausea/Vomit, days\u003c/b\u003e\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e95 (48, 505)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e140 (67, 577)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e92 (52, 139)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e44 (23, 113)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e109 (61, 523)\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eCensored, N (%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e298 (40.6%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e728 (46.5%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e160 (70.8%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e105 (49.3%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e1,291 (47.2%)\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003e\u003cb\u003ePain, days\u003c/b\u003e\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e315 (88, 828)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e175 (96, 555)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e68 (44, 134)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e60 (23, 124)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e151 (71, 538)\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eCensored, N (%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e406 (55.3%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e666 (42.6%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e124 (54.9%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e105 (49.3%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e1,301 (47.6%)\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003e\u003cb\u003eInsomnia, days\u003c/b\u003e\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e176 (86, 615)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e182 (92, 642)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e77 (47, 136)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e71 (27, 148)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e154 (71, 540)\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eCensored, N (%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e367 (50.0%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e717 (45.9%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e149 (65.9%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e139 (65.6%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e1,372 (50.2%)\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003e\u003cb\u003eGlobal Health Status/QOL, days\u003c/b\u003e\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e134 (53, 534)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e150 (69, 545)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e72 (45, 135)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e49 (24, 113)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e132 (65, 504)\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eCensored, N (%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e344 (46.9%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e632 (40.5%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e126 (55.8%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e109 (51.2%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e1,211 (44.3%)\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003e\u003cb\u003ePhysical functioning, days\u003c/b\u003e\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e486 (129, 860)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e544 (196, 908)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e94 (54, 140)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e71 (38, 157)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e382 (128, 877)\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eCensored, N (%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e538 (73.3%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e1,196 (76.7%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e183 (81.0%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e152 (71.7%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e2,069 (75.7%)\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003e\u003cb\u003eRole functioning, days\u003c/b\u003e\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e298 (53, 712)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e358 (85, 891)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e73 (46, 135)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e71 (27, 143)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e182 (69, 570)\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eCensored, N (%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e405 (55.2%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e847 (54.2%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e152 (67.3%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e145 (68.1%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e1,549 (56.6%)\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003e\u003cb\u003eCognitive functioning, days\u003c/b\u003e\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e155 (51, 531)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e353 (118, 855)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e72 (45, 136)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e49 (23, 114)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e176 (69, 547)\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eCensored, N (%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e494 (67.3%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e1,065 (68.3%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e162 (71.7%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e138 (65.1%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e1,859 (68%)\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003e\u003cb\u003eEmotional functioning, days\u003c/b\u003e\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e420 (98, 852)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e535 (145, 906)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e76 (46, 136)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e52 (24, 134)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e349 (87, 862)\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eCensored, N (%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e350 (47.7%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e802 (51.4%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e157 (69.5%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e120 (56.9%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e1,429 (52.3%)\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003e\u003cb\u003eSocial functioning, days\u003c/b\u003e\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e322 (86, 828)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e365 (130, 888)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e89 (47, 139)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e57 (24, 129)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e266 (76, 586)\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eCensored, N (%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e413 (57.2%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e858 (54.9%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e168 (74.7%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e133 (63.0%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e1,572 (57.8%)\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003c/tbody\u003e\u003c/colgroup\u003e\u003ctfoot\u003e\u003ctr\u003e\u003ctd colspan=\"6\"\u003eTime to first deterioration were summarized by median with interquartile range (IQR).\u003c/td\u003e\u003c/tr\u003e\u003c/tfoot\u003e\u003c/table\u003e\u003c/div\u003e\u003c/p\u003e\u003cp\u003e\u003c/p\u003e\u003cp\u003eIn this cohort, 113 patients who showed no deterioration across all PRO subscales during the study were classified as \u0026ldquo;non-symptomatic\u0026rdquo; breast cancer patients. Notably, 41.6% (N\u0026thinsp;=\u0026thinsp;47) of these patients experienced disease relapse, indicating that the absence of PRO deterioration does not necessarily reflect the absence of disease progression. Radiographic assessments therefore remain essential in breast cancer disease monitoring.\u003c/p\u003e\u003cp\u003e\u003cb\u003eMetastatic Sites Involvement is Associated with PRO Deterioration.\u003c/b\u003e\u003c/p\u003e\u003cp\u003eTo assess whether the location of metastases is associated with distinct patterns of PRO deterioration, we constructed Cox proportional hazards models for time to first deterioration across each PRO subscale using patients with available metastatic sites data from the CA012-0 and EFC6089 studies (N\u0026thinsp;=\u0026thinsp;432). The associations between metastatic sites and specific PROs are summarized in Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003ea. Hazard ratios (HRs) and 95% confidence intervals (CIs) for functioning domains are presented in Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003eb-e, and for other PRO subscales in \u003cb\u003eSupplementary Fig.\u0026nbsp;4a-k\u003c/b\u003e. Functional PROs (physical functioning, role functioning, and cognitive functioning) were more affected by metastatic sites compared to symptomatic PROs. Patients with bone metastases tended to show earlier deterioration in physical, role, and emotional functioning compared with those without bone metastases (HR\u0026thinsp;\u0026gt;\u0026thinsp;1, p\u0026thinsp;\u0026lt;\u0026thinsp;0.05). Although bone metastases in breast cancer are commonly associated with significant pain\u003csup\u003e\u003cspan citationid=\"CR14\" class=\"CitationRef\"\u003e14\u003c/span\u003e\u003c/sup\u003e, our analysis indicates that pain symptoms may not manifest during the early stages of metastatic involvement. Additionally, patients with lymph node metastases tended to show earlier deterioration in gastrointestinal-related symptoms, including appetite loss and nausea/vomiting. These findings suggest that different metastatic sites are linked to distinct patterns of symptoms and functional deterioration, providing potential clinical insights for disease monitoring and metastasis detection.\u003c/p\u003e\u003cp\u003e\u003c/p\u003e\u003cp\u003e\u003cb\u003eTumor Burden is Associated with PRO Appetite Loss.\u003c/b\u003e\u003c/p\u003e\u003cp\u003eWe evaluated the association between tumor size and PRO deterioration using Cox proportional hazards models based on tumor size data from the CA012-0 and EFC6089 studies (N\u0026thinsp;=\u0026thinsp;359). Larger tumor size was associated with earlier PRO deterioration (HR\u0026thinsp;\u0026gt;\u0026thinsp;1), with appetite loss showing the strongest association across all subscales (HR\u0026thinsp;=\u0026thinsp;1.74, 95% CI 1.33\u0026ndash;2.26, p\u0026thinsp;\u0026lt;\u0026thinsp;0.001; Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003ea). Pain was also strongly correlated with tumor size and patients with larger tumors reported earlier onset of pain symptoms (HR\u0026thinsp;=\u0026thinsp;1.68, 95% CI 1.33\u0026ndash;2.13, p\u0026thinsp;\u0026lt;\u0026thinsp;0.001; Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003ea).\u003c/p\u003e\u003cp\u003e\u003c/p\u003e\u003cp\u003eTo further explore the relationship between tumor burden and the two PROs most strongly associated with it\u0026mdash;appetite loss and pain\u0026mdash;we compared the tumor regrowth ratio, defined as the amount of tumor growth after its smallest size (nadir) during treatment, between patients who experienced PRO deterioration and those who did not. Patients with appetite deterioration had significantly higher tumor regrowth ratios than those without (p\u0026thinsp;\u0026lt;\u0026thinsp;0.01; Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003eb). Similarly, patients with pain deterioration had significantly higher tumor regrowth ratios (p\u0026thinsp;\u0026lt;\u0026thinsp;0.001; Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003ec). Patients whose tumors regrew by \u0026ge;\u0026thinsp;20% from their nadir (defined as disease progression per RECIST criteria\u003csup\u003e\u003cspan citationid=\"CR15\" class=\"CitationRef\"\u003e15\u003c/span\u003e\u003c/sup\u003e) experienced significantly earlier worsening of appetite (p\u0026thinsp;\u0026lt;\u0026thinsp;0.01; Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003ed) and pain (p\u0026thinsp;\u0026lt;\u0026thinsp;0.0001; Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003ee).\u003c/p\u003e\u003cp\u003eThese findings suggest that tumor progression contributes to early systemic symptoms such as appetite loss and pain, potentially reflecting metabolic or inflammatory responses to increasing tumor burden. Monitoring these symptoms may provide early signals of tumor relapse in breast cancer.\u003c/p\u003e\u003c/div\u003e\n\u003ch3\u003ePRO Deterioration is Associated with Patient Survival\u003c/h3\u003e\n\u003cp\u003eTo further examine the relationship between PRO deterioration and survival outcomes in breast cancer, we constructed time-dependent Cox proportional hazards models for OS and PFS using PRO deterioration times as covariates.\u003csup\u003e\u003cspan citationid=\"CR16\" class=\"CitationRef\"\u003e16\u003c/span\u003e\u003c/sup\u003e To account for potential confounding effects, baseline demographic and clinical variables associated with OS or PFS were identified using the LASSO algorithm and included as covariates in the multivariable models. The HRs and 95% CIs for each PRO subscale are presented for OS in Fig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003ea and for PFS in Fig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003eb.\u003c/p\u003e\u003cp\u003e\u003c/p\u003e\u003cp\u003eAcross all subscales, HRs\u0026thinsp;\u0026lt;\u0026thinsp;1 indicated that patients who experienced earlier PRO deterioration tended to have shorter OS and faster disease progression. Among all domains, fatigue showed the lowest HR in both OS (HR\u0026thinsp;=\u0026thinsp;0.13, 95% CI 0.11\u0026ndash;0.15, p\u0026thinsp;\u0026lt;\u0026thinsp;0.0001) and PFS (HR\u0026thinsp;=\u0026thinsp;0.24, 95% CI 0.21\u0026ndash;0.27, p\u0026thinsp;\u0026lt;\u0026thinsp;0.0001) models, suggesting that fatigue-related PRO deterioration was most strongly associated with poorer survival outcomes in breast cancer.\u003c/p\u003e\n\u003ch3\u003eIntegrating PRO with PFS Enhances OS Prediction Using Machine Learning\u003c/h3\u003e\n\u003cp\u003eThe prognostic value of PRO data was evaluated using machine learning approaches. We developed gradient-boosting machine learning models to predict OS status based on different feature sets: (1) PRO deterioration times from all subscales, (2) PFS only, and (3) a combination of PRO deterioration times from all subscales and PFS. Among these models, integrating PRO with PFS achieved the best predictive performance for OS, with an area under the receiver operating characteristic curve (AUC-ROC) of 0.954 (Fig.\u0026nbsp;\u003cspan refid=\"Fig5\" class=\"InternalRef\"\u003e5\u003c/span\u003ea). The AUC-ROC measures how well a model distinguishes between patients who experienced the event and those who did not. A higher value indicates better discrimination. This combined model outperformed the model using PFS alone (AUC-ROC\u0026thinsp;=\u0026thinsp;0.896), suggesting that including PRO data meaningfully improves the prediction of OS.\u003c/p\u003e\u003cp\u003e\u003c/p\u003e\u003cp\u003eBecause PFS data may include survival information that becomes available only after the outcome occurs, it can unintentionally include signals related to OS, leading to biased results in the prediction. We built a landmark machine learning model using only PFS data recorded up to one year. This approach helps prevent information leakage and provides a fair estimate of the model\u0026rsquo;s ability to predict outcomes after day 365. In this analysis, combining PRO with PFS achieved the best prediction of OS (AUC-ROC\u0026thinsp;=\u0026thinsp;0.876), outperforming models that used PRO alone (AUC-ROC\u0026thinsp;=\u0026thinsp;0.838) or PFS alone (AUC-ROC\u0026thinsp;=\u0026thinsp;0.609) (Fig.\u0026nbsp;\u003cspan refid=\"Fig5\" class=\"InternalRef\"\u003e5\u003c/span\u003eb). In this landmark setting, PRO alone performed better than PFS, suggesting that PROs may offer earlier and complementary information for predicting long-term survival.\u003c/p\u003e"},{"header":"Discussion","content":"\u003cp\u003eIn this study, we retrospectively evaluated the deterioration of 15 patient-reported symptoms, functioning scales, and quality of life using 450,147 PRO data from 2,738 breast cancer patients, including both early- and late-stage disease. Association between PRO deterioration and cancer relapse, tumor burden, and metastatic profiles were identified based on this large PRO dataset. We further observed that different metastatic sites were associated with distinct PRO deterioration patterns, and that larger tumor size correlated with earlier PRO deterioration. Using machine learning approaches, we confirmed that adding PRO to PFS can enhance the OS prediction compared to using PFS alone. Together, these findings provide quantitative evidence that longitudinal PRO monitoring can serve as early, non-invasive indicators of disease progression and survival outcomes in breast cancer.\u003c/p\u003e\u003cp\u003eIn our analysis, among breast cancer patients who later developed radiographically confirmed relapse, more than 90% exhibited at least one PRO deterioration before relapse detection, indicating that PRO decline often precedes imaging evidence of recurrence. A similar pattern was observed in early-stage breast cancer patients, with approximately 90% showing PRO deterioration prior to recurrence. These findings provide direct evidence that PROs may serve as early indicators of disease progression or recurrence in breast cancer.\u003c/p\u003e\u003cp\u003eAn interesting finding from our study is that functional PRO subscales were more affected by metastatic sites, whereas symptomatic PROs showed stronger associations with disease burden and survival. This may be because symptomatic PROs more directly reflect tumor-related physiological changes, while functional PROs capture broader aspects of overall health status that often involve multiple organ systems.\u003csup\u003e\u003cspan citationid=\"CR17\" class=\"CitationRef\"\u003e17\u003c/span\u003e\u003c/sup\u003e Among all the PRO subscales, appetite loss demonstrated the strongest correlation with tumor burden. This association may be driven by tumor-induced systemic inflammation and altered metabolic and hormonal signaling in breast cancer.\u003csup\u003e\u003cspan citationid=\"CR18\" class=\"CitationRef\"\u003e18\u003c/span\u003e, \u003cspan citationid=\"CR19\" class=\"CitationRef\"\u003e19\u003c/span\u003e, \u003cspan citationid=\"CR20\" class=\"CitationRef\"\u003e20\u003c/span\u003e\u003c/sup\u003e Furthermore, fatigue emerged as a key prognostic symptom for survival, showing the lowest HRs for both OS and PFS time-dependent Cox proportional hazards models. This indicates that earlier onset of fatigue is associated with earlier disease progression and worsening fatigue may signify the need to evaluate other indicators of progression. Both appetite loss and fatigue are common and distressing symptoms experienced by breast cancer patients, often resulting from disease progression and treatment-related toxicity.\u003csup\u003e\u003cspan citationid=\"CR21\" class=\"CitationRef\"\u003e21\u003c/span\u003e\u003c/sup\u003e Our study shows that PROs represent critical early signals of disease progression and should be systematically monitored as a part of cancer care.\u003csup\u003e\u003cspan citationid=\"CR22\" class=\"CitationRef\"\u003e22\u003c/span\u003e\u003c/sup\u003e\u003c/p\u003e\u003cp\u003eTo further explore the prognostic potential of PROs, we developed machine learning models to predict OS using PFS and PROs. Machine learning algorithms were used because PROs and PFS are potentially correlated and their associations with OS are nonlinear and complex, which may not be well captured by traditional linear regression models.\u003csup\u003e\u003cspan citationid=\"CR23\" class=\"CitationRef\"\u003e23\u003c/span\u003e, \u003cspan citationid=\"CR24\" class=\"CitationRef\"\u003e24\u003c/span\u003e\u003c/sup\u003e Using a gradient boosting machine learning framework, we found integrating PROs with PFS can substantially enhance OS prediction compared to using PFS alone. These results point to the potential of developing composite surrogate endpoints that integrate PROs and PFS for more accurate OS prediction, consistent with evidence from previous studies.\u003csup\u003e\u003cspan citationid=\"CR25\" class=\"CitationRef\"\u003e25\u003c/span\u003e, \u003cspan citationid=\"CR26\" class=\"CitationRef\"\u003e26\u003c/span\u003e\u003c/sup\u003e\u003c/p\u003e\u003cp\u003eDespite their clinical relevance, PROs have not yet been fully integrated into standard oncology practice.\u003csup\u003e\u003cspan citationid=\"CR27\" class=\"CitationRef\"\u003e27\u003c/span\u003e\u003c/sup\u003e Our findings provide strong evidence supporting the incorporation of PROs into electronic health record systems to facilitate continuous clinician monitoring and improve patient management. Integration of PROs into routine clinical workflows could enable earlier identification of disease progression and treatment-related toxicity, enhancing timely clinical intervention.\u003csup\u003e\u003cspan citationid=\"CR28\" class=\"CitationRef\"\u003e28\u003c/span\u003e, \u003cspan citationid=\"CR29\" class=\"CitationRef\"\u003e29\u003c/span\u003e\u003c/sup\u003e Furthermore, standardized electronic PRO collection could generate real-world data to inform regulatory decision-making and optimize patient-centered care models.\u003csup\u003e\u003cspan citationid=\"CR30\" class=\"CitationRef\"\u003e30\u003c/span\u003e\u003c/sup\u003e\u003c/p\u003e\u003cp\u003eWhile our results demonstrate the potential of PROs to provide early signals of disease progression before radiographic detection, we emphasize the importance of using PRO data in tandem with radiographic tools. In our dataset, we identified 113 \u0026ldquo;non-symptomatic\u0026rdquo; breast cancer patients whose all PRO measures showed no deterioration throughout the study. However, 41.6% of these patients still experienced cancer relapse. This finding suggests that while PROs can serve as valuable surveillance tools for a large proportion of breast cancer patients, clinicians should exercise caution when interpreting results for \u0026ldquo;non-symptomatic\u0026rdquo; individuals. A lack of \u0026ldquo;PRO progression\u0026rdquo; should not be interpreted as support for reducing guideline-recommended clinical assessment and surveillance.\u003c/p\u003e\u003cp\u003eOur study has limitations. We evaluated PRO changes using the time to first deterioration approach; however, as demonstrated in our previous analyses, this simplification may not be sensitive enough due to the substantial variability and noise inherent in longitudinal PRO measurements. Future work should leverage more advanced computational methods to capture the clinically meaningful changes in PRO trajectories, such as population modeling approach, which may improve the precision of evaluating how PRO changes are associated with disease progression and survival outcomes.\u003csup\u003e\u003cspan citationid=\"CR31\" class=\"CitationRef\"\u003e31\u003c/span\u003e, \u003cspan citationid=\"CR32\" class=\"CitationRef\"\u003e32\u003c/span\u003e\u003c/sup\u003e\u003c/p\u003e\u003cp\u003eIn conclusion, we analyzed the associations between PRO deterioration and cancer relapse, tumor burden, and survival outcomes in breast cancer. Our findings demonstrate that PROs can serve as complementary, patient-centered indicators to radiographic assessments, enabling non-invasive and more frequent disease monitoring. This work highlights the potential of PROs to evolve from supportive endpoints into reliable clinical biomarkers that enhance disease surveillance in breast cancer care. Our findings align with the ongoing efforts of the FDA and EMA to integrate PROs into oncology trial endpoints and labeling claims, paving the way for more patient-centered regulatory evaluation of cancer therapies.\u003c/p\u003e"},{"header":"Methods","content":"\u003cdiv id=\"Sec8\" class=\"Section2\"\u003e\u003ch2\u003eEthic Statement\u003c/h2\u003e\u003cp\u003eThe data used in this analysis are fully de-identified individual participant data for secondary analysis. The study was granted an Institutional Review Board (IRB) exemption by the University of North Carolina at Chapel Hill.\u003c/p\u003e\u003c/div\u003e\n\u003ch3\u003eClinical Study Descriptions\u003c/h3\u003e\n\u003cp\u003eData were pooled from the control arms of four Phase III clinical trials of breast cancer, including female patients with early to late-stage disease. The studies included are summarized below.\u003c/p\u003e\u003cp\u003e\u003cstrong\u003eBCIRG-001 (NCT00688740)\u003c/strong\u003e\u003cp\u003eThis was an open-label, multicenter Phase III trial that compared efficacy and safety between standard adjuvant anthracycline chemotherapy and anthracycline\u0026ndash;taxane combination chemotherapy in women with operable, node-positive, early-stage breast cancer.\u003csup\u003e\u003cspan citationid=\"CR33\" class=\"CitationRef\"\u003e33\u003c/span\u003e\u003c/sup\u003e Patients who received fluorouracil, doxorubicin, and cyclophosphamide (FAC) every 3 weeks for six cycles were included in the analysis. The study provided 10-year follow-up data on tumor recurrence.\u003c/p\u003e\u003c/p\u003e\u003cp\u003e\u003cstrong\u003eBCIRG-005 (NCT00312208)\u003c/strong\u003e\u003cp\u003eThis was a Phase III, multicenter, prospective trial evaluating efficacy and safety of sequential versus concurrent administration of doxorubicin and cyclophosphamide with docetaxel in women with node-positive, non-metastatic breast cancer.\u003csup\u003e\u003cspan citationid=\"CR34\" class=\"CitationRef\"\u003e34\u003c/span\u003e\u003c/sup\u003e The analysis included patients who received AC\u0026rarr;T (doxorubicin and cyclophosphamide every 3 weeks for four cycles, followed by docetaxel every 3 weeks for four cycles). Tumor recurrence data were available with 10-year follow-up.\u003c/p\u003e\u003c/p\u003e\u003cp\u003e\u003cstrong\u003eCA012-0 (NCT00046527)\u003c/strong\u003e\u003cp\u003eThis was a randomized, open-label, multicenter Phase III trial comparing ABI-007 (a Cremophor-free, protein-stabilized nanoparticle formulation of paclitaxel) with conventional paclitaxel in patients with metastatic breast cancer. Patients who received conventional paclitaxel were included in the analysis. This study provided longitudinal data on tumor size and metastatic organ involvement for individual patients.\u003c/p\u003e\u003c/p\u003e\u003cp\u003e\u003cstrong\u003eEFC6089 (NCT00081796)\u003c/strong\u003e\u003cp\u003eThis was a randomized, open-label Phase III trial comparing larotaxel with capecitabine in patients with metastatic breast cancer progressing after taxane and anthracycline therapy. Patients received capecitabine twice daily for 2 weeks in 3-week cycles. The dataset included longitudinal measurements of tumor size and metastatic organ involvement.\u003c/p\u003e\u003c/p\u003e\n\u003ch3\u003ePRO Measures\u003c/h3\u003e\n\u003cp\u003eThe PRO measures used in this study were the EORTC QLQ-C30 questionnaire.\u003csup\u003e\u003cspan citationid=\"CR35\" class=\"CitationRef\"\u003e35\u003c/span\u003e\u003c/sup\u003e The EORTC QLQ-C30 is a 30 item measures that assesses cancer patients disease-related symptoms, functional status, and overall quality of life (\u003cb\u003eSupplementary Table\u0026nbsp;1\u003c/b\u003e). Symptom scales assess common cancer- and treatment-related symptoms such as fatigue, appetite loss, nausea/vomiting, pain, while functional status measure physical, emotional, cognitive, role, and social functioning. The EORTC QLQ-C30, as one of the most commonly used PRO instrument, has been extensively validated across cancer types and languages, demonstrating strong reliability, construct validity, and sensitivity to clinical change.\u003csup\u003e\u003cspan citationid=\"CR36\" class=\"CitationRef\"\u003e36\u003c/span\u003e, \u003cspan citationid=\"CR37\" class=\"CitationRef\"\u003e37\u003c/span\u003e, \u003cspan citationid=\"CR38\" class=\"CitationRef\"\u003e38\u003c/span\u003e\u003c/sup\u003e\u003c/p\u003e\u003cdiv id=\"Sec11\" class=\"Section2\"\u003e\u003ch2\u003ePRO Deterioration Time\u003c/h2\u003e\u003cp\u003eThe PRO responses from the EORTC QLQ-C30 were first transformed into subscale scores according to the scoring manual\u003csup\u003e\u003cspan citationid=\"CR35\" class=\"CitationRef\"\u003e35\u003c/span\u003e\u003c/sup\u003e using the following equations:\u003cdiv id=\"Equa\" class=\"Equation\"\u003e\u003cdiv format=\"TEX\" class=\"mathdisplay\" id=\"FileID_Equa\" name=\"EquationSource\"\u003e\n$$\\:RS=({I}_{1\\:}+{I}_{2\\:}+\\dots\\:+\\:{I}_{n})/n$$\u003c/div\u003e\u003c/div\u003e\u003c/p\u003e\u003cp\u003eFunctional scales: \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:S=\\left\\{1-\\:\\frac{(RS-1)}{range}\\right\\}\\bullet\\:100\\)\u003c/span\u003e\u003c/span\u003e\u003c/p\u003e\u003cp\u003eSymptom scales/QoL scale: \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:S=\\left\\{(RS-1)/range\\right\\}\\bullet\\:100\\)\u003c/span\u003e\u003c/span\u003e\u003c/p\u003e\u003cp\u003eWhere \u003cb\u003eRS\u003c/b\u003e is the raw score for each PRO subscale calculated from the mean of the corresponding item scores, and \u003cb\u003eS\u003c/b\u003e is the transformed subscale score ranging from 0 to 100. The item composition and range for each subscale are provided in \u003cb\u003eSupplementary Table\u0026nbsp;2\u003c/b\u003e. If one PRO response was missing, the entire score was excluded from the analysis.\u003c/p\u003e\u003cp\u003eThe time to first PRO deterioration (referred to as PRO deterioration time) was defined as the first time a patient\u0026rsquo;s transformed subscale score showed a clinically meaningful change of \u0026ge;\u0026thinsp;10 points from baseline, specifically, an increase of \u0026ge;\u0026thinsp;10 points for symptom or functioning scales, or a decrease of \u0026ge;\u0026thinsp;10 points for the global QoL scale.\u003csup\u003e\u003cspan citationid=\"CR39\" class=\"CitationRef\"\u003e39\u003c/span\u003e\u003c/sup\u003e\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec12\" class=\"Section2\"\u003e\u003ch2\u003eCox Proportional Hazards Model\u003c/h2\u003e\u003cp\u003eCox proportional hazards models were used to evaluate associations between metastatic organs and PRO deterioration, with time to first deterioration as the dependent variable and metastatic organ involvement (pleura, lung, bone, lymph nodes, liver, breast, and other organs) as covariates, based on data from the CA012-0 and EFC6089 studies. Patients without tumor metastatic site information were excluded from this analysis. Separate models were constructed for each of the 15 PRO subscales. In these models, HR greater than 1 indicated that patients with a given metastatic site had a higher risk of earlier PRO deterioration compared to those without metastases at that site.\u003c/p\u003e\u003cp\u003eSimilarly, Cox proportional hazards models were applied to assess associations between tumor size and PRO deterioration using the same datasets from the CA012-0 and EFC6089 studies. Patients without tumor size data were excluded from this analysis. A total of 15 models were developed for each PRO subscale, with higher HRs indicating a stronger association between larger tumor size and earlier PRO deterioration.\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec13\" class=\"Section2\"\u003e\u003ch2\u003eTime-Dependent Cox Proportional Hazards Model\u003c/h2\u003e\u003cp\u003ePRO deterioration time was used as a time-dependent covariate in Cox proportional hazards regression models to examine its associations with OS and PFS using pooled data from four studies (BCIRG-001, BCIRG-005, CA012-0, and EFC6089). Patients without available OS or PFS data were excluded from this analysis.\u003c/p\u003e\u003cp\u003eIn the time-dependent Cox proportional hazards model, the hazard function for patient \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:i\\)\u003c/span\u003e\u003c/span\u003e at time \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:t\\)\u003c/span\u003e\u003c/span\u003e, conditional on the covariate history up to time \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:t\\)\u003c/span\u003e\u003c/span\u003e, was defined as:\u003cdiv id=\"Equb\" class=\"Equation\"\u003e\u003cdiv format=\"TEX\" class=\"mathdisplay\" id=\"FileID_Equb\" name=\"EquationSource\"\u003e\n$$\\:{\\lambda\\:}_{i}\\left(t\\:|\\:{\\mathcal{H}}_{i}\\left(t\\right)\\right)={\\lambda\\:}_{0}\\left(t\\right)exp\\left\\{{\\beta\\:}_{1}{D}_{i}\\left(t\\right)+{\\theta\\:}^{T}{Z}_{i}\\right\\},\\:t\\ge\\:0$$\u003c/div\u003e\u003c/div\u003e\u003c/p\u003e\u003cp\u003ewhere \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{\\lambda\\:}_{0}\\left(t\\right)\\)\u003c/span\u003e\u003c/span\u003e is an unspecified baseline hazard function, \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{D}_{i}\\left(t\\right)\\:\\)\u003c/span\u003e\u003c/span\u003eis a time-dependent indicator variable that equals 1 after the first 10-point deterioration and 0 otherwise, and \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{Z}_{i}\\)\u003c/span\u003e\u003c/span\u003e is a vector of baseline fixed (non-time-dependent) covariates.\u003csup\u003e\u003cspan citationid=\"CR16\" class=\"CitationRef\"\u003e16\u003c/span\u003e\u003c/sup\u003e\u003c/p\u003e\u003cp\u003eModels were constructed for each PRO subscale, incorporating LASSO-selected baseline covariates to adjust for potential confounders.\u003csup\u003e\u003cspan citationid=\"CR40\" class=\"CitationRef\"\u003e40\u003c/span\u003e\u003c/sup\u003e The most prognostic covariates were selected by the lambda value within one standard error from the minimum (lambda.1se). All variables were standardized to a mean of zero and variance of one for consistency. Forest plots were depicted to compare the contributions of each selected covariate. In these models, HRs below 1 indicated that earlier PRO deterioration was associated with shorter survival or faster disease progression.\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec14\" class=\"Section2\"\u003e\u003ch2\u003ePopulation Model Predicting Tumor Size at PRO Deterioration Times\u003c/h2\u003e\u003cp\u003eBecause tumor size obtained via imaging tools (e.g., CT scans) were measured less frequently than PRO assessments, the timing of PRO deterioration often did not align with tumor size measurements. To match the tumor size with PRO deterioration time, we developed a population tumor growth model using a non-linear mixed-effects estimation approach. This method allowed pooling of tumor size data from all patients in the study to simultaneously characterize both the population-level average tumor growth pattern and individual patient trajectories.\u003csup\u003e\u003cspan citationid=\"CR41\" class=\"CitationRef\"\u003e41\u003c/span\u003e\u003c/sup\u003e Using this model, we predicted tumor size at the time of PRO deterioration and examined their associations in the Cox proportional hazards models. This modeling approach has been validated in previous studies for assessing tumor growth heterogeneity across different organs.\u003csup\u003e\u003cspan citationid=\"CR42\" class=\"CitationRef\"\u003e42\u003c/span\u003e, \u003cspan citationid=\"CR43\" class=\"CitationRef\"\u003e43\u003c/span\u003e\u003c/sup\u003e\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec15\" class=\"Section2\"\u003e\u003ch2\u003eMachine Learning Models Predicting OS\u003c/h2\u003e\u003cp\u003eTo evaluate the prognostic value of PRO deterioration times and PFS, we applied XGBoost machine learning algorithms to predict OS using pooled data from four studies (BCIRG-001, BCIRG-005, CA012-0, and EFC6089). Patients without OS status information were excluded from this analysis. Model performance was assessed using the area under the receiver operating characteristic curve (AUC-ROC), which quantifies the model\u0026rsquo;s ability to distinguish between patients who survived and those who died (AUC\u0026thinsp;=\u0026thinsp;1.0 indicates perfect prediction; 0.8 indicates good performance; 0.5 represents random chance).\u003c/p\u003e\u003cp\u003eWe compared AUC-ROC values across three feature sets: (1) PRO deterioration times only; (2) PFS only; (3) A combination of PRO deterioration times and PFS. A landmark XGBoost model was also developed using the same dataset and feature sets. In the landmark analysis, only PFS data available up to day 365 were included to prevent information leakage from future outcomes. To ensure fair comparisons across models, we (1) used identical training and test splits (80% for training data cohort and 20% for testing data cohort), (2) applied consistent preprocessing pipelines (including scaling and feature encoding), and (3) optimized hyperparameters using the same search strategy.\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec16\" class=\"Section2\"\u003e\u003ch2\u003eStatistical Analysis\u003c/h2\u003e\u003cp\u003eFor group comparisons, Wilcoxon tests were applied to assess differences in continuous variables between two groups. Group differences in Kaplan\u0026ndash;Meier curves were evaluated using log-rank tests. Cox proportional hazards models were constructed to estimate HRs with corresponding 95% CIs and p-values. Statistical significance was defined as p\u0026thinsp;\u0026lt;\u0026thinsp;0.05. For p-value not shown, * p\u0026thinsp;\u0026lt;\u0026thinsp;0.05; ** p\u0026thinsp;\u0026lt;\u0026thinsp;0.01; *** p\u0026thinsp;\u0026lt;\u0026thinsp;0.001; **** p\u0026thinsp;\u0026lt;\u0026thinsp;0.0001.\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec17\" class=\"Section2\"\u003e\u003ch2\u003eData Availability\u003c/h2\u003e\u003cp\u003eThe raw data were accessed through Project Data Sphere under license in accordance with their Data Use Agreement. Project Data Sphere is an open-access platform that aggregates de-identified cancer clinical trial data from biopharmaceutical companies, academic medical centers, and government organizations (\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://www.projectdatasphere.org/\u003c/span\u003e\u003cspan address=\"https://www.projectdatasphere.org/\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e). The analysis data are available from the corresponding author upon reasonable request.\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec18\" class=\"Section2\"\u003e\u003ch2\u003eAnalysis Codes Availability\u003c/h2\u003e\u003cp\u003eThe statistical analyses were performed using R 4.4.1 and RStudio Version 2022.07.1\u0026thinsp;+\u0026thinsp;554. The population tumor growth model was developed using Monolix 2024R1 and model simulations were performed using Simulx 2024R1. Both Monolix and Simulx are available at \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://lixoft.com/products/\u003c/span\u003e\u003cspan address=\"https://lixoft.com/products/\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e. The figures were compiled in Adobe Illustrator 2025. The source codes were provided in \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://github.com/Whitney0208/PRO-Breast.git\u003c/span\u003e\u003cspan address=\"https://github.com/Whitney0208/PRO-Breast.git\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/p\u003e\u003c/div\u003e"},{"header":"Declarations","content":"\u003cp\u003e\u003cstrong\u003eConflicts of Interests:\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eB.M may own stock in Novartis. J.H.H is a current employee of Pfizer and may own stock in Pfizer. L.I.W W receives institutional research funding from the National Cancer Institute and previously received personal fees from Celgene/Bristol Myers Squibb as a member of the Scientific Steering Committee for the Connect Multiple Myeloma patient registry. W.A.W. has received institutional research funding from Genentech; receives consulting fees from Teladoc Health, Quantum Health, and Lantern Health; holds equity in Koneksa Health; and has a leadership position in the American Society of Hematology Research Collaborative. E.B. receives institutional research funding from the National Cancer Institute and from the Patient-Centered Outcomes Research Institute, and personal fees for scientific advising from Research Triangle Institute, Thyme Care, N-Power Medicine, Resilience Health, Canopy Care, Savor Health, and Navigating Cancer.\u0026nbsp;\u003c/p\u003e\u003cp\u003e\u003cstrong\u003eAcknowledgements\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThis work was funded by University of North Carolina at Chapel Hill. We disclose the use of AI tools (ChatGPT, Open AI) to assist with grammar and language revisions in the manuscript.\u0026nbsp;\u003c/p\u003e\n"},{"header":"References","content":"\u003col\u003e\u003cli\u003e\u003cspan\u003ePulumati A, Pulumati A, Dwarakanath BS, Verma A, Papineni RVL (2023) Technological advancements in cancer diagnostics: Improvements and limitations. Cancer Rep (Hoboken) 6:e1764\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eWilson BE, Wright K, Koven R, Booth CM (2024) Surveillance Imaging After Curative-Intent Treatment for Cancer: Benefits, Harms, and Evidence. J Clin Oncol 42:2245\u0026ndash;2249\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eThe Lancet N (2019) Patient-reported outcomes in the spotlight. Lancet Neurol 18:981\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eDi Maio M, Basch E, Bryce J, Perrone F (2016) Patient-reported outcomes in the evaluation of toxicity of anticancer treatments. Nat Rev Clin Oncol 13:319\u0026ndash;325\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eLeBlanc TW, Abernethy AP (2017) Patient-reported outcomes in cancer care - hearing the patient voice at greater volume. Nat Rev Clin Oncol 14:763\u0026ndash;772\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eBasch E (2018) Patient-reported outcomes: an essential component of oncology drug development and regulatory review. Lancet Oncol 19:595\u0026ndash;597\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eYap C et al (2024) Advancing patient-centric care: integrating patient reported outcomes for tolerability assessment in early phase clinical trials - insights from an expert virtual roundtable. EClinicalMedicine 76:102838\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eSamuel JN, Booth CM, Eisenhauer E, Brundage M, Berry SR, Gyawali B (2022) Association of Quality-of-Life Outcomes in Cancer Drug Trials With Survival Outcomes and Drug Class. JAMA Oncol 8:879\u0026ndash;886\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eBadaoui S, Kichenadasse G, Rowland A, Sorich MJ, Hopkins AM (2021) Patient-Reported Outcomes Predict Progression-Free Survival of Patients with Advanced Breast Cancer Treated with Abemaciclib. Oncologist 26:562\u0026ndash;568\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eZhou J, Muluneh B, Wang Z, Yao H, Hughes JH (2025) Leveraging Longitudinal Patient-Reported Outcome Trajectories to Predict Survival in Non-Small Cell Lung Cancer. Clin Cancer Res 31:2685\u0026ndash;2694\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eFrost MH, Sloan JA (2002) Quality of life measurements: a soft outcome\u0026ndash;or is it? Am J Manag Care 8:S574\u0026ndash;579\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eMakhni EC, Hennekes ME (2023) The Use of Patient-Reported Outcome Measures in Clinical Practice and Clinical Decision Making. J Am Acad Orthop Surg 31:1059\u0026ndash;1066\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eKaasa S et al (1995) The EORTC core quality of life questionnaire (QLQ-C30): validity and reliability when analysed with patients treated with palliative radiotherapy. Eur J Cancer 31a:2260\u0026ndash;2263\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eTakei D, Tagami K (2023) Management of cancer pain due to bone metastasis. J Bone Min Metab 41:327\u0026ndash;336\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eEisenhauer EA et al (2009) New response evaluation criteria in solid tumours: revised RECIST guideline (version 1.1). Eur J Cancer 45:228\u0026ndash;247\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eFisher LD, Lin DY (1999) Time-dependent covariates in the Cox proportional-hazards regression model. Annu Rev Public Health 20:145\u0026ndash;157\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eKluetz PG et al (2016) Focusing on Core Patient-Reported Outcomes in Cancer Clinical Trials: Symptomatic Adverse Events, Physical Function, and Disease-Related Symptoms. Clin Cancer Res 22:1553\u0026ndash;1558\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eKoh K et al (2025) Cancer-Associated Cachexia: Bridging Clinical Findings with Mechanistic Insights in Human Studies. Cancer Discov 15:1543\u0026ndash;1568\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eRoeland EJ et al (2020) Management of Cancer Cachexia: ASCO Guideline. J Clin Oncol 38:2438\u0026ndash;2453\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eYeom E, Yu K (2022) Understanding the molecular basis of anorexia and tissue wasting in cancer cachexia. Exp Mol Med 54:426\u0026ndash;432\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eAbrahams HJG, Gielissen MFM, Schmits IC, Verhagen C, Rovers MM, Knoop H (2016) Risk factors, prevalence, and course of severe fatigue after breast cancer treatment: a meta-analysis involving 12 327 breast cancer survivors. Ann Oncol 27:965\u0026ndash;974\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eGradishar WJ et al (2024) Breast cancer, version 3.2024, NCCN clinical practice guidelines in oncology. J Natl Compr Canc Netw 22:331\u0026ndash;357\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eKrepper D, Cesari M, Hubel NJ, Zelger P, Sztankay MJ (2024) Machine learning models including patient-reported outcome data in oncology: a systematic literature review and analysis of their reporting quality. J Patient Rep Outcomes 8:126\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eSiena S et al (2007) Association of progression-free survival with patient-reported outcomes and survival: results from a randomised phase 3 trial of panitumumab. Br J Cancer 97:1469\u0026ndash;1474\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eKort\u0026uuml;m M et al (2025) Symptomatic progression-free survival as an emerging patient-centered endpoint in multiple myeloma: a secondary analysis of MagnetsiMM-3 trial data. BMC Cancer 25:1288\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eZhang H et al (2025) Improving Surrogate Endpoints for Survival Prediction Through Integration of Patient-Reported Outcomes. \u003cem\u003emedRxiv\u003c/em\u003e, 2010. 2002.25337154 (2025)\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eColbert JA, Potters L (2025) Overcoming Barriers to Make Patient-Reported Outcome Collection the Standard of Care in Oncology. JAMA Oncol 11:233\u0026ndash;234\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eMazariego C et al (2022) Priority recommendations for the implementation of patient-reported outcomes in clinical cancer care: a Delphi study. J Cancer Surviv 16:33\u0026ndash;43\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eStover A et al (2015) Integrating Patient-Reported Outcome Measures into Routine Cancer Care: Cancer Patients' and Clinicians' Perceptions of Acceptability and Value. EGEMS (Wash DC) 3:1169\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eBellino S, La Salvia A (2024) The Importance of Patient Reported Outcomes in Oncology Clinical Trials and Clinical Practice to Inform Regulatory and Healthcare Decision-Making. Drugs R D 24:123\u0026ndash;127\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eZhou J, Muluneh B, Li Q, Hughes JH (2025) Revolutionizing Patient-Reported Outcomes Analysis for Oncology Drug Development Using Population Models. Clin Cancer Res 31:1580\u0026ndash;1586\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eZou Y, Sun Y, Ravva S, Wagner LI, Zhou J (2025) A Model-Based Meta-Analysis of Pembrolizumab Effects on Patient-Reported Quality of Life: Advancing Patient-Centered Oncology Drug Development. CPT Pharmacometrics Syst Pharmacol\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eMackey JR et al (2013) Adjuvant docetaxel, doxorubicin, and cyclophosphamide in node-positive breast cancer: 10-year follow-up of the phase 3 randomised BCIRG 001 trial. Lancet Oncol 14:72\u0026ndash;80\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eMackey JR et al (2016) Long-term outcomes after adjuvant treatment of sequential versus combination docetaxel with doxorubicin and cyclophosphamide in node-positive breast cancer: BCIRG-005 randomized trial. Ann Oncol 27:1041\u0026ndash;1047\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eFayers P, Aaronson NK, Bjordal K, Gr\u0026oslash;nvold M, Curran D, Bottomley A (2001) EORTC QLQ-C30 scoring manual. European Organisation for research and treatment of cancer\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eAaronson NK et al (1993) The European Organization for Research and Treatment of Cancer QLQ-C30: a quality-of-life instrument for use in international clinical trials in oncology. J Natl Cancer Inst 85:365\u0026ndash;376\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eFayers P, Bottomley A (2002) Quality of life research within the EORTC-the EORTC QLQ-C30. European Organisation for Research and Treatment of Cancer. Eur J Cancer 38(Suppl 4):S125\u0026ndash;133\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eScott NW et al (2006) Comparing translations of the EORTC QLQ-C30 using differential item functioning analyses. \u003cem\u003eQual Life Res\u003c/em\u003e 15, 1103\u0026ndash;1115; discussion 1117\u0026ndash;1120\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eCocks K et al (2024) Time to deterioration of patient-reported outcome endpoints in cancer clinical trials: targeted literature review and best practice recommendations. J Patient Rep Outcomes 8:150\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eTibshirani R (1996) Regression shrinkage and selection via the lasso. J Royal Stat Soc Ser B: Stat Methodol 58:267\u0026ndash;288\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eMould DR, Upton RN (2012) Basic concepts in population modeling, simulation, and model-based drug development. CPT Pharmacometrics Syst Pharmacol 1:e6\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eZhou J, Cipriani A, Liu Y, Fang G, Li Q, Cao Y (2023) Mapping lesion-specific response and progression dynamics and inter-organ variability in metastatic colorectal cancer. Nat Commun 14:417\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eZhou J, Liu Y, Zhang Y, Li Q, Cao Y (2020) Modeling Tumor Evolutionary Dynamics to Predict Clinical Outcomes for Patients with Metastatic Colorectal Cancer: A Retrospective Analysis. Cancer Res 80:591\u0026ndash;601\u003c/span\u003e\u003c/li\u003e\u003c/ol\u003e"}],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":true,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":true,"hideJournal":true,"highlight":"","institution":"","isAcceptedByJournal":false,"isAuthorSuppliedPdf":false,"isDeskRejected":"","isHiddenFromSearch":false,"isInQc":false,"isInWorkflow":false,"isPdf":false,"isPdfUpToDate":true,"isWithdrawnOrRetracted":false,"journal":{"display":true,"email":"[email protected]","identity":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true},"keywords":"","lastPublishedDoi":"10.21203/rs.3.rs-8013336/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-8013336/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003cp\u003ePatient-reported outcomes (PROs) offer a non-invasive, low-cost way to capture patients\u0026rsquo; experiences of symptoms, functioning, and quality of life. Yet, their potential as early indicators of tumor burden, disease progression, and survival remain unclear. In this study, we retrospectively analyzed 450,147 longitudinal PRO entries from 2,738 breast cancer patients pooled from four clinical trials, including both early- and late-stage disease, covering 15 PRO measures assessing symptoms, functioning, and quality of life. Among patients with radiographically confirmed disease progression, 90.6% experienced at least one deterioration in PROs prior to relapse detection (median PRO deterioration time 85 days vs. relapse time 1,338 days), indicating that PROs often worsen before imaging-confirmed relapse. Using Cox proportional hazards models, PRO deterioration was significantly associated with metastatic sites, tumor burden, and survival. Functional PROs were more affected by metastatic sites, whereas symptomatic PROs showed stronger associations with tumor burden and survival outcomes. Appetite loss had the strongest correlation with tumor burden, while fatigue was the most prognostic symptom for both overall survival (OS) and progression-free survival (PFS). The predictive value of PROs for OS was then evaluated using gradient boosting machine learning models. Models integrating PRO deterioration times of all subscales together with PFS achieved the best predictive performance, correctly identifying patient survival outcomes in over 95% of cases (AUC-ROC\u0026thinsp;=\u0026thinsp;0.954), outperforming models using PROs (AUC-ROC\u0026thinsp;=\u0026thinsp;0.838) or PFS alone (AUC-ROC\u0026thinsp;=\u0026thinsp;0.896). This indicates that integrating PROs with PFS enhances the prediction of OS, providing a more powerful approach than using either measure alone. Together, these findings provide quantitative evidence that PROs can serve as early and complementary predictors of disease progression and survival, supporting their use as patient-centered biomarkers in breast cancer management. Our findings align with FDA and EMA efforts to integrate PROs into oncology endpoints, supporting more patient-centered regulatory evaluation.\u003c/p\u003e","manuscriptTitle":"Patient-Reported Outcomes as Early Indicators of Disease Progression and Predictors of Survival via Machine Learning in Breast Cancer","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2025-11-13 08:20:16","doi":"10.21203/rs.3.rs-8013336/v1","editorialEvents":[{"type":"communityComments","content":0}],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true}}],"origin":"","ownerIdentity":"011cc05e-aa6f-4867-b8bf-c1e5e8ed2551","owner":[],"postedDate":"November 13th, 2025","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"posted","subjectAreas":[{"id":57336873,"name":"Health sciences/Oncology/Cancer/Breast cancer"},{"id":57336874,"name":"Health sciences/Signs and symptoms/Comorbidities"}],"tags":[],"updatedAt":"2025-11-19T19:00:16+00:00","versionOfRecord":[],"versionCreatedAt":"2025-11-13 08:20:16","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-8013336","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-8013336","identity":"rs-8013336","version":["v1"]},"buildId":"8U1c8b4HqxoKbykW_rLl7","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}