Lack of Survival Benefit from Adjuvant Chemotherapy in Stage IA Pancreatic Cancer: A Propensity Score-Matched and Causal Forest Analysis

Lack of Survival Benefit from Adjuvant Chemotherapy in Stage IA Pancreatic Cancer: A Propensity Score-Matched and Causal Forest Analysis | 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 Lack of Survival Benefit from Adjuvant Chemotherapy in Stage IA Pancreatic Cancer: A Propensity Score-Matched and Causal Forest Analysis Jinbo Shi, Zhongkai Ni, Shifei Huang, Xiaowen Li, Hai Huang This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-9236354/v1 This work is licensed under a CC BY 4.0 License Status: Under Review Version 1 posted 6 You are reading this latest preprint version Abstract The role of adjuvant chemotherapy in stage IA (T1N0M0) pancreatic cancer remains unclear. We evaluated its survival benefit using the SEER database (2010–2022). Among 727 patients with resected stage IA disease, propensity score matching generated 241 well-balanced pairs. Chemotherapy was associated with worse overall survival (log‑rank P = 0.019) and cancer‑specific survival (P = 0.006). Subgroup analyses by age, T stage, and grade showed no benefit. A causal forest model revealed a median individual treatment effect of –2.96 months (IQR: –4.05 to 2.33 months), indicating no clinically meaningful benefit for most patients. Cross‑validation failed to confirm reliable predictive performance. Adjuvant chemotherapy was not associated with improved survival in resected stage IA pancreatic cancer, challenging its routine use and supporting a more individualized approach. Prospective validation will require multi‑institutional collaboration due to disease rarity. Biological sciences/Cancer Health sciences/Gastroenterology Health sciences/Oncology Early stage pancreatic cancer SEER causal forest machine learning chemotherapy Figures Figure 1 Figure 2 Figure 3 Figure 4 Figure 5 Figure 6 Figure 7 Figure 8 Figure 9 Introduction Pancreatic ductal adenocarcinoma (PDAC) remains one of the most lethal malignancies worldwide, with a 5-year survival rate of approximately 10% despite advances in multimodality therapy [ 1 ]. For patients with localized, resectable disease, adjuvant chemotherapy has become the standard of care, supported by landmark trials demonstrating a significant survival benefit with regimens such as gemcitabine and, more recently, mFOLFIRINOX [ 2 – 5 ]. In summary, postoperative adjuvant chemotherapy is still an essential supplementation to further improve the prognosis of PDAC patients and is recommended for all patients with PDAC following resection according to the European Society for Medical Oncology-European Society of Digestive Oncology (ESMO-ESDO) and National Comprehensive Cancer Network (NCCN) Clinical Practice Guidelines[ 6 , 7 ]. However, these pivotal trials predominantly enrolled patients with node-positive or larger tumors (≥ T3) or borderline resectable tumor, leaving the role of adjuvant therapy in the earliest, node-negative stage IA (T1N0M0) disease largely unexplored. Stage IA PDAC represents a small but clinically distinct subgroup. These tumors are often discovered incidentally, are associated with a more favorable prognosis, and have a substantially lower risk of occult metastatic disease compared with more advanced stages. Extrapolating the results of trials conducted in higher-risk populations to these patients is fraught with uncertainty. The potential for overtreatment—exposing patients to the substantial toxicities of chemotherapy without a clear survival benefit—is a genuine clinical concern. Yet, the evidence base to guide treatment decisions in this subgroup remains sparse and conflicting. While some retrospective studies have suggested a possible benefit[ 8 ], others have found no association[ 9 – 11 ], leaving a significant knowledge gap. Several factors may explain the inconsistency of previous reports. Many earlier studies used older staging systems (e.g., AJCC 6th or 7th edition) that did not distinguish T1a, T1b, and T1c substages, making it difficult to isolate the truly low-risk population[ 11 ]. Others were limited by small sample sizes, lacked adequate adjustment for confounding, or did not explore the possibility of heterogeneous treatment effects—that is, whether certain subgroups might benefit while others do not[ 8 ]. Moreover, the role of modern statistical and machine learning methods in addressing these questions has not been fully leveraged. In this study, we sought to rigorously evaluate the association between adjuvant chemotherapy and survival in patients with resected stage IA PDAC using a contemporary, large-scale population-based cohort from the Surveillance, Epidemiology, and End Results (SEER) database. To address the limitations of previous work, we employed a multifaceted analytical approach. We first performed conventional survival analysis and propensity score matching (PSM) to reduce selection bias. We then conducted subgroup analyses to assess whether the effect of chemotherapy varied by age, tumor grade, or T stage. Finally, we applied a causal forest machine learning model—a method specifically designed to estimate individual-level treatment effects and to detect heterogeneity in the absence of predefined subgroups—to explore whether a subset of patients might derive benefit from adjuvant therapy that would otherwise be obscured by overall averages. By integrating these complementary approaches, we aimed to provide a more comprehensive and nuanced assessment of the value of adjuvant chemotherapy in this low-risk population. Methods Data Source and Study Population This retrospective cohort study utilized data from the SEER database. We identified patients diagnosed with primary pancreatic adenocarcinoma between 2010 and 2022. The inclusion criteria were: (1) histologically confirmed pancreatic adenocarcinoma (ICD-O-3 codes: 8140/3, 8500/3); (2) American Joint Committee on Cancer (AJCC) 8th edition stage IA (T1N0M0); (3) underwent curative-intent surgical resection. Patients with missing survival data, unknown tumor size (precluding accurate T staging), or those who died within 6 months of surgery (to exclude perioperative mortality and ensure adequate chemotherapy exposure) were excluded. The flowchart of data filtering is shown in Fig. 1 . Variables and Outcomes Demographic and clinical variables included age, sex, race, marital status, tumor grade, T stage (according to the AJCC 8th edition), and primary tumor site. The primary exposure was the receipt of adjuvant chemotherapy, coded as "Yes" or "No". The primary outcomes were overall survival (OS) and cancer-specific survival (CSS). OS was defined as time from diagnosis to death from any cause. CSS was defined as time from diagnosis to death attributed to pancreatic cancer; deaths from other causes were censored. Statistical Analysis Conventional and Propensity Score-Matched Analyses Baseline characteristics were compared between groups using standardized differences. Univariate and multivariable Cox proportional hazards models were used to identify factors associated with CSS and OS. To reduce confounding by indication, we performed 1:1 propensity score matching without replacement, using a caliper of 0.2 times the standard deviation of the logit of the propensity score. The propensity score was estimated using a logistic regression model that included age, sex, race, marital status, tumor grade, and T stage. Matched survival curves were estimated using the Kaplan-Meier method and compared with the log-rank test. Causal Forest Analysis To explore heterogeneity in treatment effects, we applied a causal survival forest model using the grf package in R. The model estimates the conditional average treatment effect (CATE), which represents the individual-level difference in 5-year restricted mean survival time (RMST) between chemotherapy and no chemotherapy. All pre-specified covariates were included as potential effect modifiers. Model performance was assessed using five-fold cross-validation, examining the interaction between predicted CATE and treatment in a Cox model, and comparing survival in the predicted benefit (highest CATE tertile) group. Feature importance was calculated to identify the most influential variables. All analyses were performed using R software (version 4.2.2). P-value < 0.05 was considered statistically significant. Results Patient Characteristics and Univariate Analysis As shown in Table 1 , in the univariate analysis of 727 patients with stage IA pancreatic cancer, several clinicopathological factors were significantly associated with cancer-specific survival (CSS). Age, marital status, tumor grade, and T stage all demonstrated prognostic significance (all P < 0.01). Specifically, median CSS decreased with advancing age: it was not reached in patients younger than 65 years, was 106 months in those aged 65–74 years, and was 72 months in patients aged 75 years or older. Married patients had a more favorable outcome (median CSS not reached) compared to unmarried patients (median CSS, 81 months; P = 0.01). Regarding tumor grade, patients with well-differentiated tumors had an unattained median CSS, whereas those with moderately and poorly differentiated or undifferentiated tumors had median CSS of 85 and 97 months, respectively (P < 0.01). For T stage, while median CSS was not reached for T1a and T1b disease, it was 90 months for T1c tumors (P < 0.01). In contrast, receipt of adjuvant chemotherapy was not associated with CSS in this unadjusted analysis (P = 0.22), with median CSS not reached in the untreated group and 106 months in the treated group. Other factors, including race, sex, and primary tumor site, showed no significant association with CSS (P = 0.06, 0.90, and 0.27, respectively). Table 1 Univariate Analysis of Cancer-Specific Survival (CSS)(n = 727) Characteristics Number of patients Median CSS (months) P Overall n = 727 Age < 0.01 =75 219 72 Race 0.06 White 587 106 Black 65 119 Asian 75 NR Sex 0.90 Male 303 103 Female 424 113 Marital status 0.01 Married 476 NR Unmarried 251 81 Tumor site 0.27 Head of pancreas 380 110 Body of pancreas 127 92 Tail of pancreas 123 NR Others 97 106 Grade < 0.01 Well differentiated 163 NR Moderately differentiated 415 85 Poorly or undifferentiated 149 97 T stage < 0.01 T1a 109 NR T1b 93 NR T1c 525 90 Chemotherapy 0.22 No chemotherapy 271 NR Yes 456 106 Note:NR indicates that the median survival time was not reached Multivariable Analysis of Factors Associated with Cancer-Specific Survival and Overall Survival The results of the multivariable Cox regression analysis for cancer-specific survival (CSS) and overall survival (OS) are presented in Tables 2 and 3 , respectively. After adjusting for potential confounders, age ≥ 75 years remained an independent predictor of worse CSS (HR 1.74, 95% CI 1.26–2.40, P < 0.01) and OS (HR 2.03, 95% CI 1.52–2.71, P < 0.01), while age 65–74 years was significantly associated with poorer OS (HR 1.57, 95% CI 1.17–2.10, P < 0.01) but not CSS (HR 1.34, 95% CI 0.96–1.86, P = 0.08). Unmarried status was independently associated with worse CSS (HR 1.41, 95% CI 1.08–1.85, P = 0.01) but not with OS (HR 1.21, 95% CI 0.95–1.54, P = 0.12). Regarding tumor grade, well-differentiated tumors were associated with improved CSS (HR 0.62, 95% CI 0.43–0.90, P = 0.01) and showed a trend toward improved OS that approached statistical significance (HR 0.74, 95% CI 0.54–1.00, P = 0.05). T1c stage was a strong independent predictor of both worse CSS (HR 2.30, 95% CI 1.39–3.79, P < 0.01) and OS (HR 2.15, 95% CI 1.42–3.28, P < 0.01) compared to T1a. In contrast, receipt of adjuvant chemotherapy was not independently associated with CSS (HR 1.09, 95% CI 0.82–1.44, P = 0.55) or OS (HR 1.00, 95% CI 0.78–1.27, P = 0.97) after multivariable adjustment. Table 2 Multivariable Cox regression analysis of cancer-specific survival (CSS) in patients with early-stage pancreatic cancer Characteristics HR (95% CI) P Age =75 1.74 (1.26–2.40) < 0.01 Marital status Married Reference Umarried 1.41 (1.08–1.85) 0.01 Grade 0.62 (0.43–0.90) Well differentiated 0.01 Moderately differentiated Reference 0.25 Poorly or undifferentiated 1.20 (0.88–1.65) T Stage T1a Reference T1b 1.44 (0.77–2.68) 0.25 T1c 2.30 (1.39–3.79) < 0.01 Chemotherapy Reference No chemotherapy Yes 1.09 (0.82–1.44) 0.55 CI: Confidence intervals, HR: Hazard ratios Table 3 Multivariable Cox regression analysis of overall survival (OS) in the same cohort Characteristics HR (95% CI) P Age < 65 Reference 65–74 1.57 (1.17–2.10) =75 2.03 (1.52–2.71) < 0.01 Marital status Married Reference Umarried 1.21 (0.95–1.54) 0.12 Grade 0.74 (0.54-1.00) Well differentiated 0.05 Moderately differentiated Reference 0.29 Poorly or undifferentiated 1.16 (0.88–1.54) T Stage T1a Reference T1b 1.47(0.88–2.47) 0.14 T1c 2.15 (1.42–3.28) < 0.01 Chemotherapy Reference No chemotherapy Yes 1.00 (0.78–1.27) 0.97 CI: Confidence intervals, HR: Hazard ratios Propensity Score-Matched Analysis: Overall Survival and Cancer-specific Survival After propensity score matching, excellent covariate balance was achieved between the chemotherapy and no‑chemotherapy groups. As shown in Fig. 2 , all standardized mean differences (SMDs) were below the commonly accepted threshold of 0.1 after matching, indicating that potential confounding factors were effectively balanced between the two groups. In the matched cohort, the survival benefit of adjuvant chemotherapy was re‑evaluated using Kaplan‑Meier analysis. As illustrated in Fig. 3a, patients who received adjuvant chemotherapy had significantly worse overall survival compared to those who did not (log‑rank P = 0.019). The 5‑year overall survival rates were 53.4% in the chemotherapy group versus 66.7% in the no‑chemotherapy group. Similarly, cancer‑specific survival was also significantly inferior in the chemotherapy group (log‑rank P = 0.006; Fig. 3b), with 5‑year cancer‑specific survival rates of 56.9% and 70.5%, respectively. These findings suggest that, after accounting for baseline imbalances, adjuvant chemotherapy was not associated with improved survival; rather, it was linked to worse outcomes in this cohort of patients with stage IA pancreatic cancer. Figure 3. Kaplan-Meier Curves for Overall Survival and Cancer-Specific Survival After Propensity Score Matching Note (a) Overall survival; (b) Cancer-specific survival. P values were calculated using the log-rank test. The number of patients at risk at each time point is shown below the curves. Subgroup Analysis of Cancer-Specific Survival After PSM To assess whether the effect of adjuvant chemotherapy varied across clinically relevant subgroups, we performed stratified analyses according to age, tumor grade, and T stage within the propensity score‑matched cohort. The results are summarized in Figs. 4 – 6 . As shown in Fig. 4 , no survival benefit from chemotherapy was observed in any age subgroup. In patients aged = 75 years (P = 0.834) ,chemotherapy was not associated with improved cancer‑specific survival. notably, among patients aged between 64–74 years, chemotherapy was associated with significantly worse survival. The analysis by tumor grade (Fig. 5 ) also revealed no survival advantage for chemotherapy across all differentiation categories. For well‑differentiated tumors, the log‑rank P value was 0.089; and for poorly differentiated or undifferentiated tumors, it was 0.651. In the moderately differentiated group, chemotherapy was associated with significantly poorer survival (P = 0.038). Finally, when stratified by T stage (Fig. 6 ), none of the subgroups demonstrated a survival benefit from chemotherapy. For T1a and T1b tumors, the log‑rank P values were 0.950 and 0.434, respectively. For T1c tumors, the difference was statistically significant but in the opposite direction (P < 0.001), with chemotherapy again linked to worse survival. In summary, across all predefined subgroups, adjuvant chemotherapy was not associated with improved cancer‑specific survival; rather, it was paradoxically associated with worse outcomes in certain subgroups. These findings are consistent with the overall null effect observed in the matched cohort and suggest that no subgroup of patients with stage IA pancreatic cancer appears to benefit from adjuvant chemotherapy based on the variables available in this dataset. Note (a) Age < 65 years; (b) Age 65–74 years; (c) Age ≥ 75 years. P values were calculated using the log-rank test for comparison between chemotherapy and no chemotherapy within each subgroup. The number of patients at risk at each time point is shown below the curves. Note (a) Well-differentiated; (b) Moderately-differentiated; (c) Poorly-differentiated or undifferentiated. P values were calculated using the log-rank test for comparison between chemotherapy and no chemotherapy within each subgroup. The number of patients at risk at each time point is shown below the curves. Note (a) T1a; (b) T1b; (v) T1c. P values were calculated using the log-rank test for comparison between chemotherapy and no chemotherapy within each subgroup. The number of patients at risk at each time point is shown below the curves. Causal Forest Analysis To further explore potential heterogeneity in treatment effects, we applied a causal survival forest model. The estimated conditional average treatment effects (CATE) for each patient are summarized in Table 4 and illustrated in Fig. 7 . The distribution of CATE was centered between − 5.0 to -2.5 months, with a median of − 2.96 months (interquartile range: − 4.05 to 2.33 months), and ranged from − 6.10 to 4.00 months. These findings indicate that, for the typical patient, adjuvant chemotherapy had a negligible effect on 5-year restricted mean survival time (RMST), with no clinically meaningful benefit for the majority of patients. Table 4 Summary of Individualized Treatment Effects (CATE) Estimated by Causal Forest Number 727 Q1 -4.046 Min -6.102 Median -2.958 Mean -1.555 Q3 2.326 Max 4.000 SD 3.111 Note: CATE (Conditional Average Treatment Effect) represents the estimated difference in 5-year restricted mean survival time (RMST) between chemotherapy and no chemotherapy for each patient, expressed in months . Positive values indicate a potential survival benefit from chemotherapy, while negative values suggest harm. Abbreviations: SD, standard deviation; Q1, first quartile; Q3, third quartile. Feature importance analysis (Fig. 8 ) identified age ≥ 75 years as the most important variable contributing to heterogeneity in treatment effects (importance score 0.27), followed by poorly differentiated grade (0.12), age 65–74 years (0.09), and male sex (0.09). However, all importance scores were modest, suggesting that none of the available clinical variables strongly dictated treatment effect heterogeneity. Note The histogram shows the distribution of conditional average treatment effects (CATE) estimated by causal forest, representing the difference in 5-year restricted mean survival time (RMST) between chemotherapy and no chemotherapy for each patient. Positive values indicate potential benefit, while negative values suggest harm. The vertical red dashed line indicates zero effect. Note The importance score represents the contribution of each variable to the heterogeneity in estimated treatment effects (CATE). Higher scores indicate greater influence on effect modification. Age ≥ 75 years was the most important factor. The stability and predictive performance of the model were assessed using five-fold cross-validation (Table 5 ). In four of the five folds, the interaction between predicted CATE and chemotherapy was not significant (all P > 0.05); in one fold, the interaction was nominally significant (P = 0.042). The log-rank test comparing cancer-specific survival between chemotherapy and no chemotherapy in the predicted benefit group (defined as the highest CATE tertile) was not significant in any validation fold (all P > 0.11). Furthermore, in the validation set, patients predicted to derive the greatest benefit from chemotherapy showed no survival advantage (log-rank P = 0.605; Fig. 9 ). Table 5 Summary of Individualized Treatment Effects (CATE) Estimated by Causal Forest Number 727 Q1 -4.046 Min -6.102 Median -2.958 Mean -1.555 Q3 2.326 Max 4.000 SD 3.111 Note: CATE (Conditional Average Treatment Effect) represents the estimated difference in 5-year restricted mean survival time (RMST) between chemotherapy and no chemotherapy for each patient, expressed in months . Positive values indicate a potential survival benefit from chemotherapy, while negative values suggest harm. Abbreviations: SD, standard deviation; Q1, first quartile; Q3, third quartile. Note The predicted benefit group was defined as patients in the highest tertile of estimated conditional average treatment effects (CATE) in the validation set. P value was calculated using the log-rank test for comparison between chemotherapy and no chemotherapy. The number of patients at risk at each time point is shown below the curves. Taken together, the causal forest analysis failed to identify a reliable subgroup of patients with stage IA pancreatic cancer who would benefit from adjuvant chemotherapy. The negative value CATE distribution, modest feature importance scores, and lack of consistent cross-validation support all converge to the same conclusion: based on the clinical variables available, no meaningful treatment effect heterogeneity could be demonstrated. Discussion In this population‑based cohort study of 727 patients with resected stage IA pancreatic cancer, we found no evidence that adjuvant chemotherapy improves survival. The absence of benefit was observed consistently across a range of analytical approaches, including conventional multivariable Cox regression, propensity score matching (PSM), subgroup analyses, and a causal forest machine learning model designed specifically to detect treatment effect heterogeneity. In the matched cohort, patients who received chemotherapy had significantly worse overall and cancer‑specific survival. Subgroup analyses by age, tumor grade, and T stage revealed that, in several subgroups, chemotherapy was paradoxically associated with poorer outcomes. The causal forest analysis further supported these findings: the distribution of individual treatment effects was centered near zero (median − 2.96 months), feature importance scores were modest, and the model failed to identify a reliable subgroup of patients who would benefit from adjuvant therapy in cross‑validation. Taken together, these results strongly suggest that, based on currently available clinical variables, adjuvant chemotherapy does not confer a survival advantage in stage IA pancreatic cancer and may even be associated with harm in certain patient subsets. Several factors may help explain why adjuvant chemotherapy appears to lack benefit in this population. First, the toxicity of standard regimens such as mFOLFIRINOX or gemcitabine cannot be overlooked[ 12 – 14 ]. In a patient group already at relatively low risk for cancer‑specific death, the cumulative burden of adverse events—including myelosuppression, neuropathy, fatigue, and gastrointestinal symptoms—may offset any marginal oncologic benefit, and in some cases may contribute to non‑cancer mortality or reduced quality of life[ 13 ]. Second, the biological behavior of stage IA tumors likely differs fundamentally from that of more advanced lesions. The likelihood of occult micrometastatic disease is presumably very low in true T1N0 tumors, leaving little therapeutic target for systemic therapy. This hypothesis is supported by our observation that the negative association between chemotherapy and survival was most pronounced in T1c disease—the subgroup with the highest tumor burden within T1—while T1a and T1b tumors showed no difference. Third, selection bias inherent to observational studies may have played a role. Even after PSM, unmeasured factors such as performance status (ECOG), comorbidity burden, and patient preferences likely influenced treatment decisions. It is possible that patients with subtle, undocumented poor prognostic features were more likely to receive chemotherapy, artificially amplifying the apparent risk associated with treatment. Fourth, the heterogeneity of chemotherapy regimens, cycles, and dosing intensity—none of which are captured in the SEER database—may have diluted any true effect. Patients receiving incomplete or suboptimal courses of therapy would still be classified as having received chemotherapy, potentially biasing the estimate toward the null. Fifth, the absence of data on recurrence patterns and subsequent salvage therapies precludes a more detailed mechanistic understanding. It remains possible that chemotherapy delays recurrence without improving overall survival, but this question cannot be addressed with the available data. Our findings align with several prior studies that have examined the role of adjuvant therapy in very early‑stage pancreatic cancer. Using the National Cancer Database (NCDB), Shaib et al. reported that adjuvant therapy did not confer a survival benefit in patients with resected sub‑centimeter (T1a/T1b) stage IA pancreatic cancer, which is consistent with our observation of no benefit in the overall stage IA population[ 10 ]. However, in that study, patients with tumors measuring 1–2 cm (T1c) did derive a survival benefit from adjuvant treatment. Similarly, using both the SEER database and a multi‑institutional dataset, Zhang et al. demonstrated that adjuvant chemotherapy did not improve survival for stage IA PDAC patients after propensity score matching, while it did confer a benefit for stage IB, IIA, IIB, and III patients. Notably, that study did not perform subgroup analyses by age or by more refined T substage stratification within the IA category, leaving open the question of whether certain patient subgroups might still derive benefit[ 11 ]. A previous SEER‑based analysis by our group also demonstrated that adjuvant chemotherapy did not improve survival in stage IA patients overall, although a potential benefit was observed in the subset with poorly differentiated or undifferentiated histology. In the present study, we strictly applied AJCC 8th edition staging to accurately define the stage IA population, used contemporary data spanning 2010–2022, and employed multiple analytical methods—including propensity score matching, subgroup analyses, and a causal forest machine learning model—to control for confounding and to rigorously explore treatment effect heterogeneity. This comprehensive approach allowed us to confirm the absence of a survival benefit from adjuvant chemotherapy across most subgroups of stage IA pancreatic cancer. Notably, even in the subgroup with poorly differentiated histology, which previously showed a potential benefit in our earlier analysis, the current causal forest model did not consistently identify a reliable treatment effect in cross‑validation. Several strengths distinguish our study from previous work. The use of a contemporary, large‑scale population‑based cohort with extended follow‑up (up to 12 years) allowed for a robust assessment of long‑term outcomes. Strict adherence to AJCC 8th edition staging ensured that the study population accurately reflects current clinical definitions of stage IA disease. Methodologically, the combination of traditional survival analyses, PSM, and a causal forest machine learning model represents a layered approach that strengthens confidence in the null finding. Causal forest, in particular, offers a principled framework for estimating individual‑level treatment effects and testing for heterogeneity, and its consistent negative results across cross‑validation provide strong evidence that, based on the variables available, no meaningful subgroup of patients derives benefit from adjuvant chemotherapy. Several limitations should be acknowledged. First, as a retrospective study derived from a single database (SEER), our findings are subject to unmeasured confounding. Important clinical variables such as performance status (ECOG), specific chemotherapy regimens, number of cycles completed, dose intensity, treatment‑related toxicities, and quality‑of‑life measures were not available. Additionally, prognostic biomarkers such as CA19‑9 were not captured. Second, the validation of the causal forest model was performed using an internal split of the data rather than an external cohort. Although cross‑validation helps guard against overfitting, external validation in an independent dataset would be required to fully confirm the generalizability of our findings. Third, the SEER database does not include information on surgical margin status, recurrence patterns, or salvage therapies, all of which could influence survival independently of the initial adjuvant treatment. Fourth, the small number of patients in certain subgroups (e.g., T1a, T1b) may have limited statistical power to detect subtle differences, although the consistent direction of effect across subgroups argues against a clinically meaningful missed benefit. Fifth, while PSM reduced observable confounding, it cannot account for unmeasured differences between groups. Given these findings and limitations, several avenues for future research merit consideration. External validation of our results in independent, multi‑institutional cohorts is essential. Pooled analyses from large consortia or national registries in different countries could help establish the generalizability of our findings. Incorporation of molecular and genomic data—such as KRAS mutation status, microsatellite instability, or expression of chemotherapy sensitivity markers—may help identify the rare patient who could derive benefit from systemic therapy. Prospective studies that include both survival and quality‑of‑life endpoints are needed to fully assess the risk‑benefit profile of adjuvant chemotherapy in this low‑risk population. Given the rarity of stage IA disease, such studies will require multi‑center collaboration and may benefit from novel trial designs, such as registry‑based randomized controlled trials or platform trials. Conclusion Using a comprehensive analytical framework that includes both conventional survival methods and a state‑of‑the‑art machine learning approach, we found no evidence that adjuvant chemotherapy improves survival in patients with resected stage IA pancreatic cancer. The consistency of these null findings across all analyses suggests that, based on currently available clinical variables, routine administration of adjuvant chemotherapy in this subgroup is unlikely to provide meaningful benefit. These results support a more individualized approach to treatment decisions and underscore the need for further research, ideally incorporating molecular biomarkers and multi‑center collaborations, to identify the rare patients who may truly benefit from systemic therapy. Declarations Data Availability Statement: The data used in this study were obtained from the Surveillance, Epidemiology, and End Results (SEER) database, a publicly available resource. The SEER data can be accessed at https://seer.cancer.gov/data/ after obtaining permission from the National Cancer Institute. The analytical dataset generated during this study, including the propensity score‑matched cohort and causal forest results, is available from the corresponding author upon reasonable request at [email protected] or [email protected] Author Contributions: Jinbo Shi conceived the study, designed the methodology, performed the data analysis, and drafted the manuscript. Hai Huang supervised the study, provided critical revisions, and approved the final manuscript. Zhongkai Ni assisted with data analysis and interpretation. Xiaowen Li contributed to data collection and validation. Shifei. Huang assisted with data curation and manuscript preparation. All authors reviewed and approved the final version. Competing interests: The authors declare no competing interests. Funding Declaration ：This work was supported by Zhejiang Provincial Health Industry Science and Technology Program Provincial–Ministerial Jointly Funded Project (WKJ-ZJ-26079) by Hai Huang. References McGuigan, A., et al., Pancreatic cancer: A review of clinical diagnosis, epidemiology, treatment and outcomes. World Journal of Gastroenterology, 2018. 24 (43): p. 4846–4861. Tranvik, H., et al., Chemotherapy Regimens and Survival in Pancreatic Cancer-A Ten-Year Single Centre Overview. Cancer Medicine, 2025. 14 (23): p. e71416. Mosalem, O.M., et al., Pancreatic ductal adenocarcinoma (PDAC): clinical progress in the last five years. Expert Opinion on Investigational Drugs, 2025. 34 (3): p. 149–160. Ducreux, M., et al., PRODIGE 29-UCGI 26 (NEOPAN): A Phase III Randomized Trial Comparing Chemotherapy With FOLFIRINOX or Gemcitabine in Locally Advanced Pancreatic Carcinoma. Journal of Clinical Oncology‌ 2025. 43 (20): p. 2255–2264. Conroy, T., et al., Five-Year Outcomes of FOLFIRINOX vs Gemcitabine as Adjuvant Therapy for Pancreatic Cancer: A Randomized Clinical Trial. JAMA Oncology, 2022. 8 (11): p. 1571–1578. Seufferlein, T., et al., Pancreatic adenocarcinoma: ESMO-ESDO Clinical Practice Guidelines for diagnosis, treatment and follow-up. Annals of Oncology, 2012. 23 Suppl 7 : p. vii33–40. Tempero, M.A., et al., Pancreatic Adenocarcinoma, version 2.2012: featured updates to the NCCN Guidelines. Journal of the National Comprehensive Cancer Network‌, 2012. 10 (6): p. 703–13. Shi, J., X. Li, and Y. Wu, Whether early stage pancreatic ductal adenocarcinoma patients could benefit from the post-operation chemotherapy regimens: a SEER-based propensity score matching study. Zhejiang Da Xue Xue Bao Yi Xue Ban, 2021. 50 (3): p. 375–382. Luo, C., et al., Comparative effectiveness of chemotherapy in different histological types of pancreatic cancer: a PSM-based study using the `SEER database. Journal of Chemotherapy, 2024. 36 (2): p. 167–178. Shaib, W.L., et al., Role of adjuvant therapy in resected stage IA subcentimeter (T1a/T1b) pancreatic cancer. Cancer, 2019. 125 (1): p. 57–67. Zhang, Y., et al., Stage IA Patients With Pancreatic Ductal Adenocarcinoma Cannot Benefit From Chemotherapy: A Propensity Score Matching Study. Frontiers in Oncology, 2020. 10 : p. 1018. Borghesani, M., et al., Efficacy and Toxicity Analysis of mFOLFIRINOX in High-Grade Gastroenteropancreatic Neuroendocrine Neoplasms. Journal of the National Comprehensive Cancer Network‌, 2024. 22 (5). Sohal, D.P.S., et al., Efficacy of Perioperative Chemotherapy for Resectable Pancreatic Adenocarcinoma: A Phase 2 Randomized Clinical Trial. JAMA Oncology, 2021. 7 (3): p. 421–427. Li, X., et al., Efficacy and safety of SOXIRI versus mFOLFIRINOX in advanced pancreatic cancer. Therapeutic Advances in Medical Oncology, 2023. 15 : p. 17588359231186029. Additional Declarations No competing interests reported. Cite Share Download PDF Status: Under Review Version 1 posted Reviewers agreed at journal 06 May, 2026 Reviewers invited by journal 01 May, 2026 Editor invited by journal 03 Apr, 2026 Editor assigned by journal 30 Mar, 2026 Submission checks completed at journal 30 Mar, 2026 First submitted to journal 26 Mar, 2026 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-9236354","acceptedTermsAndConditions":true,"allowDirectSubmit":false,"archivedVersions":[],"articleType":"Article","associatedPublications":[],"authors":[{"id":635616989,"identity":"e92a68eb-e250-481f-bc1b-a948e2771b61","order_by":0,"name":"Jinbo Shi","email":"","orcid":"","institution":"Hangzhou Hospital of Traditional Chinese Medicine","correspondingAuthor":false,"prefix":"","firstName":"Jinbo","middleName":"","lastName":"Shi","suffix":""},{"id":635616990,"identity":"a84a0963-8c3e-47bd-9246-9bfbdddc0c8e","order_by":1,"name":"Zhongkai Ni","email":"","orcid":"","institution":"Hangzhou Hospital of Traditional Chinese Medicine","correspondingAuthor":false,"prefix":"","firstName":"Zhongkai","middleName":"","lastName":"Ni","suffix":""},{"id":635616991,"identity":"8ee462d4-bd7e-4056-928a-a49a81f96d20","order_by":2,"name":"Shifei Huang","email":"","orcid":"","institution":"Hangzhou Hospital of Traditional Chinese Medicine","correspondingAuthor":false,"prefix":"","firstName":"Shifei","middleName":"","lastName":"Huang","suffix":""},{"id":635616992,"identity":"864dbacc-2e70-483d-9a26-9bbb0a858d7e","order_by":3,"name":"Xiaowen Li","email":"","orcid":"","institution":"Hangzhou Hospital of Traditional Chinese Medicine","correspondingAuthor":false,"prefix":"","firstName":"Xiaowen","middleName":"","lastName":"Li","suffix":""},{"id":635616993,"identity":"9bbe99f3-0547-40d2-94d9-eeb721c2ee21","order_by":4,"name":"Hai Huang","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAAA0UlEQVRIiWNgGAWjYHACNjBpwMDY+CChooaweh4kLc0GD84cI0kLA5vkwxZmwlrs2U+nPeapuWO3nf1wW0ViAxsDf3t3An5beHK3G/Mce5a8syex7UbiDhkGiTNnN+DXIsG7TZqH7XCywQGQljNsDAYSucRo+QfUcv5hW0FiGzORWnjbDtsZ3EhsYyBOy5ncbZJz+w4nGNx42CyRcOYYD0G/sLef3Sbx5tthe4Pz6Q8//qiokeNv78WvBQYSG2DWEqUcBOyJVjkKRsEoGAUjDwAAy3lKZCTZbLoAAAAASUVORK5CYII=","orcid":"","institution":"Hangzhou Hospital of Traditional Chinese Medicine","correspondingAuthor":true,"prefix":"","firstName":"Hai","middleName":"","lastName":"Huang","suffix":""}],"badges":[],"createdAt":"2026-03-26 16:09:41","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-9236354/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-9236354/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":108975955,"identity":"2bc0457b-9e19-42b1-a274-04305afa9d97","added_by":"auto","created_at":"2026-05-11 10:58:31","extension":"png","order_by":1,"title":"Figure 1","display":"","copyAsset":false,"role":"figure","size":224176,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eFlowchart of Patient Selection from the SEER Database\u003c/strong\u003e\u003c/p\u003e","description":"","filename":"1.png","url":"https://assets-eu.researchsquare.com/files/rs-9236354/v1/2fc13c494dd2a4472c20553b.png"},{"id":108975884,"identity":"37c7b238-7f16-41ad-8b83-71c782088f43","added_by":"auto","created_at":"2026-05-11 10:58:06","extension":"png","order_by":2,"title":"Figure 2","display":"","copyAsset":false,"role":"figure","size":101829,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eCovariate balance before and after propensity score matching\u003c/strong\u003e\u003c/p\u003e","description":"","filename":"2.png","url":"https://assets-eu.researchsquare.com/files/rs-9236354/v1/9545412151cb4f0d2572c1f7.png"},{"id":108975606,"identity":"0e966161-81fc-4a01-847a-1515890a588a","added_by":"auto","created_at":"2026-05-11 10:56:51","extension":"png","order_by":3,"title":"Figure 3","display":"","copyAsset":false,"role":"figure","size":180744,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eKaplan-Meier Curves for Overall Survival and Cancer-Specific Survival After Propensity Score Matching\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003e\u003cem\u003eNote: (a) Overall survival; (b) Cancer-specific survival. P values were calculated using the log-rank test. The number of patients at risk at each time point is shown below the curves.\u003c/em\u003e\u003c/p\u003e","description":"","filename":"3.png","url":"https://assets-eu.researchsquare.com/files/rs-9236354/v1/234c8a78522e58a7bd1b31ab.png"},{"id":108975627,"identity":"cf9d2026-f9be-48e9-bc20-7b9fb4287460","added_by":"auto","created_at":"2026-05-11 10:56:56","extension":"png","order_by":4,"title":"Figure 4","display":"","copyAsset":false,"role":"figure","size":160050,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eAge-Stratified Analysis of Cancer-Specific Survival in the Propensity Score-Matched Cohort\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003e\u003cem\u003eNote: (a) Age \u0026lt;65 years; (b) Age 65–74 years; (c) Age ≥75 years. P values were calculated using the log-rank test for comparison between chemotherapy and no chemotherapy within each subgroup. The number of patients at risk at each time point is shown below the curves.\u003c/em\u003e\u003c/p\u003e","description":"","filename":"4.png","url":"https://assets-eu.researchsquare.com/files/rs-9236354/v1/9118760cedf8d9f41f147d6e.png"},{"id":108975604,"identity":"859c90cc-53b7-4383-b1f1-6489c40bd53a","added_by":"auto","created_at":"2026-05-11 10:56:51","extension":"png","order_by":5,"title":"Figure 5","display":"","copyAsset":false,"role":"figure","size":154387,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eGrade-Stratified Analysis of Cancer-Specific Survival in the Propensity Score-Matched Cohort\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003e\u003cem\u003eNote: (a) Well-differentiated; (b) Moderately-differentiated; (c) Poorly-differentiated or undifferentiated. P values were calculated using the log-rank test for comparison between chemotherapy and no chemotherapy within each subgroup. The number of patients at risk at each time point is shown below the curves.\u003c/em\u003e\u003c/p\u003e","description":"","filename":"5.png","url":"https://assets-eu.researchsquare.com/files/rs-9236354/v1/932ed92b65bb0600a0ed2aa3.png"},{"id":108975935,"identity":"676a971d-c842-48c2-a54a-98fd48eb0a25","added_by":"auto","created_at":"2026-05-11 10:58:24","extension":"png","order_by":6,"title":"Figure 6","display":"","copyAsset":false,"role":"figure","size":142150,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eT Stage-Stratified Analysis of Cancer-Specific Survival in the Propensity Score-Matched Cohort\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003e\u003cem\u003eNote：(a) T1a; (b) T1b; (v) T1c. P values were calculated using the log-rank test for comparison between chemotherapy and no chemotherapy within each subgroup. The number of patients at risk at each time point is shown below the curves.\u003c/em\u003e\u003c/p\u003e","description":"","filename":"6.png","url":"https://assets-eu.researchsquare.com/files/rs-9236354/v1/98ace42730b0c160769455eb.png"},{"id":108975631,"identity":"2ed6156b-6642-442b-9d8b-1e2283a3b17b","added_by":"auto","created_at":"2026-05-11 10:56:56","extension":"png","order_by":7,"title":"Figure 7","display":"","copyAsset":false,"role":"figure","size":62296,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eDistribution of Individual Treatment Effects (CATE) Estimated by Causal Forest\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003e\u003cem\u003eNote：The histogram shows the distribution of conditional average treatment effects (CATE) estimated by causal forest, representing the difference in 5-year restricted mean survival time (RMST) between chemotherapy and no chemotherapy for each patient. Positive values indicate potential benefit, while negative values suggest harm. The vertical red dashed line indicates zero effect.\u003c/em\u003e\u003c/p\u003e","description":"","filename":"7.png","url":"https://assets-eu.researchsquare.com/files/rs-9236354/v1/8a1cc030cfdbd20f765de54e.png"},{"id":108975536,"identity":"9f0656ea-c80e-4a19-9d60-91130c273aec","added_by":"auto","created_at":"2026-05-11 10:56:30","extension":"png","order_by":8,"title":"Figure 8","display":"","copyAsset":false,"role":"figure","size":59901,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eFeature Importance for Treatment Effect Heterogeneity Estimated by Causal Forest\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003e\u003cem\u003eNote：The importance score represents the contribution of each variable to the heterogeneity in estimated treatment effects (CATE). Higher scores indicate greater influence on effect modification. Age ≥75 years was the most important factor.\u003c/em\u003e\u003c/p\u003e","description":"","filename":"8.png","url":"https://assets-eu.researchsquare.com/files/rs-9236354/v1/1a1993a64a189daaa96d2089.png"},{"id":108975602,"identity":"4560882d-e068-4a95-a975-f29b4f9156fb","added_by":"auto","created_at":"2026-05-11 10:56:50","extension":"png","order_by":9,"title":"Figure 9","display":"","copyAsset":false,"role":"figure","size":91452,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eKaplan-Meier Curves for Cancer-Specific Survival in the Predicted Benefit Group (Validation Set)\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003e\u003cem\u003eNote：The predicted benefit group was defined as patients in the highest tertile of estimated conditional average treatment effects (CATE) in the validation set. P value was calculated using the log-rank test for comparison between chemotherapy and no chemotherapy. The number of patients at risk at each time point is shown below the curves.\u003c/em\u003e\u003c/p\u003e","description":"","filename":"9.png","url":"https://assets-eu.researchsquare.com/files/rs-9236354/v1/a7179e97f417ebff86cbba13.png"},{"id":108978824,"identity":"c634cf18-96a3-4d6a-8366-0ed0a28b028e","added_by":"auto","created_at":"2026-05-11 11:49:09","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":1358124,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-9236354/v1/d3bc7fa9-c0d3-4b39-9c94-061e7156dc12.pdf"}],"financialInterests":"No competing interests reported.","formattedTitle":"Lack of Survival Benefit from Adjuvant Chemotherapy in Stage IA Pancreatic Cancer: A Propensity Score-Matched and Causal Forest Analysis","fulltext":[{"header":"Introduction","content":"\u003cp\u003ePancreatic ductal adenocarcinoma (PDAC) remains one of the most lethal malignancies worldwide, with a 5-year survival rate of approximately 10% despite advances in multimodality therapy [\u003cspan citationid=\"CR1\" class=\"CitationRef\"\u003e1\u003c/span\u003e]. For patients with localized, resectable disease, adjuvant chemotherapy has become the standard of care, supported by landmark trials demonstrating a significant survival benefit with regimens such as gemcitabine and, more recently, mFOLFIRINOX [\u003cspan additionalcitationids=\"CR3 CR4\" citationid=\"CR2\" class=\"CitationRef\"\u003e2\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR5\" class=\"CitationRef\"\u003e5\u003c/span\u003e]. In summary, postoperative adjuvant chemotherapy is still an essential supplementation to further improve the prognosis of PDAC patients and is recommended for all patients with PDAC following resection according to the European Society for Medical Oncology-European Society of Digestive Oncology (ESMO-ESDO) and National Comprehensive Cancer Network (NCCN) Clinical Practice Guidelines[\u003cspan citationid=\"CR6\" class=\"CitationRef\"\u003e6\u003c/span\u003e, \u003cspan citationid=\"CR7\" class=\"CitationRef\"\u003e7\u003c/span\u003e]. However, these pivotal trials predominantly enrolled patients with node-positive or larger tumors (\u0026ge;\u0026thinsp;T3) or borderline resectable tumor, leaving the role of adjuvant therapy in the earliest, node-negative stage IA (T1N0M0) disease largely unexplored.\u003c/p\u003e \u003cp\u003eStage IA PDAC represents a small but clinically distinct subgroup. These tumors are often discovered incidentally, are associated with a more favorable prognosis, and have a substantially lower risk of occult metastatic disease compared with more advanced stages. Extrapolating the results of trials conducted in higher-risk populations to these patients is fraught with uncertainty. The potential for overtreatment\u0026mdash;exposing patients to the substantial toxicities of chemotherapy without a clear survival benefit\u0026mdash;is a genuine clinical concern. Yet, the evidence base to guide treatment decisions in this subgroup remains sparse and conflicting. While some retrospective studies have suggested a possible benefit[\u003cspan citationid=\"CR8\" class=\"CitationRef\"\u003e8\u003c/span\u003e], others have found no association[\u003cspan additionalcitationids=\"CR10\" citationid=\"CR9\" class=\"CitationRef\"\u003e9\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR11\" class=\"CitationRef\"\u003e11\u003c/span\u003e], leaving a significant knowledge gap.\u003c/p\u003e \u003cp\u003eSeveral factors may explain the inconsistency of previous reports. Many earlier studies used older staging systems (e.g., AJCC 6th or 7th edition) that did not distinguish T1a, T1b, and T1c substages, making it difficult to isolate the truly low-risk population[\u003cspan citationid=\"CR11\" class=\"CitationRef\"\u003e11\u003c/span\u003e]. Others were limited by small sample sizes, lacked adequate adjustment for confounding, or did not explore the possibility of heterogeneous treatment effects\u0026mdash;that is, whether certain subgroups might benefit while others do not[\u003cspan citationid=\"CR8\" class=\"CitationRef\"\u003e8\u003c/span\u003e]. Moreover, the role of modern statistical and machine learning methods in addressing these questions has not been fully leveraged.\u003c/p\u003e \u003cp\u003eIn this study, we sought to rigorously evaluate the association between adjuvant chemotherapy and survival in patients with resected stage IA PDAC using a contemporary, large-scale population-based cohort from the Surveillance, Epidemiology, and End Results (SEER) database. To address the limitations of previous work, we employed a multifaceted analytical approach. We first performed conventional survival analysis and propensity score matching (PSM) to reduce selection bias. We then conducted subgroup analyses to assess whether the effect of chemotherapy varied by age, tumor grade, or T stage. Finally, we applied a causal forest machine learning model\u0026mdash;a method specifically designed to estimate individual-level treatment effects and to detect heterogeneity in the absence of predefined subgroups\u0026mdash;to explore whether a subset of patients might derive benefit from adjuvant therapy that would otherwise be obscured by overall averages. By integrating these complementary approaches, we aimed to provide a more comprehensive and nuanced assessment of the value of adjuvant chemotherapy in this low-risk population.\u003c/p\u003e"},{"header":"Methods","content":"\u003cdiv id=\"Sec3\" class=\"Section2\"\u003e \u003ch2\u003eData Source and Study Population\u003c/h2\u003e \u003cp\u003eThis retrospective cohort study utilized data from the SEER database. We identified patients diagnosed with primary pancreatic adenocarcinoma between 2010 and 2022. The inclusion criteria were: (1) histologically confirmed pancreatic adenocarcinoma (ICD-O-3 codes: 8140/3, 8500/3); (2) American Joint Committee on Cancer (AJCC) 8th edition stage IA (T1N0M0); (3) underwent curative-intent surgical resection. Patients with missing survival data, unknown tumor size (precluding accurate T staging), or those who died within 6 months of surgery (to exclude perioperative mortality and ensure adequate chemotherapy exposure) were excluded. The flowchart of data filtering is shown in Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003e.\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003c/div\u003e\n\u003ch3\u003eVariables and Outcomes\u003c/h3\u003e\n\u003cp\u003eDemographic and clinical variables included age, sex, race, marital status, tumor grade, T stage (according to the AJCC 8th edition), and primary tumor site. The primary exposure was the receipt of adjuvant chemotherapy, coded as \"Yes\" or \"No\". The primary outcomes were overall survival (OS) and cancer-specific survival (CSS). OS was defined as time from diagnosis to death from any cause. CSS was defined as time from diagnosis to death attributed to pancreatic cancer; deaths from other causes were censored.\u003c/p\u003e \u003cdiv id=\"Sec5\" class=\"Section2\"\u003e \u003ch2\u003eStatistical Analysis\u003c/h2\u003e \u003cdiv id=\"Sec6\" class=\"Section3\"\u003e \u003ch2\u003eConventional and Propensity Score-Matched Analyses\u003c/h2\u003e \u003cp\u003eBaseline characteristics were compared between groups using standardized differences. Univariate and multivariable Cox proportional hazards models were used to identify factors associated with CSS and OS. To reduce confounding by indication, we performed 1:1 propensity score matching without replacement, using a caliper of 0.2 times the standard deviation of the logit of the propensity score. The propensity score was estimated using a logistic regression model that included age, sex, race, marital status, tumor grade, and T stage. Matched survival curves were estimated using the Kaplan-Meier method and compared with the log-rank test.\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e\n\u003ch3\u003eCausal Forest Analysis\u003c/h3\u003e\n\u003cp\u003eTo explore heterogeneity in treatment effects, we applied a causal survival forest model using the grf package in R. The model estimates the conditional average treatment effect (CATE), which represents the individual-level difference in 5-year restricted mean survival time (RMST) between chemotherapy and no chemotherapy. All pre-specified covariates were included as potential effect modifiers. Model performance was assessed using five-fold cross-validation, examining the interaction between predicted CATE and treatment in a Cox model, and comparing survival in the predicted benefit (highest CATE tertile) group. Feature importance was calculated to identify the most influential variables.\u003c/p\u003e \u003cp\u003eAll analyses were performed using R software (version 4.2.2). P-value\u0026thinsp;\u0026lt;\u0026thinsp;0.05 was considered statistically significant.\u003c/p\u003e"},{"header":"Results","content":"\u003cdiv id=\"Sec9\" class=\"Section2\"\u003e \u003ch2\u003ePatient Characteristics and Univariate Analysis\u003c/h2\u003e \u003cp\u003eAs shown in Table\u0026nbsp;\u003cspan refid=\"Tab1\" class=\"InternalRef\"\u003e1\u003c/span\u003e, in the univariate analysis of 727 patients with stage IA pancreatic cancer, several clinicopathological factors were significantly associated with cancer-specific survival (CSS). Age, marital status, tumor grade, and T stage all demonstrated prognostic significance (all P\u0026thinsp;\u0026lt;\u0026thinsp;0.01). Specifically, median CSS decreased with advancing age: it was not reached in patients younger than 65 years, was 106 months in those aged 65\u0026ndash;74 years, and was 72 months in patients aged 75 years or older. Married patients had a more favorable outcome (median CSS not reached) compared to unmarried patients (median CSS, 81 months; P\u0026thinsp;=\u0026thinsp;0.01). Regarding tumor grade, patients with well-differentiated tumors had an unattained median CSS, whereas those with moderately and poorly differentiated or undifferentiated tumors had median CSS of 85 and 97 months, respectively (P\u0026thinsp;\u0026lt;\u0026thinsp;0.01). For T stage, while median CSS was not reached for T1a and T1b disease, it was 90 months for T1c tumors (P\u0026thinsp;\u0026lt;\u0026thinsp;0.01). In contrast, receipt of adjuvant chemotherapy was not associated with CSS in this unadjusted analysis (P\u0026thinsp;=\u0026thinsp;0.22), with median CSS not reached in the untreated group and 106 months in the treated group. Other factors, including race, sex, and primary tumor site, showed no significant association with CSS (P\u0026thinsp;=\u0026thinsp;0.06, 0.90, and 0.27, respectively).\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\u003eUnivariate Analysis of Cancer-Specific Survival (CSS)(n\u0026thinsp;=\u0026thinsp;727)\u003c/p\u003e \u003c/div\u003e \u003c/caption\u003e \u003ccolgroup cols=\"4\"\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=\"char\" char=\".\" class=\"colspec\" colname=\"c4\" colnum=\"4\"\u003e\u003c/div\u003e \u003cthead\u003e \u003ctr\u003e \u003cth align=\"left\" colname=\"c1\"\u003e \u003cp\u003eCharacteristics\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c2\"\u003e \u003cp\u003eNumber of patients\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c3\"\u003e \u003cp\u003eMedian CSS (months)\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c4\"\u003e \u003cp\u003eP\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003c/thead\u003e \u003ctbody\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eOverall\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003en\u0026thinsp;=\u0026thinsp;727\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eAge\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=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e\u003cb\u003e\u0026lt;\u0026thinsp;0.01\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u0026lt;\u0026thinsp;65\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e266\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eNR\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e65\u0026ndash;74\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e242\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e106\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u0026gt;=75\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e219\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e72\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eRace\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=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e0.06\u003c/p\u003e \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\u003e587\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e106\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eBlack\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e65\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e119\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eAsian\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e75\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eNR\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eSex\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=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e0.90\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eMale\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e303\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e103\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eFemale\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e424\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e113\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eMarital status\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=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e\u003cb\u003e0.01\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eMarried\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e476\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eNR\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eUnmarried\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e251\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e81\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eTumor site\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=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e0.27\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eHead of pancreas\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e380\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e110\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eBody of pancreas\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e127\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e92\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eTail of pancreas\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e123\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eNR\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eOthers\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e97\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e106\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eGrade\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=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e\u003cb\u003e\u0026lt;\u0026thinsp;0.01\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eWell differentiated\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e163\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eNR\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eModerately differentiated\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e415\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e85\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003ePoorly or undifferentiated\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e149\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e97\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eT stage\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=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e\u003cb\u003e\u0026lt;\u0026thinsp;0.01\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eT1a\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e109\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eNR\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eT1b\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e93\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eNR\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eT1c\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e525\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e90\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eChemotherapy\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=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e0.22\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eNo chemotherapy\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e271\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eNR\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eYes\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e456\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e106\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003c/tbody\u003e \u003c/colgroup\u003e \u003c/table\u003e\u003c/div\u003e \u003c/p\u003e \u003c/div\u003e\n\u003ch3\u003eNote:NR indicates that the median survival time was not reached\u003c/h3\u003e\n\u003cdiv id=\"Sec11\" class=\"Section2\"\u003e \u003ch2\u003eMultivariable Analysis of Factors Associated with Cancer-Specific Survival and Overall Survival\u003c/h2\u003e \u003cp\u003eThe results of the multivariable Cox regression analysis for cancer-specific survival (CSS) and overall survival (OS) are presented in Tables\u0026nbsp;\u003cspan refid=\"Tab2\" class=\"InternalRef\"\u003e2\u003c/span\u003e and \u003cspan refid=\"Tab3\" class=\"InternalRef\"\u003e3\u003c/span\u003e, respectively. After adjusting for potential confounders, age\u0026thinsp;\u0026ge;\u0026thinsp;75 years remained an independent predictor of worse CSS (HR 1.74, 95% CI 1.26\u0026ndash;2.40, P\u0026thinsp;\u0026lt;\u0026thinsp;0.01) and OS (HR 2.03, 95% CI 1.52\u0026ndash;2.71, P\u0026thinsp;\u0026lt;\u0026thinsp;0.01), while age 65\u0026ndash;74 years was significantly associated with poorer OS (HR 1.57, 95% CI 1.17\u0026ndash;2.10, P\u0026thinsp;\u0026lt;\u0026thinsp;0.01) but not CSS (HR 1.34, 95% CI 0.96\u0026ndash;1.86, P\u0026thinsp;=\u0026thinsp;0.08). Unmarried status was independently associated with worse CSS (HR 1.41, 95% CI 1.08\u0026ndash;1.85, P\u0026thinsp;=\u0026thinsp;0.01) but not with OS (HR 1.21, 95% CI 0.95\u0026ndash;1.54, P\u0026thinsp;=\u0026thinsp;0.12). Regarding tumor grade, well-differentiated tumors were associated with improved CSS (HR 0.62, 95% CI 0.43\u0026ndash;0.90, P\u0026thinsp;=\u0026thinsp;0.01) and showed a trend toward improved OS that approached statistical significance (HR 0.74, 95% CI 0.54\u0026ndash;1.00, P\u0026thinsp;=\u0026thinsp;0.05). T1c stage was a strong independent predictor of both worse CSS (HR 2.30, 95% CI 1.39\u0026ndash;3.79, P\u0026thinsp;\u0026lt;\u0026thinsp;0.01) and OS (HR 2.15, 95% CI 1.42\u0026ndash;3.28, P\u0026thinsp;\u0026lt;\u0026thinsp;0.01) compared to T1a. In contrast, receipt of adjuvant chemotherapy was not independently associated with CSS (HR 1.09, 95% CI 0.82\u0026ndash;1.44, P\u0026thinsp;=\u0026thinsp;0.55) or OS (HR 1.00, 95% CI 0.78\u0026ndash;1.27, P\u0026thinsp;=\u0026thinsp;0.97) after multivariable adjustment.\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\u003eMultivariable Cox regression analysis of cancer-specific survival (CSS) in patients with early-stage pancreatic cancer\u003c/p\u003e \u003c/div\u003e \u003c/caption\u003e \u003ccolgroup cols=\"5\"\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 \u003cthead\u003e \u003ctr\u003e \u003cth align=\"left\" colname=\"c1\"\u003e \u003cp\u003eCharacteristics\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colspan=\"3\" nameend=\"c4\" namest=\"c2\"\u003e \u003cp\u003eHR (95% CI)\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c5\"\u003e \u003cp\u003eP\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003c/thead\u003e \u003ctbody\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eAge\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 \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u0026lt;\u0026thinsp;65\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colspan=\"3\" nameend=\"c4\" namest=\"c2\"\u003e \u003cp\u003eReference\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e65\u0026ndash;74\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colspan=\"3\" nameend=\"c4\" namest=\"c2\"\u003e \u003cp\u003e1.34 (0.96\u0026ndash;1.86)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.08\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u0026gt;=75\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colspan=\"3\" nameend=\"c4\" namest=\"c2\"\u003e \u003cp\u003e1.74 (1.26\u0026ndash;2.40)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e\u003cb\u003e\u0026lt;\u0026thinsp;0.01\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eMarital status\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 \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eMarried\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colspan=\"3\" nameend=\"c4\" namest=\"c2\"\u003e \u003cp\u003eReference\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eUmarried\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colspan=\"3\" nameend=\"c4\" namest=\"c2\"\u003e \u003cp\u003e1.41 (1.08\u0026ndash;1.85)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e\u003cb\u003e0.01\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eGrade\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colspan=\"3\" morerows=\"1\" nameend=\"c4\" namest=\"c2\" rowspan=\"2\"\u003e \u003cp\u003e0.62 (0.43\u0026ndash;0.90)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eWell differentiated\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e\u003cb\u003e0.01\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eModerately differentiated\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colspan=\"3\" nameend=\"c4\" namest=\"c2\"\u003e \u003cp\u003eReference\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\" morerows=\"1\" rowspan=\"2\"\u003e \u003cp\u003e0.25\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003ePoorly or undifferentiated\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colspan=\"3\" nameend=\"c4\" namest=\"c2\"\u003e \u003cp\u003e1.20 (0.88\u0026ndash;1.65)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eT Stage\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 \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eT1a\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colspan=\"3\" nameend=\"c4\" namest=\"c2\"\u003e \u003cp\u003eReference\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eT1b\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colspan=\"3\" nameend=\"c4\" namest=\"c2\"\u003e \u003cp\u003e1.44 (0.77\u0026ndash;2.68)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.25\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eT1c\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colspan=\"3\" nameend=\"c4\" namest=\"c2\"\u003e \u003cp\u003e2.30 (1.39\u0026ndash;3.79)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e\u003cb\u003e\u0026lt;\u0026thinsp;0.01\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eChemotherapy\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colspan=\"3\" morerows=\"1\" nameend=\"c4\" namest=\"c2\" rowspan=\"2\"\u003e \u003cp\u003eReference\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eNo chemotherapy\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eYes\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colspan=\"3\" nameend=\"c4\" namest=\"c2\"\u003e \u003cp\u003e1.09 (0.82\u0026ndash;1.44)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.55\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003c/tbody\u003e \u003c/colgroup\u003e \u003c/table\u003e\u003c/div\u003e \u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec12\" class=\"Section2\"\u003e \u003ch2\u003eCI: Confidence intervals, HR: Hazard ratios\u003c/h2\u003e \u003cp\u003e \u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab3\" border=\"1\"\u003e \u003ccaption language=\"En\"\u003e \u003cdiv class=\"CaptionNumber\"\u003eTable 3\u003c/div\u003e \u003cdiv class=\"CaptionContent\"\u003e \u003cp\u003eMultivariable Cox regression analysis of overall survival (OS) in the same cohort\u003c/p\u003e \u003c/div\u003e \u003c/caption\u003e \u003ccolgroup cols=\"5\"\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 \u003cthead\u003e \u003ctr\u003e \u003cth align=\"left\" colname=\"c1\"\u003e \u003cp\u003eCharacteristics\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colspan=\"3\" nameend=\"c4\" namest=\"c2\"\u003e \u003cp\u003eHR (95% CI)\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c5\"\u003e \u003cp\u003eP\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003c/thead\u003e \u003ctbody\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eAge\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 \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u0026lt;\u0026thinsp;65\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colspan=\"3\" nameend=\"c4\" namest=\"c2\"\u003e \u003cp\u003eReference\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e65\u0026ndash;74\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colspan=\"3\" nameend=\"c4\" namest=\"c2\"\u003e \u003cp\u003e1.57 (1.17\u0026ndash;2.10)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e\u003cb\u003e\u0026lt;\u0026thinsp;0.01\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u0026gt;=75\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colspan=\"3\" nameend=\"c4\" namest=\"c2\"\u003e \u003cp\u003e2.03 (1.52\u0026ndash;2.71)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e\u003cb\u003e\u0026lt;\u0026thinsp;0.01\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eMarital status\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 \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eMarried\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colspan=\"3\" nameend=\"c4\" namest=\"c2\"\u003e \u003cp\u003eReference\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eUmarried\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colspan=\"3\" nameend=\"c4\" namest=\"c2\"\u003e \u003cp\u003e1.21 (0.95\u0026ndash;1.54)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.12\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eGrade\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colspan=\"3\" morerows=\"1\" nameend=\"c4\" namest=\"c2\" rowspan=\"2\"\u003e \u003cp\u003e0.74 (0.54-1.00)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eWell differentiated\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e\u003cb\u003e0.05\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eModerately differentiated\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colspan=\"3\" nameend=\"c4\" namest=\"c2\"\u003e \u003cp\u003eReference\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\" morerows=\"1\" rowspan=\"2\"\u003e \u003cp\u003e0.29\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003ePoorly or undifferentiated\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colspan=\"3\" nameend=\"c4\" namest=\"c2\"\u003e \u003cp\u003e1.16 (0.88\u0026ndash;1.54)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eT Stage\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 \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eT1a\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colspan=\"3\" nameend=\"c4\" namest=\"c2\"\u003e \u003cp\u003eReference\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eT1b\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colspan=\"3\" nameend=\"c4\" namest=\"c2\"\u003e \u003cp\u003e1.47(0.88\u0026ndash;2.47)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.14\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eT1c\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colspan=\"3\" nameend=\"c4\" namest=\"c2\"\u003e \u003cp\u003e2.15 (1.42\u0026ndash;3.28)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e\u003cb\u003e\u0026lt;\u0026thinsp;0.01\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eChemotherapy\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colspan=\"3\" morerows=\"1\" nameend=\"c4\" namest=\"c2\" rowspan=\"2\"\u003e \u003cp\u003eReference\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eNo chemotherapy\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eYes\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colspan=\"3\" nameend=\"c4\" namest=\"c2\"\u003e \u003cp\u003e1.00 (0.78\u0026ndash;1.27)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.97\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003c/tbody\u003e \u003c/colgroup\u003e \u003c/table\u003e\u003c/div\u003e \u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec13\" class=\"Section2\"\u003e \u003ch2\u003eCI: Confidence intervals, HR: Hazard ratios\u003c/h2\u003e \u003cdiv id=\"Sec14\" class=\"Section3\"\u003e \u003ch2\u003ePropensity Score-Matched Analysis: Overall Survival and Cancer-specific Survival\u003c/h2\u003e \u003cp\u003eAfter propensity score matching, excellent covariate balance was achieved between the chemotherapy and no‑chemotherapy groups. As shown in Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003e, all standardized mean differences (SMDs) were below the commonly accepted threshold of 0.1 after matching, indicating that potential confounding factors were effectively balanced between the two groups.\u003c/p\u003e \u003cp\u003eIn the matched cohort, the survival benefit of adjuvant chemotherapy was re‑evaluated using Kaplan‑Meier analysis. As illustrated in Fig.\u0026nbsp;3a, patients who received adjuvant chemotherapy had significantly worse overall survival compared to those who did not (log‑rank P\u0026thinsp;=\u0026thinsp;0.019). The 5‑year overall survival rates were 53.4% in the chemotherapy group versus 66.7% in the no‑chemotherapy group. Similarly, cancer‑specific survival was also significantly inferior in the chemotherapy group (log‑rank P\u0026thinsp;=\u0026thinsp;0.006; Fig.\u0026nbsp;3b), with 5‑year cancer‑specific survival rates of 56.9% and 70.5%, respectively. These findings suggest that, after accounting for baseline imbalances, adjuvant chemotherapy was not associated with improved survival; rather, it was linked to worse outcomes in this cohort of patients with stage IA pancreatic cancer.\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cp\u003e \u003cb\u003eFigure 3. Kaplan-Meier Curves for Overall Survival and Cancer-Specific Survival After Propensity Score Matching\u003c/b\u003e \u003c/p\u003e \u003cp\u003e \u003cstrong\u003eNote\u003c/strong\u003e \u003cp\u003e \u003cem\u003e(a) Overall survival; (b) Cancer-specific survival. P values were calculated using the log-rank test. The number of patients at risk at each time point is shown below the curves.\u003c/em\u003e \u003c/p\u003e \u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv id=\"Sec15\" class=\"Section2\"\u003e \u003ch2\u003eSubgroup Analysis of Cancer-Specific Survival After PSM\u003c/h2\u003e \u003cp\u003eTo assess whether the effect of adjuvant chemotherapy varied across clinically relevant subgroups, we performed stratified analyses according to age, tumor grade, and T stage within the propensity score‑matched cohort. The results are summarized in Figs.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e4\u003c/span\u003e\u0026ndash;\u003cspan refid=\"Fig5\" class=\"InternalRef\"\u003e6\u003c/span\u003e.\u003c/p\u003e \u003cp\u003eAs shown in Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e4\u003c/span\u003e, no survival benefit from chemotherapy was observed in any age subgroup. In patients aged\u0026thinsp;\u0026lt;\u0026thinsp;65 years, cancer‑specific survival did not differ significantly between the chemotherapy and no‑chemotherapy groups (log‑rank P\u0026thinsp;=\u0026thinsp;0.746). Similarly, among patients aged\u0026thinsp;\u0026gt;\u0026thinsp;=\u0026thinsp;75 years (P\u0026thinsp;=\u0026thinsp;0.834) ,chemotherapy was not associated with improved cancer‑specific survival. notably, among patients aged between 64\u0026ndash;74 years, chemotherapy was associated with significantly worse survival.\u003c/p\u003e \u003cp\u003eThe analysis by tumor grade (Fig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e5\u003c/span\u003e) also revealed no survival advantage for chemotherapy across all differentiation categories. For well‑differentiated tumors, the log‑rank P value was 0.089; and for poorly differentiated or undifferentiated tumors, it was 0.651. In the moderately differentiated group, chemotherapy was associated with significantly poorer survival (P\u0026thinsp;=\u0026thinsp;0.038).\u003c/p\u003e \u003cp\u003eFinally, when stratified by T stage (Fig.\u0026nbsp;\u003cspan refid=\"Fig5\" class=\"InternalRef\"\u003e6\u003c/span\u003e), none of the subgroups demonstrated a survival benefit from chemotherapy. For T1a and T1b tumors, the log‑rank P values were 0.950 and 0.434, respectively. For T1c tumors, the difference was statistically significant but in the opposite direction (P\u0026thinsp;\u0026lt;\u0026thinsp;0.001), with chemotherapy again linked to worse survival.\u003c/p\u003e \u003cp\u003eIn summary, across all predefined subgroups, adjuvant chemotherapy was not associated with improved cancer‑specific survival; rather, it was paradoxically associated with worse outcomes in certain subgroups. These findings are consistent with the overall null effect observed in the matched cohort and suggest that no subgroup of patients with stage IA pancreatic cancer appears to benefit from adjuvant chemotherapy based on the variables available in this dataset.\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cp\u003e \u003cstrong\u003eNote\u003c/strong\u003e \u003cp\u003e \u003cem\u003e(a) Age\u0026thinsp;\u0026lt;\u0026thinsp;65 years; (b) Age 65\u0026ndash;74 years; (c) Age\u0026thinsp;\u0026ge;\u0026thinsp;75 years. P values were calculated using the log-rank test for comparison between chemotherapy and no chemotherapy within each subgroup. The number of patients at risk at each time point is shown below the curves.\u003c/em\u003e \u003c/p\u003e \u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cp\u003e \u003cstrong\u003eNote\u003c/strong\u003e \u003cp\u003e \u003cem\u003e(a) Well-differentiated; (b) Moderately-differentiated; (c) Poorly-differentiated or undifferentiated. P values were calculated using the log-rank test for comparison between chemotherapy and no chemotherapy within each subgroup. The number of patients at risk at each time point is shown below the curves.\u003c/em\u003e \u003c/p\u003e \u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cp\u003e \u003cstrong\u003eNote\u003c/strong\u003e \u003cp\u003e \u003cem\u003e(a) T1a; (b) T1b; (v) T1c. P values were calculated using the log-rank test for comparison between chemotherapy and no chemotherapy within each subgroup. The number of patients at risk at each time point is shown below the curves.\u003c/em\u003e \u003c/p\u003e \u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec16\" class=\"Section2\"\u003e \u003ch2\u003eCausal Forest Analysis\u003c/h2\u003e \u003cp\u003eTo further explore potential heterogeneity in treatment effects, we applied a causal survival forest model. The estimated conditional average treatment effects (CATE) for each patient are summarized in Table\u0026nbsp;\u003cspan refid=\"Tab4\" class=\"InternalRef\"\u003e4\u003c/span\u003e and illustrated in Fig.\u0026nbsp;\u003cspan refid=\"Fig6\" class=\"InternalRef\"\u003e7\u003c/span\u003e. The distribution of CATE was centered between \u0026minus;\u0026thinsp;5.0 to -2.5 months, with a median of \u0026minus;\u0026thinsp;2.96 months (interquartile range: \u0026minus;\u0026thinsp;4.05 to 2.33 months), and ranged from \u0026minus;\u0026thinsp;6.10 to 4.00 months. These findings indicate that, for the typical patient, adjuvant chemotherapy had a negligible effect on 5-year restricted mean survival time (RMST), with no clinically meaningful benefit for the majority of patients.\u003c/p\u003e \u003cp\u003e \u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab4\" border=\"1\"\u003e \u003ccaption language=\"En\"\u003e \u003cdiv class=\"CaptionNumber\"\u003eTable 4\u003c/div\u003e \u003cdiv class=\"CaptionContent\"\u003e \u003cp\u003eSummary of Individualized Treatment Effects (CATE) Estimated by Causal Forest\u003c/p\u003e \u003c/div\u003e \u003c/caption\u003e \u003ccolgroup cols=\"4\"\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 \u003cthead\u003e \u003ctr\u003e \u003cth align=\"left\" colname=\"c1\"\u003e \u003cp\u003eNumber\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c2\"\u003e \u003cp\u003e727\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c3\"\u003e \u003cp\u003eQ1\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c4\"\u003e \u003cp\u003e-4.046\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003c/thead\u003e \u003ctbody\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eMin\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e-6.102\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e\u003cb\u003eMedian\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e-2.958\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eMean\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e-1.555\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e\u003cb\u003eQ3\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e2.326\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eMax\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e4.000\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e\u003cb\u003eSD\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e3.111\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003c/tbody\u003e \u003c/colgroup\u003e \u003ctfoot\u003e \u003ctr\u003e\u003ctd colspan=\"4\"\u003e\u003cem\u003eNote: CATE (Conditional Average Treatment Effect) represents the estimated difference in 5-year restricted mean survival time (RMST) between chemotherapy and no chemotherapy for each patient, expressed in\u003c/em\u003e \u003cb\u003emonths\u003c/b\u003e. \u003cem\u003ePositive values indicate a potential survival benefit from chemotherapy, while negative values suggest harm. Abbreviations: SD, standard deviation; Q1, first quartile; Q3, third quartile.\u003c/em\u003e\u003c/td\u003e\u003c/tr\u003e \u003c/tfoot\u003e \u003c/table\u003e\u003c/div\u003e \u003c/p\u003e \u003cp\u003eFeature importance analysis (Fig.\u0026nbsp;\u003cspan refid=\"Fig7\" class=\"InternalRef\"\u003e8\u003c/span\u003e) identified age\u0026thinsp;\u0026ge;\u0026thinsp;75 years as the most important variable contributing to heterogeneity in treatment effects (importance score 0.27), followed by poorly differentiated grade (0.12), age 65\u0026ndash;74 years (0.09), and male sex (0.09). However, all importance scores were modest, suggesting that none of the available clinical variables strongly dictated treatment effect heterogeneity.\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cp\u003e \u003cstrong\u003eNote\u003c/strong\u003e \u003cp\u003e \u003cem\u003eThe histogram shows the distribution of conditional average treatment effects (CATE) estimated by causal forest, representing the difference in 5-year restricted mean survival time (RMST) between chemotherapy and no chemotherapy for each patient. Positive values indicate potential benefit, while negative values suggest harm. The vertical red dashed line indicates zero effect.\u003c/em\u003e \u003c/p\u003e \u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cp\u003e \u003cstrong\u003eNote\u003c/strong\u003e \u003cp\u003e \u003cem\u003eThe importance score represents the contribution of each variable to the heterogeneity in estimated treatment effects (CATE). Higher scores indicate greater influence on effect modification. Age\u0026thinsp;\u0026ge;\u0026thinsp;75 years was the most important factor.\u003c/em\u003e \u003c/p\u003e \u003c/p\u003e \u003cp\u003eThe stability and predictive performance of the model were assessed using five-fold cross-validation (Table\u0026nbsp;\u003cspan refid=\"Tab5\" class=\"InternalRef\"\u003e5\u003c/span\u003e). In four of the five folds, the interaction between predicted CATE and chemotherapy was not significant (all P\u0026thinsp;\u0026gt;\u0026thinsp;0.05); in one fold, the interaction was nominally significant (P\u0026thinsp;=\u0026thinsp;0.042). The log-rank test comparing cancer-specific survival between chemotherapy and no chemotherapy in the predicted benefit group (defined as the highest CATE tertile) was not significant in any validation fold (all P\u0026thinsp;\u0026gt;\u0026thinsp;0.11). Furthermore, in the validation set, patients predicted to derive the greatest benefit from chemotherapy showed no survival advantage (log-rank P\u0026thinsp;=\u0026thinsp;0.605; Fig.\u0026nbsp;\u003cspan refid=\"Fig8\" class=\"InternalRef\"\u003e9\u003c/span\u003e).\u003c/p\u003e \u003cp\u003e \u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab5\" border=\"1\"\u003e \u003ccaption language=\"En\"\u003e \u003cdiv class=\"CaptionNumber\"\u003eTable 5\u003c/div\u003e \u003cdiv class=\"CaptionContent\"\u003e \u003cp\u003eSummary of Individualized Treatment Effects (CATE) Estimated by Causal Forest\u003c/p\u003e \u003c/div\u003e \u003c/caption\u003e \u003ccolgroup cols=\"4\"\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 \u003cthead\u003e \u003ctr\u003e \u003cth align=\"left\" colname=\"c1\"\u003e \u003cp\u003eNumber\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c2\"\u003e \u003cp\u003e727\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c3\"\u003e \u003cp\u003eQ1\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c4\"\u003e \u003cp\u003e-4.046\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003c/thead\u003e \u003ctbody\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eMin\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e-6.102\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e\u003cb\u003eMedian\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e-2.958\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eMean\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e-1.555\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e\u003cb\u003eQ3\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e2.326\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eMax\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e4.000\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e\u003cb\u003eSD\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e3.111\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003c/tbody\u003e \u003c/colgroup\u003e \u003ctfoot\u003e \u003ctr\u003e\u003ctd colspan=\"4\"\u003e\u003cem\u003eNote: CATE (Conditional Average Treatment Effect) represents the estimated difference in 5-year restricted mean survival time (RMST) between chemotherapy and no chemotherapy for each patient, expressed in\u003c/em\u003e \u003cb\u003emonths\u003c/b\u003e. \u003cem\u003ePositive values indicate a potential survival benefit from chemotherapy, while negative values suggest harm. Abbreviations: SD, standard deviation; Q1, first quartile; Q3, third quartile.\u003c/em\u003e\u003c/td\u003e\u003c/tr\u003e \u003c/tfoot\u003e \u003c/table\u003e\u003c/div\u003e \u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cp\u003e \u003cstrong\u003eNote\u003c/strong\u003e \u003cp\u003e \u003cem\u003eThe predicted benefit group was defined as patients in the highest tertile of estimated conditional average treatment effects (CATE) in the validation set. P value was calculated using the log-rank test for comparison between chemotherapy and no chemotherapy. The number of patients at risk at each time point is shown below the curves.\u003c/em\u003e \u003c/p\u003e \u003c/p\u003e \u003cp\u003eTaken together, the causal forest analysis failed to identify a reliable subgroup of patients with stage IA pancreatic cancer who would benefit from adjuvant chemotherapy. The negative value CATE distribution, modest feature importance scores, and lack of consistent cross-validation support all converge to the same conclusion: based on the clinical variables available, no meaningful treatment effect heterogeneity could be demonstrated.\u003c/p\u003e \u003c/div\u003e"},{"header":"Discussion","content":"\u003cp\u003eIn this population‑based cohort study of 727 patients with resected stage IA pancreatic cancer, we found no evidence that adjuvant chemotherapy improves survival. The absence of benefit was observed consistently across a range of analytical approaches, including conventional multivariable Cox regression, propensity score matching (PSM), subgroup analyses, and a causal forest machine learning model designed specifically to detect treatment effect heterogeneity. In the matched cohort, patients who received chemotherapy had significantly worse overall and cancer‑specific survival. Subgroup analyses by age, tumor grade, and T stage revealed that, in several subgroups, chemotherapy was paradoxically associated with poorer outcomes. The causal forest analysis further supported these findings: the distribution of individual treatment effects was centered near zero (median \u0026minus;\u0026thinsp;2.96 months), feature importance scores were modest, and the model failed to identify a reliable subgroup of patients who would benefit from adjuvant therapy in cross‑validation. Taken together, these results strongly suggest that, based on currently available clinical variables, adjuvant chemotherapy does not confer a survival advantage in stage IA pancreatic cancer and may even be associated with harm in certain patient subsets.\u003c/p\u003e \u003cp\u003eSeveral factors may help explain why adjuvant chemotherapy appears to lack benefit in this population. First, the toxicity of standard regimens such as mFOLFIRINOX or gemcitabine cannot be overlooked[\u003cspan additionalcitationids=\"CR13\" citationid=\"CR12\" class=\"CitationRef\"\u003e12\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR14\" class=\"CitationRef\"\u003e14\u003c/span\u003e]. In a patient group already at relatively low risk for cancer‑specific death, the cumulative burden of adverse events\u0026mdash;including myelosuppression, neuropathy, fatigue, and gastrointestinal symptoms\u0026mdash;may offset any marginal oncologic benefit, and in some cases may contribute to non‑cancer mortality or reduced quality of life[\u003cspan citationid=\"CR13\" class=\"CitationRef\"\u003e13\u003c/span\u003e]. Second, the biological behavior of stage IA tumors likely differs fundamentally from that of more advanced lesions. The likelihood of occult micrometastatic disease is presumably very low in true T1N0 tumors, leaving little therapeutic target for systemic therapy. This hypothesis is supported by our observation that the negative association between chemotherapy and survival was most pronounced in T1c disease\u0026mdash;the subgroup with the highest tumor burden within T1\u0026mdash;while T1a and T1b tumors showed no difference. Third, selection bias inherent to observational studies may have played a role. Even after PSM, unmeasured factors such as performance status (ECOG), comorbidity burden, and patient preferences likely influenced treatment decisions. It is possible that patients with subtle, undocumented poor prognostic features were more likely to receive chemotherapy, artificially amplifying the apparent risk associated with treatment. Fourth, the heterogeneity of chemotherapy regimens, cycles, and dosing intensity\u0026mdash;none of which are captured in the SEER database\u0026mdash;may have diluted any true effect. Patients receiving incomplete or suboptimal courses of therapy would still be classified as having received chemotherapy, potentially biasing the estimate toward the null. Fifth, the absence of data on recurrence patterns and subsequent salvage therapies precludes a more detailed mechanistic understanding. It remains possible that chemotherapy delays recurrence without improving overall survival, but this question cannot be addressed with the available data.\u003c/p\u003e \u003cp\u003eOur findings align with several prior studies that have examined the role of adjuvant therapy in very early‑stage pancreatic cancer. Using the National Cancer Database (NCDB), Shaib et al. reported that adjuvant therapy did not confer a survival benefit in patients with resected sub‑centimeter (T1a/T1b) stage IA pancreatic cancer, which is consistent with our observation of no benefit in the overall stage IA population[\u003cspan citationid=\"CR10\" class=\"CitationRef\"\u003e10\u003c/span\u003e]. However, in that study, patients with tumors measuring 1\u0026ndash;2 cm (T1c) did derive a survival benefit from adjuvant treatment. Similarly, using both the SEER database and a multi‑institutional dataset, Zhang et al. demonstrated that adjuvant chemotherapy did not improve survival for stage IA PDAC patients after propensity score matching, while it did confer a benefit for stage IB, IIA, IIB, and III patients. Notably, that study did not perform subgroup analyses by age or by more refined T substage stratification within the IA category, leaving open the question of whether certain patient subgroups might still derive benefit[\u003cspan citationid=\"CR11\" class=\"CitationRef\"\u003e11\u003c/span\u003e]. A previous SEER‑based analysis by our group also demonstrated that adjuvant chemotherapy did not improve survival in stage IA patients overall, although a potential benefit was observed in the subset with poorly differentiated or undifferentiated histology.\u003c/p\u003e \u003cp\u003eIn the present study, we strictly applied AJCC 8th edition staging to accurately define the stage IA population, used contemporary data spanning 2010\u0026ndash;2022, and employed multiple analytical methods\u0026mdash;including propensity score matching, subgroup analyses, and a causal forest machine learning model\u0026mdash;to control for confounding and to rigorously explore treatment effect heterogeneity. This comprehensive approach allowed us to confirm the absence of a survival benefit from adjuvant chemotherapy across most subgroups of stage IA pancreatic cancer. Notably, even in the subgroup with poorly differentiated histology, which previously showed a potential benefit in our earlier analysis, the current causal forest model did not consistently identify a reliable treatment effect in cross‑validation. Several strengths distinguish our study from previous work. The use of a contemporary, large‑scale population‑based cohort with extended follow‑up (up to 12 years) allowed for a robust assessment of long‑term outcomes. Strict adherence to AJCC 8th edition staging ensured that the study population accurately reflects current clinical definitions of stage IA disease. Methodologically, the combination of traditional survival analyses, PSM, and a causal forest machine learning model represents a layered approach that strengthens confidence in the null finding. Causal forest, in particular, offers a principled framework for estimating individual‑level treatment effects and testing for heterogeneity, and its consistent negative results across cross‑validation provide strong evidence that, based on the variables available, no meaningful subgroup of patients derives benefit from adjuvant chemotherapy.\u003c/p\u003e \u003cp\u003eSeveral limitations should be acknowledged. First, as a retrospective study derived from a single database (SEER), our findings are subject to unmeasured confounding. Important clinical variables such as performance status (ECOG), specific chemotherapy regimens, number of cycles completed, dose intensity, treatment‑related toxicities, and quality‑of‑life measures were not available. Additionally, prognostic biomarkers such as CA19‑9 were not captured. Second, the validation of the causal forest model was performed using an internal split of the data rather than an external cohort. Although cross‑validation helps guard against overfitting, external validation in an independent dataset would be required to fully confirm the generalizability of our findings. Third, the SEER database does not include information on surgical margin status, recurrence patterns, or salvage therapies, all of which could influence survival independently of the initial adjuvant treatment. Fourth, the small number of patients in certain subgroups (e.g., T1a, T1b) may have limited statistical power to detect subtle differences, although the consistent direction of effect across subgroups argues against a clinically meaningful missed benefit. Fifth, while PSM reduced observable confounding, it cannot account for unmeasured differences between groups.\u003c/p\u003e \u003cp\u003eGiven these findings and limitations, several avenues for future research merit consideration. External validation of our results in independent, multi‑institutional cohorts is essential. Pooled analyses from large consortia or national registries in different countries could help establish the generalizability of our findings. Incorporation of molecular and genomic data\u0026mdash;such as KRAS mutation status, microsatellite instability, or expression of chemotherapy sensitivity markers\u0026mdash;may help identify the rare patient who could derive benefit from systemic therapy. Prospective studies that include both survival and quality‑of‑life endpoints are needed to fully assess the risk‑benefit profile of adjuvant chemotherapy in this low‑risk population. Given the rarity of stage IA disease, such studies will require multi‑center collaboration and may benefit from novel trial designs, such as registry‑based randomized controlled trials or platform trials.\u003c/p\u003e"},{"header":"Conclusion","content":"\u003cp\u003eUsing a comprehensive analytical framework that includes both conventional survival methods and a state‑of‑the‑art machine learning approach, we found no evidence that adjuvant chemotherapy improves survival in patients with resected stage IA pancreatic cancer. The consistency of these null findings across all analyses suggests that, based on currently available clinical variables, routine administration of adjuvant chemotherapy in this subgroup is unlikely to provide meaningful benefit. These results support a more individualized approach to treatment decisions and underscore the need for further research, ideally incorporating molecular biomarkers and multi‑center collaborations, to identify the rare patients who may truly benefit from systemic therapy.\u003c/p\u003e"},{"header":"Declarations","content":"\u003cp\u003e\u003cstrong\u003eData Availability Statement:\u003c/strong\u003e The data used in this study were obtained from the Surveillance, Epidemiology, and End Results (SEER) database, a publicly available resource. The SEER data can be accessed at https://seer.cancer.gov/data/ after obtaining permission from the National Cancer Institute. The analytical dataset generated during this study, including the propensity score‑matched cohort and causal forest results, is available from the corresponding author upon reasonable request at [email protected] or [email protected] \u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eAuthor Contributions:\u003c/strong\u003e Jinbo Shi conceived the study, designed the methodology, performed the data analysis, and drafted the manuscript. Hai Huang supervised the study, provided critical revisions, and approved the final manuscript. Zhongkai Ni assisted with data analysis and interpretation. Xiaowen Li contributed to data collection and validation. Shifei. Huang assisted with data curation and manuscript preparation. All authors reviewed and approved the final version. \u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eCompeting interests: \u003c/strong\u003eThe authors declare no competing interests.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eFunding Declaration\u003c/strong\u003e：This work was supported by Zhejiang Provincial Health Industry Science and Technology Program Provincial–Ministerial Jointly Funded Project (WKJ-ZJ-26079) by Hai Huang.\u003c/p\u003e"},{"header":"References","content":"\u003col\u003e\n \u003cli\u003eMcGuigan, A., et al., \u003cem\u003ePancreatic cancer: A review of clinical diagnosis, epidemiology, treatment and outcomes.\u003c/em\u003e World Journal of Gastroenterology, 2018. \u003cstrong\u003e24\u003c/strong\u003e(43): p. 4846\u0026ndash;4861.\u003c/li\u003e\n \u003cli\u003eTranvik, H., et al., \u003cem\u003eChemotherapy Regimens and Survival in Pancreatic Cancer-A Ten-Year Single Centre Overview.\u003c/em\u003e Cancer Medicine, 2025. \u003cstrong\u003e14\u003c/strong\u003e(23): p. e71416.\u003c/li\u003e\n \u003cli\u003eMosalem, O.M., et al., \u003cem\u003ePancreatic ductal adenocarcinoma (PDAC): clinical progress in the last five years.\u003c/em\u003e Expert Opinion on Investigational Drugs, 2025. \u003cstrong\u003e34\u003c/strong\u003e(3): p. 149\u0026ndash;160.\u003c/li\u003e\n \u003cli\u003eDucreux, M., et al., \u003cem\u003ePRODIGE 29-UCGI 26 (NEOPAN): A Phase III Randomized Trial Comparing Chemotherapy With FOLFIRINOX or Gemcitabine in Locally Advanced Pancreatic Carcinoma.\u003c/em\u003e Journal of Clinical Oncology\u0026zwnj; 2025. \u003cstrong\u003e43\u003c/strong\u003e(20): p. 2255\u0026ndash;2264.\u003c/li\u003e\n \u003cli\u003eConroy, T., et al., \u003cem\u003eFive-Year Outcomes of FOLFIRINOX vs Gemcitabine as Adjuvant Therapy for Pancreatic Cancer: A Randomized Clinical Trial.\u003c/em\u003e JAMA Oncology, 2022. \u003cstrong\u003e8\u003c/strong\u003e(11): p. 1571\u0026ndash;1578.\u003c/li\u003e\n \u003cli\u003eSeufferlein, T., et al., \u003cem\u003ePancreatic adenocarcinoma: ESMO-ESDO Clinical Practice Guidelines for diagnosis, treatment and follow-up.\u003c/em\u003e Annals of Oncology, 2012. \u003cstrong\u003e23 Suppl 7\u003c/strong\u003e: p. vii33\u0026ndash;40.\u003c/li\u003e\n \u003cli\u003eTempero, M.A., et al., \u003cem\u003ePancreatic Adenocarcinoma, version 2.2012: featured updates to the NCCN Guidelines.\u003c/em\u003e Journal of the National Comprehensive Cancer Network\u0026zwnj;, 2012. \u003cstrong\u003e10\u003c/strong\u003e(6): p. 703\u0026ndash;13.\u003c/li\u003e\n \u003cli\u003eShi, J., X. Li, and Y. Wu, \u003cem\u003eWhether early stage pancreatic ductal adenocarcinoma patients could benefit from the post-operation chemotherapy regimens: a SEER-based propensity score matching study.\u003c/em\u003e Zhejiang Da Xue Xue Bao Yi Xue Ban, 2021. \u003cstrong\u003e50\u003c/strong\u003e(3): p. 375\u0026ndash;382.\u003c/li\u003e\n \u003cli\u003eLuo, C., et al., \u003cem\u003eComparative effectiveness of chemotherapy in different histological types of pancreatic cancer: a PSM-based study using the `SEER database.\u003c/em\u003e Journal of Chemotherapy, 2024. \u003cstrong\u003e36\u003c/strong\u003e(2): p. 167\u0026ndash;178.\u003c/li\u003e\n \u003cli\u003eShaib, W.L., et al., \u003cem\u003eRole of adjuvant therapy in resected stage IA subcentimeter (T1a/T1b) pancreatic cancer.\u003c/em\u003e Cancer, 2019. \u003cstrong\u003e125\u003c/strong\u003e(1): p. 57\u0026ndash;67.\u003c/li\u003e\n \u003cli\u003eZhang, Y., et al., \u003cem\u003eStage IA Patients With Pancreatic Ductal Adenocarcinoma Cannot Benefit From Chemotherapy: A Propensity Score Matching Study.\u003c/em\u003e Frontiers in Oncology, 2020. \u003cstrong\u003e10\u003c/strong\u003e: p. 1018.\u003c/li\u003e\n \u003cli\u003eBorghesani, M., et al., \u003cem\u003eEfficacy and Toxicity Analysis of mFOLFIRINOX in High-Grade Gastroenteropancreatic Neuroendocrine Neoplasms.\u003c/em\u003e Journal of the National Comprehensive Cancer Network\u0026zwnj;, 2024. \u003cstrong\u003e22\u003c/strong\u003e(5).\u003c/li\u003e\n \u003cli\u003eSohal, D.P.S., et al., \u003cem\u003eEfficacy of Perioperative Chemotherapy for Resectable Pancreatic Adenocarcinoma: A Phase 2 Randomized Clinical Trial.\u003c/em\u003e JAMA Oncology, 2021. \u003cstrong\u003e7\u003c/strong\u003e(3): p. 421\u0026ndash;427.\u003c/li\u003e\n \u003cli\u003eLi, X., et al., \u003cem\u003eEfficacy and safety of SOXIRI versus mFOLFIRINOX in advanced pancreatic cancer.\u003c/em\u003e Therapeutic Advances in Medical Oncology, 2023. \u003cstrong\u003e15\u003c/strong\u003e: p. 17588359231186029.\u003c/li\u003e\n\u003c/ol\u003e"}],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":true,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":false,"hideJournal":false,"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":"scientific-reports","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":false,"externalIdentity":"scirep","sideBox":"Learn more about [Scientific Reports](http://www.nature.com/srep/)","snPcode":"","submissionUrl":"","title":"Scientific Reports","twitterHandle":"","acdcEnabled":true,"dfaEnabled":true,"editorialSystem":"stoa","reportingPortfolio":"Scientific Reports","inReviewEnabled":true,"inReviewRevisionsEnabled":true},"keywords":"Early stage pancreatic cancer, SEER, causal forest, machine learning, chemotherapy","lastPublishedDoi":"10.21203/rs.3.rs-9236354/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-9236354/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003cp\u003eThe role of adjuvant chemotherapy in stage IA (T1N0M0) pancreatic cancer remains unclear. We evaluated its survival benefit using the SEER database (2010–2022). Among 727 patients with resected stage IA disease, propensity score matching generated 241 well-balanced pairs. Chemotherapy was associated with worse overall survival (log‑rank P = 0.019) and cancer‑specific survival (P = 0.006). Subgroup analyses by age, T stage, and grade showed no benefit. A causal forest model revealed a median individual treatment effect of –2.96 months (IQR: –4.05 to 2.33 months), indicating no clinically meaningful benefit for most patients. Cross‑validation failed to confirm reliable predictive performance. Adjuvant chemotherapy was not associated with improved survival in resected stage IA pancreatic cancer, challenging its routine use and supporting a more individualized approach. Prospective validation will require multi‑institutional collaboration due to disease rarity.\u003c/p\u003e","manuscriptTitle":"Lack of Survival Benefit from Adjuvant Chemotherapy in Stage IA Pancreatic Cancer: A Propensity Score-Matched and Causal Forest Analysis","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2026-05-11 10:53:36","doi":"10.21203/rs.3.rs-9236354/v1","editorialEvents":[{"type":"communityComments","content":0},{"type":"reviewerAgreed","content":"290824745746146824607581340219968348050","date":"2026-05-06T12:59:35+00:00","index":"hide","fulltext":""},{"type":"reviewersInvited","content":"","date":"2026-05-01T12:33:16+00:00","index":"","fulltext":""},{"type":"editorInvited","content":"","date":"2026-04-03T05:43:37+00:00","index":"","fulltext":""},{"type":"editorAssigned","content":"","date":"2026-03-30T05:29:47+00:00","index":"","fulltext":""},{"type":"checksComplete","content":"","date":"2026-03-30T05:29:36+00:00","index":"","fulltext":""},{"type":"submitted","content":"Scientific Reports","date":"2026-03-26T15:57:10+00:00","index":"","fulltext":""}],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"scientific-reports","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":false,"externalIdentity":"scirep","sideBox":"Learn more about [Scientific Reports](http://www.nature.com/srep/)","snPcode":"","submissionUrl":"","title":"Scientific Reports","twitterHandle":"","acdcEnabled":true,"dfaEnabled":true,"editorialSystem":"stoa","reportingPortfolio":"Scientific Reports","inReviewEnabled":true,"inReviewRevisionsEnabled":true}}],"origin":"","ownerIdentity":"6a1ff60d-65f5-4ba2-8bd8-1fbfb74663d5","owner":[],"postedDate":"May 11th, 2026","published":true,"recentEditorialEvents":[{"type":"reviewerAgreed","content":"290824745746146824607581340219968348050","date":"2026-05-06T12:59:35+00:00","index":102,"fulltext":""},{"type":"reviewersInvited","content":"18","date":"2026-05-01T12:33:16+00:00","index":"","fulltext":""}],"rejectedJournal":[],"revision":"","amendment":"","status":"under-review","subjectAreas":[{"id":67650281,"name":"Biological sciences/Cancer"},{"id":67650282,"name":"Health sciences/Gastroenterology"},{"id":67650283,"name":"Health sciences/Oncology"}],"tags":[],"updatedAt":"2026-05-11T10:53:43+00:00","versionOfRecord":[],"versionCreatedAt":"2026-05-11 10:53:36","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-9236354","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-9236354","identity":"rs-9236354","version":["v1"]},"buildId":"XKTyCvWXoU3ODBz1xrDgd","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}