Diet and survival after a diagnosis of ovarian cancer: a pooled analysis from the Ovarian Cancer Association Consortium.

The authors’ responsibilities were as follows – PMW, CLP, KLT, GEH, JR, AB, MCP: conceived and designed the research; EVB, DC, JAD, GGG, MTG, HRH, AJ, SKK, AL, VM, RLM, BQ, JMS, NS, WS, KLT, LT, BT, NW, AHW, PMW: collected and provided the data; TII, RN, PMW: conducted the analysis; CMN, PMW: wrote the original draft of the paper; PMW: has primary responsibility for the final content; and all authors: have read and approved the final manuscript.

The Multidisciplinary Ovarian Cancer Outcomes Group was supported by the Department of Defense Congressionally Directed Medical Research Program (grant number W81XSH-16-2-0010). The Ovarian Cancer Association Consortium is currently funded by the generous contributions of its research investigators and through anonymous donations; previous funding came from a grant from the Ovarian Cancer Research Fund . Individual studies and investigators were supported as follows: AUS: U.S. Army Medical Research and Materiel Command (DAMD17-01-1-0729), National Health & Medical Research Council of Australia (199600, 400413, and 400281), Cancer Councils of New South Wales, Victoria, Queensland, South Australia and Tasmania and Cancer Foundation of Western Australia (Multi-State Applications 191, 211 and 182). AOCS gratefully acknowledges additional support from Ovarian Cancer Australia and the Peter MacCallum Foundation ; DOV: National Institutes of Health R01-CA112523 and R01-CA87538; HAW: U.S. National Institutes of Health (R01-CA58598, N01-CN-55424 and N01-PC-67001); LAC: National Institutes of Health P01-CA17054, P30-CA014089, R01-CA61132, N01-PC67010, R03-CA113148, R03-CA115195, N01-CN025403, P30-CA046592, P30-CA008748 and California Cancer Research Program (00-01389V-20170, 2II0200); MAL: funding for this study was provided by research grant R01-CA61107 from the National Cancer Institute , Bethesda, MD, research grant 94 222 52 from the Danish Cancer Society, Copenhagen, Denmark ; and the Mermaid I and Mermaid III projects; MCC: Melbourne Collaborative Cohort Study (MCCS) cohort recruitment was funded by VicHealth and Cancer Council Victoria . The MCCS was further augmented by Australian National Health and Medical Research Council grants 209057, 396414 and 1074383 and by infrastructure provided by Cancer Council Victoria; NCO: National Institutes of Health (R01-CA076016) and the Department of Defense (DAMD17-02-1-0666); NEC: National Institutes of Health R01-CA54419 and P50-CA105009 and Department of Defense W81XWH-10-1-02802; NJO: National Cancer Institute (NIH-K07 CA095666, R01-CA83918, NIH-K22-CA138563, and P30-CA072720) and the Cancer Institute of New Jersey; OPL: National Health and Medical Research Council of Australia (APP1025142, APP1120431); PLC: The PLCO trial was funded through contracts administered by the Division of Cancer Prevention at the National Cancer Institute, National Institutes of Health, with support from the National Cancer Institute Intramural Research Program; POL: Intramural Research Program of the National Cancer Institute; STA: National Institutes of Health U01 CA71966 and U01 CA69417. CMN, TII, RN, and PMW were funded by the National Health and Medical Research Council of Australia (GNT1173346). MCP was supported in part through the National Institutes of Health/National Cancer Institute Support Grant P30 CA008748 (P.I. S.M. Vickers) to Memorial Sloan Kettering Cancer Center.

We pooled primary data from 2 cohorts, 1 case-only and 10 case-control studies of ovarian cancer ( Supplemental Table 1 ) that provided dietary data (representing the prediagnosis period) through the Multidisciplinary Ovarian Cancer Outcomes Group; all studies also participated in the Ovarian Cancer Association Consortium (OCAC) [ 10 ]. Eight studies were from the United States (Diseases of the Ovary and their Evaluation Study [DOV] [ 11 ], Hawaii Ovarian Cancer Study [HAW] [ 12 ], New England Case–Control Study of Ovarian Cancer [NEC] [ 13 ], North Carolina Ovarian Cancer Study [NCO] [ 14 ], New Jersey Ovarian Cancer Study [NJO] [ 15 ], The Prostate, Lung, Colorectal, and Ovarian Cancer Screening Trial (PLC) [ 16 ], Genetic Epidemiology of Ovarian Cancer Study [STA] [ 17 ], Los Angeles County Case Control Studies of Ovarian Cancer [LAC] [ 18 ]), 2 from Europe (Danish Malignant Ovarian Tumour Study (MAL) [ 19 ], Polish Ovarian Cancer Study (POL) [ 20 ]), and 3 from Australia (Australian Ovarian Cancer Study (AUS) [ 21 ], Melbourne Collaborative Cohort Study (MCC) [ 22 ], and Ovarian Cancer Prognosis and Lifestyle Study (OPL) [ 23 ]). The 13 studies included 10,965 individuals diagnosed with invasive epithelial ovarian cancer between 1990 and 2015. We excluded those without dietary data ( n = 2322), missing vital status information ( n = 296), or with invalid values for the timing of data collection or follow-up ( n = 37). Among the remaining 8310 individuals, we excluded 69 with implausible energy intake (defined as log intake >3 SDs from the mean for that study [ 24 ]), and 220 for whom dietary data were collected >2 y after their diagnosis (because of concerns the passage of time might have affected recall of their prediagnosis diet). In our primary analyses, we further excluded 321 individuals who died in the first year after diagnosis (or had <12 mo follow-up) as it is less likely that diet would affect their outcome. The primary analytic dataset thus included 7700 individuals ( Supplemental Figure 1 ). Researchers at QIMR Berghofer Medical Research Institute were responsible for pooling and harmonizing the dietary data. All studies were approved by the relevant institutional review boards and all participants provided informed consent. Most studies used a food-frequency questionnaire (FFQ) that asked about the usual diet in the year (5 y for POL) before diagnosis. The exceptions were NCO which collected information about specific foods only; and PLC and MCC, which asked about diet at cohort enrolment before diagnosis. Three studies provided food data only (MAL, NCO, and POL) and the other 10 provided both food and nutrient data. Dietary information was collected at a mean of 6.8 (SD 5.1) mo after diagnosis in the case-control and case-only studies, and an average of ∼10 y before diagnosis in the cohort studies. Study sites provided food intake data in different forms (for example, servings, weight or frequency, and portion size). These data were converted to “servings per day” using standard serving sizes where necessary, and foods with similar nutrients were combined to create 15 major food groups (for example, beef, lamb, pork, and ground meat were grouped as “red meat”). We also created a group of “ultraprocessed” foods using the NOVA classification [ 25 ]. After harmonization, the nutrient data were energy adjusted using the residual method [ 26 ]; 45 common nutrients reported by the majority of studies were included in analyses. Using the food and nutrient data, we calculated scores for a range of dietary indices. The Dietary Inflammatory Index (DII) [ 27 ], Novel Dietary Inflammatory Score (NDIS) [ 28 ], and Empirical Dietary Inflammatory Pattern (EDIP) [ 29 ] estimate the inflammatory potential of diet from intakes of a range of nutrients, food groups, and specific food items, respectively. A higher score indicates greater inflammatory potential and thus a less healthy diet. High scores on the Healthy Eating Index-2015 (HEI) [ 30 ], alternate HEI-2010 (aHEI) [ 31 ], WCRF [ 32 ]. and alternate Mediterranean diet (aMed) [ 33 ] indicate a healthier diet. The items used to calculate the indices and the studies for which each index could be calculated are shown in Supplemental Figure 2 . Most indices (except NDIS and HEI) include alcohol as a component: WCRF assigns the highest score to nondrinkers, aHEI and aMed to moderate drinkers (0.5–1.5 standard drinks/day), and alcohol and wine/beer reduce DII and EDIP scores, respectively. We evaluated these scores both including and excluding the alcohol component. Pearson correlations between the indices were generally below 0.5 and only exceeded 0.6 for HEI and aHEI ( r = 0.63) and the DII and aMED ( r = –0.64) ( Supplemental Figure 3 ). With the exception of fruits and vegetables ( r = 0.50), correlations between the main food groups were generally below 0.2 ( Supplemental Figure 4 ). With the exception of alcoholic drinks and soy foods, scores for each index, food/food group, and nutrient were categorized using study-specific quartile cut-points. This was because variation across studies was likely due, at least in part, to the differing numbers and types of food items on the different FFQs. In LAC, and for the dietary indices in NEC, which were conducted in phases, phase-specific cut-points were used as not all variables were available for all phases. Alcohol intake was quantified in standard drinks (containing ∼10 g alcohol) and categorized as 0, ≤1/wk, >1/wk to ≤1/d, >1/d. Because the intake of soy foods was low, this was categorized as none, below, and above the median (in consumers) for the study. Information regarding factors potentially associated with diet or survival was accessed from the OCAC database. This included age at diagnosis (years); race/ethnicity (non-Hispanic white, Hispanic, Asian, Black; racial groups with N ≤ 15 in a study were combined as "other"); education (less than high school, completed high school, college/university); BMI (<25, 25–29, ≥30 kg/m 2 ) reported at diagnosis (STA), for the period 1 y (AUS, LAC, NCO, NEC, and NJO) or 5 y before diagnosis (DOV, HAW, MAL, OPL, and POL), or at cohort entry (PLC and MCC); prediagnosis smoking status (never, former, and current), physical activity (“active” compared with “inactive,” defined as no regular weekly recreational physical activity [ 34 ]) and menopausal hormone therapy (MHT) use (any, none). Clinical information included tumor stage (local, regional, and distant), histotype (high-grade serous, low-grade serous, mucinous, endometrioid, clear cell, and other), and presence of residual disease after surgery (none, any). Missing indicators were used when nondietary variables were missing; sensitivity analyses showed that key results were essentially unchanged when we used multiple imputation for missing data (data not shown). Each study reported vital status and survival time, calculated from the date of diagnosis to the date of death from any cause or date of last follow-up for those still alive. This information was obtained from a variety of sources including medical record review, patient contact, and linkage with state cancer registries, Surveillance, Epidemiology, and End Results (SEER) registries (in the United States), and national death registries. Cause of death information was only available for 7 studies (AUS, DOV, HAW, MAL, NCO, OPL, and PLC); in these studies, the vast majority of deaths (92% overall, 95% in the first 5 y) were from ovarian cancer. We used multivariable Cox proportional hazards regression to estimate hazard ratios (HR) and 95% confidence intervals (CI) for the associations between dietary indices, foods and food groups, and nutrients and survival. Because the vast majority of deaths were from ovarian cancer, we used death from any cause as the primary outcome to maximize power and conducted a secondary analysis looking at death from ovarian cancer in the subset of studies with this information. In our primary analyses, we excluded those who died in the first year; we therefore left truncated survival to 1 y after diagnosis or the date of questionnaire completion if this occurred >1 y after diagnosis. This was to avoid immortal time bias and reduce the potential of survivorship bias arising from the exclusion of eligible individuals who died before recruitment. All models were stratified by study, and also by tumor stage (except stage-specific analyses) and histotype as the baseline hazard varied greatly by these factors. A priori, all models were adjusted for age at diagnosis (continuous), year of diagnosis (continuous), race, education, BMI, smoking status, physical activity, and MHT use. Models for foods and nutrients were additionally adjusted for total energy intake (log); energy was not included in models for dietary indices as most of these were standardized for energy intake and including energy made no appreciable difference to the estimates. We also considered other variables including parity, oral contraceptive pill use, tubal ligation, endometriosis, use of nonsteroidal anti-inflammatory drugs, and residual disease but, as none were associated with both diet and survival, they were not included in the final models. Most studies had limited or no information regarding comorbidities and the available information was very heterogeneous. Adjusting for Charlson comorbidity score [ 35 ] in 2 studies that had this information (AUS, OPL) made little difference to the results; hence, the final models were not adjusted for comorbidity. We modeled quartile categories for each food/nutrient and dietary index and assessed linear trends by assigning each quartile a number from 0 (Q1) to 3 (Q4) and modeling this as a continuous variable. The only variables to violate the proportional hazards assumption were age at diagnosis and stage but, as including log-time interactions for these variables in the models did not alter the estimates for the dietary variables of interest, the interactions were not included in the final models. We investigated whether associations differed according to the stage of ovarian cancer at diagnosis (local or regional compared with distant), histotype [high-grade serous cancer (HGSC), other], residual disease, menopausal status, education level, BMI, smoking or physical activity by including a cross-product term between quartile of the dietary score (as a continuous variable) and the potential effect modifier in the models. A P value of <0.05 was considered evidence of significant heterogeneity. We also assessed whether associations differed for short-term follow-up (≤5 y from diagnosis). In sensitivity analyses, we assessed the effects of including those who died within the first 12 mo after diagnosis as well as additionally excluding those who provided data >1 y after diagnosis ( N = 1112), those from the cohort studies when the dietary questionnaire could have been completed many years before diagnosis (MCC, PLC; N = 253) and those from the studies with no information about energy intake (MAL, NCO, and POL). Analyses were performed using SAS version 9.4 (SAS Institute) and STATA version 13. No adjustment was performed for multiple comparisons.

The total number of individuals included in each study ranged from 93 (MAL) to 1455 (NEC) ( Supplemental Table 1 ). Of the 8021 participants included, 4988 (62%) had died, including 316 who died in the first year; 5-y mortality ranged from 38% (NEC and NJO) to 58% (MCC). The median and maximum follow-up among those who had not died were 11 and 26 y, respectively. Table 1 shows the characteristics of the participants included in primary analyses (that is, excluding those who died in the first year). Older age at diagnosis, tumor stage and residual disease after surgery had clear adverse effects on survival and, as expected, survival was worst for those with high-grade serous cancers. Age-adjusted survival was somewhat worse for smokers and Black individuals, and somewhat better among those with a college or university education and of Asian ethnicity. TABLE 1 Characteristics of the study population used for primary analysis ( N = 7700). TABLE 1 Characteristic Subgroup N % Age-adjusted HR (95% CI) Age at diagnosis (mean and range) 7700 56.9 (20–87) Vital status Alive 3028 39.3 Deceased 4672 60.7 Age in categories (y) <40 492 6.4 0.42 (0.36, 0.50) 40–49 1463 19.0 0.79 (0.72, 0.86) 50–59 2460 32.0 1.0 (Ref) 60–69 2251 29.2 1.28 (1.19, 1.38) 70+ 1034 13.4 1.63 (1.49, 1.78) Tumor stage Local 1505 19.6 1.0 Regional 1199 15.6 1.56 (1.35, 1.80) Distant 4785 62.1 5.79 (5.18, 6.47) Missing 211 2.8 Histotype High-grade serous 4543 59.0 1.0 Low-grade serous 273 3.6 0.65 (0.56, 0.76) Mucinous 434 5.6 0.27 (0.22, 0.32) Endometrioid 1107 14.4 0.31 (0.28, 0.35) Clear cell 572 7.4 0.31 (0.27, 0.36) Mixed/other 771 10.0 0.62 (0.56, 0.68) Residual disease after surgery 1 Nil 1097 14.2 1.0 Any 1147 14.9 3.48 (3.08, 3.92) Missing 5456 70.9 Menopausal status 2 Pre- and perimenopause 2137 27.8 1.0 Post menopause 5273 68.5 1.09 (0.99, 1.21) Missing 290 3.8 Menopausal hormone No 4860 63.1 1.0 use Yes 2731 35.5 1.00 (0.94, 1.06) Missing 109 1.4 Race/ethnicity Non-Hispanic White 6462 83.9 1.0 Hispanic 245 3.2 1.03 (0.86, 1.22) Black 222 2.9 1.24 (0.98, 1.56) Asian 458 6.0 0.83 (0.67, 1.03) Other 313 4.1 1.08 (0.85, 1.37) Education Less than high school 1206 15.7 1.0 High school 3984 51.7 0.97 (0.89, 1.06) College or university 2461 32.0 0.92 (0.84, 1.02) Missing 49 0.6 Smoking status Never smoker 4143 53.8 1.0 Former smoker 2495 32.4 1.12 (1.05, 1.19) Current smoker 1020 13.2 1.22 (1.12, 1.34) Missing 42 0.6 BMI <25.0 3705 48.1 1.0 (kg/m 2 ) ≥25 – <30 2206 28.7 0.98 (0.92, 1.05) ≥30 1715 22.3 1.04 (0.97, 1.12) Missing 74 1.0 Physical activity 3 Active 4594 59.7 1.0 Inactive 1343 17.4 1.03 (0.95, 1.12) Missing 1763 22.9 Abbreviations: CI, confidence interval; DOV, Diseases of the Ovary and their Evaluation Study; HR, hazard ratio, adjusted for age and stratified by site; LAC, Los Angeles County Case Control Studies of Ovarian Cancer; MCC, Melbourne Collaborative Cohort Study; NCO, North Carolina Ovarian Cancer Study; NEC, New England Case–Control Study of Ovarian Cancer; NJO, New Jersey Ovarian Cancer Study; PLC, Prostate, Lung, Colorectal, and Ovarian Cancer Screening Trial; POL, Polish Ovarian Cancer Study; STA, STA, Genetic Epidemiology of Ovarian Cancer Study (Stanford). 1 Residual disease information was not available for DOV, LAC, MCC, NJO, PLC, POL or STA ( N = 3152) or for >50% of those from NCO and NEC ( N = 1899). 2 Menopausal status data were not available for MCC or PLC ( N = 222). 3 Physical activity data were not available for MCC, NCO, POL, or STA ( N = 1598). Characteristics of the study population used for primary analysis ( N = 7700). Abbreviations: CI, confidence interval; DOV, Diseases of the Ovary and their Evaluation Study; HR, hazard ratio, adjusted for age and stratified by site; LAC, Los Angeles County Case Control Studies of Ovarian Cancer; MCC, Melbourne Collaborative Cohort Study; NCO, North Carolina Ovarian Cancer Study; NEC, New England Case–Control Study of Ovarian Cancer; NJO, New Jersey Ovarian Cancer Study; PLC, Prostate, Lung, Colorectal, and Ovarian Cancer Screening Trial; POL, Polish Ovarian Cancer Study; STA, STA, Genetic Epidemiology of Ovarian Cancer Study (Stanford). Residual disease information was not available for DOV, LAC, MCC, NJO, PLC, POL or STA ( N = 3152) or for >50% of those from NCO and NEC ( N = 1899). Menopausal status data were not available for MCC or PLC ( N = 222). Physical activity data were not available for MCC, NCO, POL, or STA ( N = 1598). Table 2 shows that, overall, none of the selected dietary indices were strongly associated with survival after adjusting for potential confounders. There was no significant heterogeneity by study site ( Supplemental Figures 5 and 6 ); however, associations appeared to differ by stage of disease and this heterogeneity was statistically significant for the aHEI ( P -interaction=0.01), aMed ( P -interaction 0.02), and NDIS ( P -interaction=0.04). A similar pattern was seen when we restricted the cohort to those with HGSC, suggesting the stage difference was not driven by histotype; slightly stronger associations among those with non-HGSC likely arose because this group was more likely to have early-stage disease (data not shown). There were no consistent differences when we stratified by residual disease, BMI, smoking, or physical activity. A stronger positive association was seen for DII among Blacks, but an inverse association among Asians, and slightly stronger associations for HEI and aHEI among the group with the least education, but these differences are potentially due to small numbers, especially for Black individuals ( n = 85). TABLE 2 Association between dietary indices and survival after a diagnosis of ovarian cancer. TABLE 2 Dietary index Model 1 N 2 Q1 Q2 Q3 Q4 P -trend 3 Reference HR (95% CI) 1 HR (95% CI) 1 HR (95% CI) 1 DII 4 1 6476 1.0 1.10 (1.00, 1.20) 1.05 (0.96, 1.15) 1.09 (0.99, 1.19) 0.17 2 6449 1.0 1.09 (0.99, 1.19) 1.03 (0.94, 1.12) 1.05 (0.96, 1.15) 0.5 NDIS 4 1 5468 1.0 0.99 (0.89, 1.09) 1.09 (0.99, 1.20) 1.12 (1.02, 1.24) 0.005 2 5443 1.0 0.97 (0.88, 1.07) 1.06 (0.96, 1.17) 1.06 (0.96, 1.18) 0.1 EDIP 4 1 5392 1.0 1.01 (0.91, 1.11) 1.04 (0.94, 1.15) 1.09 (0.98, 1.20) 0.1 2 5367 1.0 1.00 (0.90, 1.11) 1.03 (0.93, 1.14) 1.07 (0.96, 1.19) 0.2 HEI-2015 5 1 5986 1.0 0.94 (0.85, 1.03) 0.92 (0.83, 1.01) 0.90 (0.81, 0.99) 0.02 2 5959 1.0 0.96 (0.88, 1.06) 0.96 (0.87, 1.05) 0.95 (0.85, 1.05) 0.3 aHEI 5 1 4981 1.0 0.87 (0.78, 0.97) 0.90 (0.83, 1.01) 0.90 (0.81, 0.99) 0.2 2 4958 1.0 0.89 (0.80, 0.99) 0.95 (0.85, 1.05) 0.95 (0.85, 1.06) 0.7 WCRF 5 1 5390 1.0 1.00 (0.90, 1.10) 0.99 (0.89, 1.09) 0.98 (0.88, 1.09) 0.7 2 5365 1.0 1.02 (0.92, 1.13) 1.02 (0.93, 1.13) 1.02 (0.91, 1.13) 0.7 aMed 5 1 5986 1.0 0.91 (0.83, 1.00) 0.96 (0.87, 1.06) 0.97 (0.88, 1.06) 0.8 2 5959 1.0 0.95 (0.86, 1.05) 1.01 (0.91, 1.12) 1.01 (0.92, 1.12) 0.5 Abbreviations: aHEI, Alternate Healthy Eating Index; aMed, alternate Mediterranean diet; CI, confidence interval; DII, Dietary Inflammatory Index; EDIP, Empirical Dietary Inflammatory Pattern; HEI, Healthy Eating Index; HR, hazard ratio; MHT, menopausal hormone therapy; NCO, North Carolina Ovarian Cancer Study; NDIS, Novel Dietary Inflammation Score; NEC, New England Case–Control Study of Ovarian Cancer; Q, quartile; WCRF, World Cancer Research Fund. 1 HRs and 95% CIs from Cox proportional hazards models stratified by study, stage and histotype, and adjusted for (1) age and (2) age, race/ethnicity, education, smoking, BMI, menopausal hormone use, physical activity, and year of diagnosis. 2 Excludes MAL, NCO, POL, and those who died in the first year; see Supplemental Table 2 for a list of studies contributing to each index. 3 Assessed using a Wald test by assigning each level a number from 0 (Q1) to 3 (Q4) and modeling this as a continuous variable. 4 Q1 = least inflammatory category (that is, most healthy), Q4 = most proinflammatory (that is, least healthy). 5 Q1 = least adherent to dietary guidelines or recommendation (that is, least healthy), Q4 = most adherent (that is, most healthy). Association between dietary indices and survival after a diagnosis of ovarian cancer. Abbreviations: aHEI, Alternate Healthy Eating Index; aMed, alternate Mediterranean diet; CI, confidence interval; DII, Dietary Inflammatory Index; EDIP, Empirical Dietary Inflammatory Pattern; HEI, Healthy Eating Index; HR, hazard ratio; MHT, menopausal hormone therapy; NCO, North Carolina Ovarian Cancer Study; NDIS, Novel Dietary Inflammation Score; NEC, New England Case–Control Study of Ovarian Cancer; Q, quartile; WCRF, World Cancer Research Fund. HRs and 95% CIs from Cox proportional hazards models stratified by study, stage and histotype, and adjusted for (1) age and (2) age, race/ethnicity, education, smoking, BMI, menopausal hormone use, physical activity, and year of diagnosis. Excludes MAL, NCO, POL, and those who died in the first year; see Supplemental Table 2 for a list of studies contributing to each index. Assessed using a Wald test by assigning each level a number from 0 (Q1) to 3 (Q4) and modeling this as a continuous variable. Q1 = least inflammatory category (that is, most healthy), Q4 = most proinflammatory (that is, least healthy). Q1 = least adherent to dietary guidelines or recommendation (that is, least healthy), Q4 = most adherent (that is, most healthy). Table 3 shows that, among those with earlier stage disease (local or regional spread), those with scores in the higher quartiles of the HEI, aHEI, and aMed indices had better survival than those in the lowest quartile (adjusted HR Q4vQ1 0.75, 95% CI: 0.59, 0.97; 0.66, 95% CI: 0.50, 0.87 and 0.76, 95% CI: 0.60, 0.98, respectively). Conversely, survival was worse for those with the highest scores on the NDIS and EDIP inflammatory diet patterns. The associations between the healthy diet indices and survival were driven largely by higher intake of vegetables, fruit, nuts, and seeds and moderate alcohol intake (associated with higher aHEI and aMED scores but lower WCRF scores) ( Supplemental Figure 7 ). When we excluded alcohol from the indices that included this to focus specifically on the dietary components, the associations with DII, EDIP, and aMED were essentially unchanged but there was now a weak association with WCRF (HR Q4vQ1 0.80, 95% CI: 0.62, 1.02, P -trend = 0.1) and the association with aHEI was slightly attenuated (HR Q4vQ1 0.77, 95% CI: 0.59, 1.00, P -trend = 0.05). We had limited power to consider local and regional disease separately; however, the associations were generally stronger among those with localized disease ( Supplemental Table 2 ). TABLE 3 Association between dietary indices and survival among individuals with early-stage (local and regional) ovarian cancer. TABLE 3 Dietary index Model 1 N 2 Q1 Q2 Q3 Q4 P -trend 3 Reference HR (95% CI) 1 HR (95% CI) 1 HR (95% CI) 1 DII 4 1 2338 1.0 0.97 (0.78, 1.21) 0.91 (0.73, 1.13) 1.03 (0.82, 1.28) 0.99 2 2327 1.0 0.97 (0.78, 1.21) 0.89 (0.71, 1.11) 0.99 (0.79, 1.24) 0.7 NDIS 4 1 2000 1.0 0.93 (0.72, 1.21) 1.19 (0.93, 1.53) 1.38 (1.09, 1.76) 0.002 2 1989 1.0 0.91 (0.70, 1.18) 1.11 (0.86, 1.42) 1.25 (0.98, 1.61) 0.03 EDIP 4 1 2004 1.0 1.02 (0.79, 1.31) 1.18 (0.93, 1.51) 1.33 (1.04, 1.70) 0.01 2 1993 1.0 1.01 (0.78, 1.30) 1.19 (0.93, 1.54) 1.27 (0.98, 1.65) 0.03 HEI-2015 5 1 2184 1.0 0.95 (0.76, 1.20) 0.84 (0.67, 1.06) 0.71 (0.56, 0.91) 0.003 2 2173 1.0 0.95 (0.75, 1.19) 0.87 (0.68, 1.10) 0.75 (0.59, 0.97) 0.02 aHEI 5 1 1831 1.0 0.72 (0.57, 0.92) 0.70 (0.55, 0.90) 0.62 (0.47, 0.80) <0.001 2 1821 1.0 0.73 (0.57, 0.94) 0.72 (0.56, 0.93) 0.66 (0.50, 0.87) 0.004 WCRF 5 1 2002 1.0 0.90 (0.71, 1.15) 1.03 (0.82, 1.30) 0.85 (0.66, 1.09) 0.4 2 1991 1.0 0.93 (0.73, 1.19) 1.06 (0.83, 1.34) 0.89 (0.68, 1.15) 0.6 aMed 5 1 2184 1.0 0.73 (0.58, 0.92) 0.80 (0.63, 1.01) 0.71 (0.56, 0.90) 0.02 2 2173 1.0 0.76 (0.61, 0.96) 0.85 (0.66, 1.08) 0.76 (0.60, 0.98) 0.09 Abbreviations: aHEI, Alternate Healthy Eating Index; aMed, alternate Mediterranean diet; CI, confidence interval; DII, Dietary Inflammatory Index; EDIP, Empirical Dietary Inflammatory Pattern; HEI, Healthy Eating Index; HR, hazard ratio; MAL, Danish Malignant Ovarian Tumour Study; MCC, Melbourne Collaborative Cohort Study; NCO, North Carolina Ovarian Cancer Study; NDIS, Novel Dietary Inflammation Score; POL, Polish Ovarian Cancer Study; Q, quartile; WCRF, World Cancer Research Fund. 1 HRs and 95% CIs from models stratified by study and histotype, and adjusted for (1) age and (2) age, race/ethnicity, education, smoking, BMI, menopausal hormone use, physical activity, and year of diagnosis. 2 Excludes MAL, MCC (no stage data), NCO, POL, and those who died in the first year; see Supplemental Table 2 for list of studies contributing to each index. 3 Assessed using a Wald test by assigning each level a number from 0 (Q1) to 3 (Q4) and modeling this as a continuous variable. 4 Q1 = least inflammatory category (that is, most healthy), Q4 = most proinflammatory (that is, least healthy). 5 Q1 = least adherent to dietary guidelines or recommendation (that is, least healthy), Q4 = most adherent (that is, most healthy). Association between dietary indices and survival among individuals with early-stage (local and regional) ovarian cancer. Abbreviations: aHEI, Alternate Healthy Eating Index; aMed, alternate Mediterranean diet; CI, confidence interval; DII, Dietary Inflammatory Index; EDIP, Empirical Dietary Inflammatory Pattern; HEI, Healthy Eating Index; HR, hazard ratio; MAL, Danish Malignant Ovarian Tumour Study; MCC, Melbourne Collaborative Cohort Study; NCO, North Carolina Ovarian Cancer Study; NDIS, Novel Dietary Inflammation Score; POL, Polish Ovarian Cancer Study; Q, quartile; WCRF, World Cancer Research Fund. HRs and 95% CIs from models stratified by study and histotype, and adjusted for (1) age and (2) age, race/ethnicity, education, smoking, BMI, menopausal hormone use, physical activity, and year of diagnosis. Excludes MAL, MCC (no stage data), NCO, POL, and those who died in the first year; see Supplemental Table 2 for list of studies contributing to each index. Assessed using a Wald test by assigning each level a number from 0 (Q1) to 3 (Q4) and modeling this as a continuous variable. Q1 = least inflammatory category (that is, most healthy), Q4 = most proinflammatory (that is, least healthy). Q1 = least adherent to dietary guidelines or recommendation (that is, least healthy), Q4 = most adherent (that is, most healthy). In contrast, none of the dietary indices were associated with survival among those with advanced disease ( Supplemental Table 2 ). In the 5 studies with dietary scores and information about the cause of death, the associations between NDIS, EDIP, aHEI, WCRF, and ovarian cancer-specific survival were somewhat stronger than in the full cohort. Although there was no association between HEI-2015 or aMED and ovarian cancer-specific survival, the results for overall survival in this subset of studies were similar suggesting the differences were largely because of the restricted study sample, and not the end-point of ovarian cancer death ( Supplemental Table 3 ). When we censored follow-up at 5 y after diagnosis, the estimates for NDIS and EDIP were somewhat further from the null, and those for HEI and aHEI were closer ( Supplemental Table 3 ). In further sensitivity analyses, the associations were essentially unchanged when we included those who died during the first year but they were attenuated when we excluded the 19% of participants who completed the FFQ 1–2 y after diagnosis ( Supplemental Table 3 ). Exclusion of the cohort studies ( N = 31 from PLC with early-stage disease) made no difference to the results. As none of the dietary indices were associated with survival among those with advanced cancer, subsequent analyses focused on those with local or regional disease. The associations between food groups and survival among those with early-stage disease are shown in Table 4 . Most notable were the better survival with higher fruit (excluding fruit juice, HR Q4vQ1 0.78, 95% CI: 0.61, 0.99, P -trend 0.04) and vegetable (excluding potatoes) intake (HR Q4vQ1 0.71, 95% CI: 0.56, 0.91, P -trend 0.01), and particularly orange/yellow vegetables ( P -trend 0.001) and tomatoes ( P -trend 0.003). Better survival was also seen for the highest intake of nuts and seeds (0.71, 95% CI: 0.55, 0.92, P -trend 0.01). Survival was also better among those who consumed an average of >1 standard drink per day compared with no alcohol (HR: 0.77, 95% CI: 0.61, 0.97, P -trend 0.01), but this association was attenuated when we excluded the studies without information on energy intake (HR: 0.83, 95% CI: 0.65, 1.08, P -trend 0.1). Median alcohol intake in the highest group was 17 g alcohol per day, equivalent to 1.7 standard drinks (or 1.2 drinks in the United States where a standard drink contains 14 g alcohol). These food groups were consistent with those that contributed most to the associations with the diet quality scores above. TABLE 4 Association between food and beverage groups and survival among individuals with early-stage (local or regional) ovarian cancer 1 . TABLE 4 Food/food group Q1 Q2 HR (95% CI) 2 Q3 HR (95% CI) 2 Q4 HR (95% CI) 2 P -trend 3 Red meat 1.0 1.05 (0.84, 1.32) 1.02 (0.81, 1.28) 1.02 (0.79, 1.31) 1.0 Processed meat 1.0 1.12 (0.89, 1.40) 0.98 (0.76, 1.26) 0.98 (0.75, 1.27) 0.6 Poultry 1.0 0.97 (0.77, 1.21) 0.93 (0.75, 1.15) 0.82 (0.64, 1.05) 0.1 All fish 1.0 0.97 (0.77, 1.21) 1.04 (0.84, 1.29) 0.80 (0.63, 1.00) 0.1  Oily fish 1.0 0.92 (0.73, 1.15) 0.97 (0.78, 1.21) 0.84 (0.66, 1.08) 0.3  Other fish 1.0 0.96 (0.76, 1.21) 1.04 (0.83, 1.30) 0.90 (0.71, 1.15) 0.6 Eggs 1.0 1.13 (0.91, 1.41) 0.94 (0.76, 1.17) 0.94 (0.70, 1.25) 0.4 All dairy 1.0 1.00 (0.81, 1.25) 0.93 (0.74, 1.17) 1.06 (0.84, 1.34) 0.8  Reduced-fat dairy 1.0 1.18 (0.94, 1.48) 0.98 (0.78, 1.23) 0.94 (0.75, 1.18) 0.3  High-fat dairy 1.0 0.89 (0.71, 1.11) 0.95 (0.75, 1.19) 1.18 (0.93, 1.51) 0.2 All grains 1.0 1.23 (0.97, 1.58) 1.13 (0.87, 1.45) 1.05 (0.79, 1.40) 1.0  Whole grains 1.0 0.85 (0.68, 1.07) 0.96 (0.77, 1.21) 0.84 (0.66, 1.07) 0.3  Refined grains 1.0 1.11 (0.87, 1.41) 1.26 (0.98, 1.61) 1.01 (0.76, 1.32) 0.8 All fruits with juice 1.0 0.78 (0.61, 0.98) 0.70 (0.55, 0.89) 0.90 (0.70, 1.15) 0.3 All fruits without juice 1.0 0.78 (0.62, 0.98) 0.74 (0.59, 0.94) 0.78 (0.61, 0.99) 0.04  Citrus fruits 1.0 0.85 (0.67, 1.07) 0.91 (0.72, 1.15) 0.85 (0.66, 1.08) 0.3  Yellow/orange fruits 1.0 1.00 (0.79, 1.25) 0.74 (0.58, 0.95) 0.83 (0.64, 1.06) 0.03  Other fruits 1.0 0.88 (0.71, 1.10) 0.76 (0.61, 0.95) 0.86 (0.69, 1.09) 0.11  Juices 1.0 1.07 (0.85, 1.34) 0.89 (0.71, 1.13) 1.02 (0.81, 1.28) 0.8 All vegetables incl. potato 1.0 0.90 (0.72, 1.14) 0.88 (0.70, 1.12) 0.78 (0.60, 1.01) 0.06 All vegetables excl. potato 1.0 0.80 (0.64, 1.00) 0.83 (0.67, 1.04) 0.71 (0.56, 0.91) 0.01  Cruciferous vegetables 1.0 1.08 (0.86, 1.37) 0.99 (0.78, 1.25) 0.95 (0.74, 1.21) 0.5  Orange/yellow vegetables 1.0 0.86 (0.69, 1.06) 0.66 (0.52, 0.83) 0.78 (0.62, 0.98) 0.01  Leafy green vegetables 1.0 1.06 (0.86, 1.30) 0.80 (0.63, 1.00) 0.80 (0.63, 1.02) 0.01  Other vegetables 1.0 0.86 (0.69, 1.08) 0.85 (0.68, 1.07) 0.81 (0.63, 1.03) 0.1  Tomatoes 1.0 0.89 (0.72, 1.10) 0.77 (0.62, 0.97) 0.72 (0.57, 0.91) 0.003  Potato or tubers 1.0 0.85 (0.67, 1.08) 0.96 (0.76, 1.22) 1.05 (0.81, 1.34) 0.6 Soy food 1.0 0.86 (0.67, 1.12) 0.98 (0.75, 1.27) N/A 0.7 Legumes 1.0 0.91 (0.74, 1.13) 0.90 (0.71, 1.14) 0.96 (0.74, 1.24) 0.6 Nuts and seeds 1.0 0.95 (0.77, 1.19) 0.90 (0.72, 1.14) 0.71 (0.55, 0.92) 0.01 Sweets and sugary foods 1.0 1.00 (0.79, 1.28) 1.01 (0.79, 1.30) 1.01 (0.77, 1.32) 0.9 Sugar-sweetened drinks 1.0 0.96 (0.76, 1.21) 1.09 (0.86, 1.37) 0.94 (0.74, 1.20) 0.9 Ultraprocessed foods 1.0 0.91 (0.71, 1.16) 0.87 (0.68, 1.10) 1.05 (0.83, 1.34) 0.8 Alcoholic drinks 4 1.0 0.94 (0.73, 1.20) 0.81 (0.65, 0.99) 0.77 (0.61, 0.97) 0.01 Alcoholic drinks 4 , 5 1.0 1.03 (0.79, 1.34) 0.89 (0.71, 1.11) 0.83 (0.65, 1.08) 0.1 Abbreviations: CI, confidence interval; HR, hazard ratio; N/A, not applicable; Q, quartile. 1 Excludes those who died <1 y after diagnosis; numbers vary from 2040 to 2646 in individual models depending on the availability of data. 2 HRs and 95% CIs stratified by study and histotype and adjusted for age, log-energy, race, education, smoking, BMI, menopausal hormone use, physical activity, and year of diagnosis. 3 Assessed using a Wald test by assigning each level a number from 0 (Q1) to 3 (Q4) and modeling this as a continuous variable. 4 Categorized as 0, ≤1 standard drink/week, >1/wk to ≤1/d and >1/d; 1 standard drink ∼10 g alcohol. 5 Restricted to studies with information about energy intake; this restriction had little effect on the other food groups. Association between food and beverage groups and survival among individuals with early-stage (local or regional) ovarian cancer 1 . Abbreviations: CI, confidence interval; HR, hazard ratio; N/A, not applicable; Q, quartile. Excludes those who died <1 y after diagnosis; numbers vary from 2040 to 2646 in individual models depending on the availability of data. HRs and 95% CIs stratified by study and histotype and adjusted for age, log-energy, race, education, smoking, BMI, menopausal hormone use, physical activity, and year of diagnosis. Assessed using a Wald test by assigning each level a number from 0 (Q1) to 3 (Q4) and modeling this as a continuous variable. Categorized as 0, ≤1 standard drink/week, >1/wk to ≤1/d and >1/d; 1 standard drink ∼10 g alcohol. Restricted to studies with information about energy intake; this restriction had little effect on the other food groups. Results for those with advanced disease were generally null ( Supplemental Table 4 ); the only exceptions were a suggestion of worse survival associated with higher consumption of oily fish ( P -trend = 0.002) and better survival with higher dairy food intake ( P -trend = 0.047). Table 5 shows the associations between intake of nutrients from food sources and survival among those with early-stage disease. Better survival was seen for higher intakes of galactose (HR Q4vQ1 0.70, 95% CI: 0.52, 0.93, P -trend 0.02), magnesium (0.74, 95% CI: 0.58, 0.95, P -trend = 0.03), and beta-carotene (0.73, 95% CI: 0.57, 0.93, P -trend=0.01). Worse survival was seen with a higher intake of trans-fatty acids (1.40, 95% CI: 1.05, 1.88, P -trend = 0.02). There was no association with lycopene, despite the strong association seen with consumption of tomatoes which are a major contributor to dietary lycopene intake. These associations were not seen among those with advanced cancers ( Supplemental Table 5 ). TABLE 5 Associations between nutrient intake and survival among individuals diagnosed with early-stage (local or regional) ovarian cancer 1 . TABLE 5 Nutrient Q1 Q2 Q3 Q4 P -trend 3 Ref HR (95% CI) 2 HR (95% CI) 2 HR (95% CI) 2 Energy 1.0 1.08 (0.86, 1.36) 1.15 (0.92, 1.44) 1.11 (0.88, 1.39) 0.3 Total fiber 1.0 0.75 (0.60, 0.93) 0.87 (0.70, 1.09) 0.84 (0.67, 1.05) 0.3 Protein 1.0 0.93 (0.74, 1.15) 1.02 (0.82, 1.27) 0.90 (0.72, 1.13) 0.6 Carbohydrate 1.0 0.89 (0.72, 1.12) 0.90 (0.71, 1.13) 0.93 (0.75, 1.16) 0.6 Starch 1.0 1.06 (0.81, 1.38) 1.03 (0.79, 1.34) 1.07 (0.82, 1.41) 0.7 Sucrose 1.0 0.80 (0.62, 1.03) 0.84 (0.66, 1.07) 0.84 (0.66, 1.08) 0.2 Glucose 1.0 0.88 (0.67, 1.16) 0.72 (0.55, 0.95) 0.87 (0.66, 1.15) 0.2 Fructose 1.0 0.94 (0.74, 1.20) 0.92 (0.72, 1.19) 0.88 (0.69, 1.12) 0.3 Lactose 1.0 1.03 (0.80, 1.31) 1.01 (0.78, 1.31) 1.12 (0.88, 1.42) 0.4 Galactose 1.0 0.97 (0.73, 1.28) 0.90 (0.68, 1.19) 0.70 (0.52, 0.93) 0.02 Total sugar 1.0 0.95 (0.71, 1.25) 0.91 (0.68, 1.21) 0.86 (0.65, 1.14) 0.3 Glycemic index 1.0 1.11 (0.72, 1.70) 0.73 (0.47, 1.12) 1.15 (0.75, 1.75) 0.95 Glycemic load 1.0 0.76 (0.52, 1.11) 0.68 (0.46, 1.02) 0.92 (0.63, 1.33) 0.6 Fat 1.0 0.85 (0.67, 1.07) 1.01 (0.81, 1.26) 1.08 (0.87, 1.35) 0.3 Cholesterol 1.0 0.86 (0.69, 1.07) 1.03 (0.83, 1.28) 0.93 (0.74, 1.16) 0.9 Omega-3 fatty acids 1.0 0.95 (0.77, 1.19) 0.77 (0.61, 0.97) 1.04 (0.84, 1.29) 0.9 Omega-6 fatty acids 1.0 0.97 (0.75, 1.25) 0.85 (0.65, 1.10) 1.06 (0.82, 1.36) 0.9 Omega-3:6 ratio 1.0 1.08 (0.83, 1.41) 1.20 (0.93, 1.56) 1.11 (0.86, 1.44) 0.3 Polyunsaturated fats 1.0 0.86 (0.68, 1.07) 0.98 (0.79, 1.23) 0.99 (0.79, 1.23) 0.8 Monounsaturated fats 1.0 1.07 (0.83, 1.38) 1.05 (0.81, 1.36) 1.12 (0.87, 1.44) 0.4 Saturated fat 1.0 0.92 (0.73, 1.16) 0.95 (0.75, 1.19) 1.13 (0.90, 1.42) 0.3 Trans-fatty acids 1.0 1.19 (0.89, 1.59) 1.29 (0.96, 1.73) 1.40 (1.05, 1.88) 0.02 Calcium 1.0 0.95 (0.77, 1.19) 0.94 (0.75, 1.18) 0.95 (0.76, 1.19) 0.7 Iron 1.0 1.00 (0.80, 1.25) 1.00 (0.80, 1.26) 1.00 (0.80, 1.25) 1.0 Magnesium 1.0 0.92 (0.73, 1.16) 0.91 (0.72, 1.15) 0.74 (0.58, 0.95) 0.03 Selenium 1.0 0.97 (0.75, 1.25) 0.88 (0.68, 1.14) 1.03 (0.80, 1.33) 0.96 Sodium 1.0 0.90 (0.72, 1.12) 0.79 (0.63, 1.00) 1.02 (0.82, 1.27) 0.9 Zinc 1.0 0.91 (0.73, 1.13) 0.90 (0.72, 1.12) 0.93 (0.75, 1.16) 0.5 Vitamin A 1.0 0.96 (0.77, 1.19) 0.90 (0.72, 1.12) 0.83 (0.66, 1.04) 0.08 Beta-carotene 1.0 0.99 (0.79, 1.25) 0.92 (0.73, 1.17) 0.73 (0.57, 0.93) 0.01 Vitamin B1 (Thiamin) 1.0 1.00 (0.80, 1.25) 0.95 (0.75, 1.19) 1.02 (0.82, 1.28) 0.97 Vitamin B2 (Riboflavin) 1.0 0.99 (0.79, 1.24) 0.88 (0.70, 1.11) 1.09 (0.87, 1.36) 0.7 Vitamin B3 (Niacin) 1.0 0.85 (0.67, 1.07) 0.82 (0.65, 1.03) 0.85 (0.67, 1.07) 0.1 Vitamin B5 1.0 0.80 (0.63, 1.01) 0.75 (0.59, 0.95) 0.80 (0.63, 1.01) 0.05 Vitamin B6 1.0 0.82 (0.66, 1.03) 0.90 (0.72, 1.11) 0.84 (0.67, 1.06) 0.2 Food folate (B9) 1.0 0.93 (0.72, 1.20) 0.76 (0.58, 0.99) 0.91 (0.70, 1.19) 0.3 Folacin from foods 1.0 1.21 (0.91, 1.63) 1.10 (0.82, 1.47) 1.05 (0.78, 1.42) 0.9 Dietary folate equiv. 1.0 1.09 (0.85, 1.39) 0.70 (0.53, 0.93) 0.88 (0.67, 1.14) 0.06 Vitamin B 12 1.0 1.01 (0.81, 1.26) 0.92 (0.73, 1.15) 0.97 (0.78, 1.21) 0.6 Vitamin C 1.0 0.73 (0.58, 0.92) 0.83 (0.66, 1.03) 0.75 (0.60, 0.95) 0.05 Vitamin D 1.0 1.05 (0.84, 1.30) 0.96 (0.77, 1.21) 1.04 (0.83, 1.30) 0.9 Vitamin E 1.0 0.90 (0.72, 1.13) 0.80 (0.64, 1.00) 0.93 (0.75, 1.15) 0.3 Lycopene 1.0 0.88 (0.68, 1.13) 0.96 (0.75, 1.23) 1.06 (0.83, 1.35) 0.9 Isoflavones 1.0 1.18 (0.87, 1.60) 1.00 (0.73, 1.37) 1.09 (0.79, 1.49) 0.9 Total phytoestrogens 1.0 1.02 (0.78, 1.33) 0.93 (0.71, 1.22) 0.90 (0.68, 1.19) 0.4 Abbreviations: CI, confidence interval; HR, hazard ratio; Q, quartile. 1 Excludes those who died <1 y after diagnosis [sample size ranges from 1261–2282, except glycemic index (839) and glycemic load (1037)]. 2 HR) and 95% CI from models stratified by study and histotype and adjusted for age, log-energy, race, education, smoking, BMI, menopausal hormone use, physical activity, and year of diagnosis. 3 Assessed using a Wald test by assigning each level a number from 0 (Q1) to 3 (Q4) and modeling this as a continuous variable. Associations between nutrient intake and survival among individuals diagnosed with early-stage (local or regional) ovarian cancer 1 . Abbreviations: CI, confidence interval; HR, hazard ratio; Q, quartile. Excludes those who died <1 y after diagnosis [sample size ranges from 1261–2282, except glycemic index (839) and glycemic load (1037)]. HR) and 95% CI from models stratified by study and histotype and adjusted for age, log-energy, race, education, smoking, BMI, menopausal hormone use, physical activity, and year of diagnosis. Assessed using a Wald test by assigning each level a number from 0 (Q1) to 3 (Q4) and modeling this as a continuous variable.

Our findings support a possible association between prediagnosis diet and survival after a diagnosis of early-stage ovarian cancer, but we saw little association among those with advanced disease. This may be because the disease has progressed further in these individuals and so is affected less by dietary factors, or they are more likely to have changed their diet because of symptoms experienced before diagnosis or to change it after diagnosis. Specifically, we found that for those with early-stage disease, a healthier diet (measured by HEI, aHEI, aMED; or WCRF when alcohol, which counts negatively to WCRF scores, is not considered) or a less-inflammatory diet (measured by NDIS or EDIP, particularly during the first 5 y after diagnosis) was associated with better survival. Consistent with this, higher consumption of “healthy” foods (that is, fruits, vegetables, nuts, and seeds) was also associated with better survival. There was no evidence of variation by histotype or by menopausal status, education level, smoking, BMI, or physical activity. We saw few associations with individual nutrients, but the better survival associated with a higher intake of galactose (found in dairy foods and fruit and vegetables), magnesium, and beta-carotene and worse survival with trans-fats is consistent with a beneficial effect of a healthier diet. Our observation that, overall, few individual dietary components were strongly associated with survival supports previous suggestions that the overall pattern of food consumption is more important than single foods and nutrients [ 6 ]. We also saw better survival among those with higher alcohol consumption, but this association was attenuated when we excluded studies without information on energy intake. Of note, alcohol intake in this sample was generally low: 81% consumed less than an average of 1 standard drink per day; median intake among those consuming more than this was 17 g alcohol (1.7 standard drinks) per day. We cannot therefore draw any conclusions about the potential effects of higher alcohol intake on survival. Two potential mechanisms through which diet might influence cancer outcomes include its known effects on inflammation [ 36 ] and immunity [ 37 ]—2 processes integral to the development and progression of cancer [ 38 ]. Early studies examining the association between diet quality and survival among individuals with ovarian cancer gave mixed results, although most did not stratify by stage of disease. In the United States, higher diet quality, assessed using the HEI-2005 ≥12 mo before diagnosis, was associated with better survival in the Women’s Health Initiative ( N = 636 cases) [ 3 ]. In contrast, there was no association between survival and aHEI scores measured before or 1–4 y after diagnosis in the Nurses’ Health Studies (1003 cases) [ 39 ], or between a variety of diet quality indices before and after diagnosis in an Australian study (OPL, 650 cases, also included here) [ 40 ]. However, more recent reports have been more consistent. Better survival was associated with higher aHEI and aMED scores, a median of 20 y before diagnosis in the NIH-AARP cohort (1107 cases) [ 41 ] and with prediagnosis HEI and aHEI scores (measured after diagnosis) in the Black Women’s Study (483 cases) [ 42 ]. A Chinese study (550–703 cases) has also reported markedly better survival (HR ∼0.5) with higher diet quality measured using a number of scales [ [43] , [44] , [45] , [46] , [47] ]. Three previous studies have investigated the association between the inflammatory potential of diet and survival after a diagnosis of ovarian cancer. The first ( n = 857), which is also included in the current analysis (AUS), found no association with DII or EDIP scores but did not consider the stage of disease [ 48 ]. In contrast, Sasamoto et al. reported worse survival associated with a higher EDIP score (indicating greater inflammatory potential) among those with non-HGSC (who would be more likely to have early-stage disease) in the Nurses’ Health Studies [ 39 ]. Peres et al. have also reported an association between higher DII scores and worse survival, but this was restricted to HGSC, particularly smokers [ 49 ]. It should be noted that that study was conducted among African-American women ( n = 490) whose dietary patterns may differ from the predominantly white populations included in this analysis and the median follow-up was only 3.5 y. Evidence regarding the relationship among food groups, individual food items, nutrients, and survival after a diagnosis of ovarian cancer is also limited. Our results are in agreement with previous studies suggesting that prediagnosis diets higher in fruit and vegetables [ 4 , 5 ] could be associated with better survival, but we did not confirm reports of better survival with higher intake of fiber and fish [ 6 , 50 , 51 ] or worse survival with higher intakes of red and processed meats and dairy products [ [4] , [5] , [6] , 51 , 52 ]. Strengths of the current study include its large sample size and detailed information on potential confounders. Furthermore, the inclusion of studies from the United States, Australia, and Europe suggests the results would be applicable to most high-income countries. Limitations are that individuals diagnosed with more advanced disease were likely underrepresented (as shown by the low mortality at 1 y) and the dietary information related only to the period before diagnosis and so may not reflect intake after diagnosis. From a health promotion perspective, the primary interest is whether diet after diagnosis can influence survival but there are currently insufficient data to reliably assess this. Although some people change their diet after a diagnosis of cancer, overall, evidence suggests the changes are fairly modest [ 6 , 53 , 54 ]; thus, identification of associations based on prediagnosis diet can provide valuable information about potentially relevant dietary components. We did not include intake of nutrients from supplements in these analyses because their use is not recommended unless advised by a health professional [ 9 ]. Although some degree of measurement error in self-reported diet is inevitable, this is unlikely to have varied systematically between those with good and poor survival and, in this cohort, those with more advanced disease were, if anything, more likely to report higher diet quality. Any error is thus likely to have attenuated associations between the highest and lowest levels of intake. Although it is also possible that the observed associations are because of unknown or unmeasured confounding factors, we adjusted for key factors known to be associated with diet and to influence survival and this had little impact on our results. We cannot rule out the possibility of residual confounding by socioeconomic factors that might affect access and adherence to treatment; however, the patterns were generally similar in the United States studies and those from Australia where all residents are entitled to free healthcare. Finally, it is important to acknowledge that we conducted a large number of analyses; thus, some associations may have arisen by chance. However, among those with early-stage disease, the overall consistency between our results for the dietary indices and the individual foods and nutrients that contribute to these adds weight to the belief that the associations may be real. Maintaining healthy lifestyle behaviors is consistent with current recommendations for improved quality of life and prevention of comorbidities among cancer survivors. However, there is currently little evidence to recommend lifestyle changes to ovarian cancer survivors specifically for the purpose of improving survival outcomes. Our results suggest a more healthy, less-inflammatory diet may confer a modest survival benefit for individuals with localized or regional ovarian cancer, and support the need for further prospective research assessing the effect of diet after diagnosis on survival.

Ovarian cancer is the eighth most common cause of female cancer death worldwide and, although survival has slowly improved over time, the overall 5-y survival rate remains below 50%, even in high-income countries [ 1 , 2 ]. Diet is a potentially modifiable factor that might influence survival. To date, observational studies have linked better survival outcomes to higher diet quality [ 3 ], intake of fruit and vegetables [ [4] , [5] , [6] ], fiber [ 6 ], and soybean curd [ 7 ] before a diagnosis of ovarian cancer. Conversely, worse outcomes have been reported among those who, before their ovarian cancer diagnosis, consumed more red and cured/processed meat [ 4 ], milk [ 4 ], lactose, calcium and dairy products [ 5 ], salted fish and Chinese cabbage [ 7 ], or folate and folic acid [ 8 ]. Higher glycemic index has also been associated with worse survival [ 6 ]. However, results are not consistent across studies and limited sample sizes have prevented well-powered analyses of the potential heterogeneity of these associations by clinical and patient characteristics. Given the poor prognosis associated with a diagnosis of ovarian cancer, patients often want to know what they can do to improve this. The World Cancer Research Fund (WCRF) recommends that once treatment is finished, cancer survivors should, as far as possible, aim to have a healthy diet and limit alcohol consumption in addition to maintaining a healthy weight and being physically active [ 9 ]. There are currently no specific dietary guidelines for ovarian cancer survivors. In this context, our aim was to conduct a comprehensive analysis of the association between indices of diet quality and intake of individual foods and nutrients from foods before diagnosis and survival after a diagnosis of ovarian cancer, using data from a large international consortium.

PMW has received funding from AstraZeneca for an unrelated study of ovarian cancer. EVB has served on an Advisory Board for Pfizer to enhance minoritized and underrepresented populations in clinical trials, unrelated to this study. All other authors report no conflicts of interest.

Data described in the manuscript, code book, and analytic code cannot be made available because of privacy and ethical reasons. Contact the corresponding author to discuss access to data through existing data request processes for OCAC.