Plasma 25-Hydroxyvitamin D Levels and Survival in Stage III Colon Cancer: Findings from CALGB/SWOG 80702 (Alliance)

This article is featured in Highlights of This Issue, p. 2559

Vitamin D is predominantly obtained from dietary intake, sunlight exposure, and supplement intake, and is hypothesized to reduce the risk of cancer progression through proapoptotic, antiproliferative, and antiangiogenic actions (1). However, based on current findings in humans, the U.S. Preventive Services Task Force concluded that there is inadequate evidence on the benefit of treating asymptomatic vitamin D deficiency on cancer outcomes, and encouraged more research on the role of vitamin D in chronic diseases (2). Vitamin D deficiency is common among patients with colorectal cancer, which is the second most common cause of cancer-related death in the United States (3–5). Approximately 36% of patients with colorectal cancer are diagnosed with locally advanced disease (mainly stage III; ref. 6). Our previous study showed that higher predicted vitamin D status (derived using diet and lifestyle questionnaires) was associated with decreased recurrence and improved survival among stage III cancer (7). A recent study reported an improved disease-free survival (DFS) and time to recurrence (TTR) among men with the highest tertile of measured serum 25-hydroxyvitamin D [25(OH)D)] level (8, 9). Larger studies with measured plasma 25(OH)D levels are therefore warranted.

Compelling evidence suggests that vitamin D may improve cancer prognosis through reducing inflammation (10), and higher inflammation status measured by C-reactive protein (CRP), IL6, and soluble TNF receptor 2 (sTNF-R2) is associated with worse colorectal cancer survival (11–15). In this study, we prospectively assessed the association between plasma 25(OH)D levels and clinical outcomes of patients with stage III colon cancer who were enrolled in a completed phase III randomized clinical trial of adjuvant chemotherapy and also explored the role of circulating inflammatory cytokines in mediating this association.

This prospective cohort was derived from the NCI-sponsored phase III randomized trial (CALGB/SWOG 80702) in patients with stage III colon cancer who had undergone curative-intent resection. Patients were randomized to treatment with six versus 12 cycles of adjuvant 5-fluorouracil, leucovorin, and oxaliplatin (FOLFOX) with 3 years of celecoxib versus placebo (16, 17). A detailed trial description can be found in the Supplementary Materials and Methods. Briefly, from June 2010 to November 2015, a total of 2,524 patients with colon cancer were enrolled in the treatment trial. All trial patients were invited to participate in companion studies. A total of 1,498 (59%) patients participated in the plasma companion biomarker study and had blood drawn before the planned treatment. And 1,892 (75%) patients participated in the diet and lifestyle companion study and completed a semiquantitative food frequency questionnaire (FFQ) within the first 6 weeks of randomization (FFQ1) and again during the 14–16 months after randomization (FFQ2). The FFQ has been externally validated and collected information on 131 food items, vitamin and supplement use, and physical activity (18).

A detailed flow diagram for patient enrollment and exclusion for the current study is depicted in Supplementary Fig. S1. The final population included 1,437 patients in the main analysis of plasma 25(OH)D, and 1,643 patients in the sensitivity analysis of predicted vitamin D level, which may better reflect long-term vitamin D status. No appreciable differences were detected between these two companion study populations and the overall treatment trial population with regard to baseline characteristics and patient outcome, with a median survival of 8.6 years in all three groups (Supplementary Table S1). All patients provided written informed consent for trial participation. This study was conducted in accordance with ethical guideline of U.S. Common Rule, and Institutional Review Board (IRB) approval was obtained at each participating center.

The primary exposure was baseline plasma 25(OH)D. Plasma samples were collected by individual centers and sent on ice to the CALGB Pathology Coordinating Lab (Ohio State University, Columbus, OH). The median time from randomization to blood collection was 0.2 months (range, 0–5.9 months). Plasma 25(OH)D concentrations were measured by radioimmunoassay at Heartland Assays, with a mean intra-assay coefficient of variation of 4% (19). All laboratory personnel were blinded to the clinical outcomes.

A predicted vitamin D score was calculated using previously validated regression coefficients (7, 20, 21). Information on race and geographic region from FFQ1 was applied directly, and we used cumulative averaging values of physical activity, body mass index (BMI), and dietary and supplementary vitamin D intake from both FFQ1 and FFQ2, with weighted proportion to times between FFQ1 and FFQ2 and then between FFQ2 and DFS time (7). Median time from randomization to FFQ1 was 0.6 months (range, 0–5.8 months). A full description of the covariates and plasma measurement of CRP, IL6, and sTNF-R2 is provided in the Supplementary Materials and Methods. The mean day-to-day intra-assay coefficient of variability was 3% for CRP, 8% for IL6, and 4% for sTNF-R2.

The primary study endpoints were DFS (calculated from randomization to disease recurrence or death from any cause), overall survival (OS; calculated from randomization to death from any cause), and TTR (calculated from randomization to disease recurrence). Colon cancer–specific mortality was also evaluated (calculated from randomization to death from colon cancer). Adverse events were graded using NCI Common Terminology Criteria for Adverse Events version 4.0 (16). All patients were followed up through August 10, 2020.

There is no consensus on the definition of vitamin D deficiency, sufficiency, and excess (22). This study used the cut-off points recommended by the National Academy of Medicine: vitamin D deficiency was defined as plasma 25(OH)D < 12 ng/mL (1 ng/mL = 2.5 nmol/L) and nondeficiency as ≥12 ng/mL, while nondeficiency was further subcategorized as inadequacy (12–19 ng/mL), sufficiency (20–29 ng/mL), and beyond sufficiency (≥30 ng/mL; refs. 23–25). Kaplan–Meier curves were calculated for DFS, OS, and TTR, and compared according to plasma 25(OH)D levels using the log-rank test. Cox proportional hazards models were conducted to assess the HRs for DFS, OS, and TTR and their 95% confidence intervals (CI) comparing predefined plasma 25(OH)D levels, adjusting for a priori defined covariates, including age (years), sex (female or male), race (White, Black, Others), BMI (<25, 25–30, >30 kg/m²), season of blood collection (winter/spring or summer/autumn), Eastern Cooperative Oncology Group (ECOG) performance status (0 or 1/2), tumor location (left-sided, right-sided, multiple), tumor stage (T1/2 or T3/4), nodal stage (N1 or N2), and assigned treatment arms (six or 12 cycles of FOLFOX with or without celecoxib). Multivariable-adjusted HRs of DFS, OS, and TTS by quartiles of predicted vitamin D scores were also computed. Assumptions of proportional hazards were tested and met by the product of 25(OH)D levels and the natural logarithm of time. Nonlinear trends were assessed with likelihood-ratio tests of restricted cubic splines (26). Cause-specific Cox regression was applied to estimate colon cancer–specific mortality by treating noncancer deaths as a competing event. Stratification analyses were conducted by covariates of interest, and sensitivity analyses were done by excluding patients with tumor recurrence or death within 3 months of blood collection, and by including participants enrolled both in the biomarker study and the diet and lifestyle study. Using multivariable-adjusted logistic regression, the OR of significant adverse events (≥ grade 2) by plasma 25(OH)D levels was estimated.

To examine the proportion of the association of plasma 25(OH)D (modeled as continuous variable per 10 ng/mL increase) and recurrence and death that is due to circulating inflammatory biomarkers, mediation analyses were conducted using the difference method to estimate primary multivariable-adjusted HRs with and without adjustment for each specific biomarker (27, 28). Percent of mediation was calculated by β coefficients from two models using the following formula: (β_{biomarker-unadjusted} − β_{biomarker-adjusted})/β_{biomarker-unadjusted}.

The Alliance Statistics and Data Management Center collected all data and confirmed data quality following Alliance policies (data locked on October 30, 2022). Statistical analyses were conducted using SAS version 9.4 (SAS Institute Inc). A two-sided P value of <0.05 was considered statistically significant.

Deidentified patient data may be requested from Alliance for Clinical Trials in Oncology via [email protected] if data are not publicly available. A formal review process includes verifying the availability of data, conducting a review of any existing agreements that may have implications for the project, and ensuring that any transfer is in compliance with the IRB. The investigator will be required to sign a data release form prior to transfer. The code is available upon proper request.

Among 1,437 patients with stage III colon cancer, the median plasma 25(OH)D level was 21.0 ng/mL (range, 5.7–96.1 ng/mL). Vitamin D deficiency was present in 182 patients (13%; Table 1). Vitamin D deficiency was more prevalent among Black patients (32%) than White patients (10%). Compared with patients with deficient vitamin D, patients with higher plasma 25(OH)D tended to be older, more physically active, and were more likely to have normal BMI, better ECOG performance status, and higher calcium and vitamin D intake. The study participants generally represent well the patients with stage III colon cancer in the United States (Supplementary Table S2).

During a median follow-up time of 5.8 years, a total of 373 (26%) patients died or had a recurrence. Compared with vitamin D–deficient patients, nondeficient patients had significantly improved DFS (P_log-rank = 0.02, Fig. 1A), OS (P_log-rank = 0.002, Fig. 1B), and TTR (P_log-rank = 0.049, Fig. 1C). After adjusting for potential confounders, patients with nondeficient vitamin D continued to have significantly improved DFS (HR, 0.68; 95% CI, 0.51–0.92), OS (HR, 0.57; 95% CI, 0.40–0.80), and TTR (HR, 0.71; 95% CI, 0.52–0.98; Table 2). Results were similar after excluding patients who died or recurred within 3 months after blood draw (Supplementary Table S3). A significant U-shaped dose–response relationship of 25(OH)D was found for DFS (P_nonlinearity = 0.04, Fig. 2A) and OS (P_nonlinearity = 0.03, Fig. 2B), but not for TTR (P_nonlinearity = 0.87, Fig. 2C). Compared with those with deficient vitamin D, patients with plasma 25(OH)D of 12 to 50 ng/mL had improved DFS and OS, but not those with higher levels (>50 ng/mL), though the sample size was small (n = 25).

Among patients who participated in the diet and lifestyle companion study, the predicted vitamin D score ranged from 16.3 to 36.4 ng/mL. Improved survival outcomes were seen among patients in the higher quartiles of predicted vitamin D scores (Supplementary Table S4), with a linear trend observed for DFS (P_linearity = 0.02) and OS (P_linearity = 0.0002; Supplementary Fig. S2). Similar findings were observed when the analysis was restricted to patients who participated in both the diet and lifestyle and plasma biomarker companion studies (Supplementary Table S4). A statistically significant Spearman correlation between measured plasma 25(OH)D levels and predicted vitamin D scores was observed (r2 = 0.34, P < 0.0001).

In subgroup analyses, associations between plasma 25(OH)D levels and DFS, OS, and TTR were largely unchanged across strata of clinically relevant covariables (all P_interaction >0.05; Supplementary Fig. S3A–S3C). Colon cancer–specific mortality was also lower in patients with nondeficient plasma 25(OH)D levels (HR, 0.59; 95% CI, 0.39–0.90) and patients with higher predicted vitamin D levels (P_trend = 0.02; Supplementary Table S5).

There was no significant association between 25(OH)D levels and the occurrence of the most common ≥ grade 2 adverse events, including diarrhea, fatigue, hypertension, nausea, neutropenia, peripheral neuropathy, and thrombocytopenia (Table 3). However, in the subgroup treated with celecoxib, patients with plasma 25(OH)D of ≥ 12 ng/mL had a lower risk of fatigue (OR, 0.57; 95% CI, 0.33–0.97) and hypertension (OR, 0.55; 95% CI, 0.29–1.04), compared with those with plasma 25(OH)D < 12 ng/mL.

Further adjustment of the primary multivariable model with sTNF-R2 slightly attenuated the HRs of DFS and OS, with 10.6% (P_{mediation =} 0.04) of the association with DFS, and 11.8% (P_{mediation =} 0.05) with OS, but not with TTR (8.4%, P_mediation = 0.13; Table 4). HRs of DFS, OS, and TTR by plasma 25(OH)D were basically unaltered after adjustment of CRP or IL6 and both these inflammatory biomarkers did not mediate the association of plasma 25(OH)D.

This prospective cohort study of patients with stage III colon cancer showed that nondeficient plasma 25(OH)D levels were associated with lower cancer recurrence and improved survival, and the association between 25(OH)D and survival may potentially follow a U-shaped pattern with survival benefits seen among patients with 25(OH)D levels of 12 to 50 ng/mL. We also observed that the protective effect of higher vitamin D was independent of inflammatory cytokines of CRP and IL6 while being marginally mediated by sTNF-R2. Plasma 25(OH)D levels were not significantly associated with the occurrence of ≥ grade 2 adverse events, although a possible lower risk of fatigue and hypertension may exist with higher plasma 25(OH)D levels among patients treated with celecoxib. To our knowledge, this is the largest study to characterize the relationship between plasma 25(OH)D levels and outcomes among patients with stage III colon cancer.

Higher plasma 25(OH)D levels were significantly associated with improved survival, which is consistent with findings using the predicted vitamin D score in a previous trial study of stage III colon cancer (7), and also in the current study. Sinicrope and colleagues reported an improved DFS and TTR by higher plasma 25(OH)D levels only among male patients with stage III colon cancer but not all patients (8, 9). The current study had a relatively wide range of plasma 25(OH)D levels, which allowed us to observe a significant U-shaped association between plasma 25(OH)D and survival among the entire study population. However, this U-shaped relationship currently should be considered hypothesis-generating only, because the analysis may be underpowered for patients with plasma 25(OH)D levels of >50 ng/mL. Further studies are needed to confirm whether a true negative association exists between high plasma 25(OH)D and survival among patients with stage III colon cancer. Several other observational studies also evaluated the relationship between 25(OH)D levels and patient outcomes among patients with any stage of colon and rectal cancer; however, conflicting results were reported and none of them were adequately powered to look at patients with stage III colon cancer specifically (4, 29, 30, 31). The U-shaped association of DFS and OS by 25(OH)D levels found in the current study may partly explain why conflicting results were seen in previous reports that used different cut-off points to define high and low 25(OH)D levels. Interestingly, we did not find a U-shaped association between predicted vitamin D scores and survival outcomes; this could potentially be attributed to the much narrower range of predicted vitamin D levels. Moreover, predicted vitamin D scores were derived from two serial questionnaires, which indirectly reflect relative levels of vitamin D status, and it is also possible that diet, physical activity, or supplementary vitamin D use could have changed after chemotherapy and led to variations in predicted vitamin D levels.

In comparison with the general population, patients with colon cancer tend to have a higher prevalence of vitamin D deficiency (3, 24). Potential reasons for this difference include the fact that patients with cancer or patients with more advanced tumor stage are likely to be older and have less sun exposure, lower dietary vitamin D intake, reduced physical activity, and increased obesity compared with the general population. We also noticed that Black patients had a higher prevalence of vitamin D deficiency (32%) compared with White patients (10%), which could partly explain the worse survival among Black patients in previous publications (5, 32). However, in this study, both Black and White patients with higher vitamin D had improved survival compared with their counterparts with lower vitamin D levels, suggesting the importance of detecting and correcting vitamin D deficiency. Higher plasma 25(OH)D levels were mainly due to a higher intake of supplemental vitamin D, which supports future investigation of adjuvant supplement use among patients with stage III colon cancer.

Our results are supported by compelling preclinical evidence showing that vitamin D may inhibit colon cancer progression through several signaling pathways (1). The active form of vitamin D, 1,25-dihydroxyvitamin D, binds to the vitamin D receptor–retinoid X receptor complex to regulate gene expression, leading to induction of apoptosis, cell differentiation, inhibition of cell proliferation, invasion, metastasis, and angiogenesis (1). Vitamin D also inhibits the production of proinflammatory cytokines and demonstrates anti-inflammatory actions in cancers (10). The current study did not find strong evidence that CRP or IL6 confounded or mediated associations between 25(OH)D and colon cancer survival, suggesting that the reported association of 25(OH)D could be independent of CRP or IL6 levels. Notwithstanding, we observed sTNF-R2 as a potential mediator in the causal pathway between vitamin D and survival, indicating potential other mechanisms. Previous studies showed that TNF-R2 not only directly promoted cancer proliferation, activated immunosuppressive cells, and supported immune escape (33), but may also work together with vitamin D to coregulate extracellular signal–regulated kinase (ERK) and NFκB signaling in colon cancer, which may explain the marginal mediation effect of TNF-R2 in the current study (1, 33, 34). These findings suggest that clinical interventions to increase plasma 25(OH)D levels or to decrease circulation inflammations may benefit patients with colon cancer in the future. Further investigation of sTNF-R2 in colorectal cancer pathogenesis and underlying mechanisms of action, as well as interactions with vitamin D signaling, is warranted.

The strengths of this study include a large study population from a rigorously conducted multicenter phase III randomized clinical trial with uniform treatment and follow-up. Detailed data on lifestyle and clinical characteristics were prospectively collected using a validated instrument. In addition, direct plasma measurement of 25(OH)D levels provides reliable vitamin D status at the study baseline before chemotherapy, while a predicted vitamin D score could be more representative of longer-term vitamin D status among patients with colon cancer before and after the chemotherapy. The reduced recurrence and improved survival among patients with higher vitamin D levels measured by both direct assay and predicted vitamin D score further enhance the validity and biological plausibility of the study results.

However, several limitations deserve attention. It is possible that higher plasma 25(OH)D levels may be acting as a surrogate for other favorable prognostic factors, although higher vitamin D levels continued to be significantly associated with improved survival after adjusting for several of these factors in multivariable analyses. Plasma 25(OH)D was measured only once at the study baseline; therefore, the impact of changes in 25(OH)D on patient outcome was not able to be evaluated. But similar results were found using a predicted vitamin D score that may better reflect longstanding vitamin D status. Because our study population was comprised of patients enrolled in a clinical trial, the generalizability of our results may be limited, and findings may not apply to patients with colon cancer at large. Nevertheless, the relatively wide range of 25(OH)D levels in the cohort diminishes this concern and our patients were recruited from both community and academic centers throughout North America. Finally, the majority of the patients were of White race; additional studies focusing on other races are encouraged, particularly because disparities exist in vitamin D status and colorectal cancer incidence and mortality by race and ethnicity (35, 36).

In conclusion, this study found that nondeficient vitamin D levels are significantly associated with decreased recurrence and improved survival in patients with stage III colon cancer, independent of inflammation cytokines of CRP or IL6. This survival benefit may be mediated in part through sTNF-R2. Additional research to understand the mechanisms of vitamin D activity is needed. Given the promising results from the current study, a randomized clinical trial of adjuvant vitamin D supplementation in patients with stage III colon cancer should be considered to confirm causality.

Q. Shi reports personal fees from Regeneron Pharmaceuticals, Inc., Kronos Bio, Yiviva Inc., Hoosier Cancer Research Network, and Mirati Therapeutics Inc; grants from Janssen, BMS, Genetech, and Novartis outside the submitted work. C.S. Fuchs reports other support from Genentech/Roche and Evolveimmune Therapeutics outside the submitted work. P. Kumthekar reports other support from Roche, Mirati, Novocure, Clario, Enclear, Biocept, Janssen, and Seagen outside the submitted work. K.A. Guthrie reports grants from NIH/NCI during the conduct of the study. C. Blanke reports grants from NCI during the conduct of the study. E.M. O'Reilly reports grants from Genentech/Roche, BioNTech, AstraZeneca, Arcus, Elicio Therapeutics, Parker Institute, NIH/NCI, and Pertzye; personal fees from Boehringer Ingelheim, BioNTech, Ipsen, Merck, Novartis, AstraZeneca, BioSapien, Astellas, Thetis, Autem, Neogene, BMS, Tempus, Fibrogen, and Merus; other support from Agios, Genentech-Roche, and Eisai outside the submitted work. J.A. Meyerhardt reports personal fees from Merck Pharmaceutical outside the submitted work. K. Ng reports grants from NCI during the conduct of the study; non-financial support from Pharmavite; grants from Evergrande Group, Janssen, and Revolution Medicines; personal fees from Bayer, SeaGen, BiomX, Bicara Therapeutics, GlaxoSmithKline, Pfizer, X-Biotix Therapeutics, and Redesign Health outside the submitted work. No disclosures were reported by the other authors.

Q.-L. Wang: Conceptualization, data curation, software, formal analysis, investigation, visualization, methodology, writing–original draft, writing–review and editing. C. Ma: Data curation, software, formal analysis, validation, methodology, project administration, writing–review and editing. C. Yuan: Data curation, software, validation, investigation, methodology, writing–review and editing. Q. Shi: Resources, software, validation, investigation, methodology, writing–review and editing. B.M. Wolpin: Supervision, funding acquisition, investigation, writing–review and editing. Y. Zhang: Software, validation, investigation, methodology, writing–review and editing. C.S. Fuchs: Supervision, investigation, project administration, writing–review and editing. J. Meyer: Validation, project administration, writing–review and editing. T. Zemla: Validation, project administration, writing–review and editing. E. Cheng: Data curation, investigation, writing–review and editing. P. Kumthekar: Resources, funding acquisition, investigation, project administration, writing–review and editing. K.A. Guthrie: Resources, funding acquisition, investigation, project administration, writing–review and editing. F. Couture: Resources, funding acquisition, investigation, project administration, writing–review and editing. P. Kuebler: Resources, funding acquisition, investigation, project administration, writing–review and editing. P. Kumar: Resources, funding acquisition, investigation, project administration, writing–review and editing. B. Tan: Resources, funding acquisition, investigation, project administration, writing–review and editing. S. Krishnamurthi: Resources, funding acquisition, investigation, project administration, writing–review and editing. R.M. Goldberg: Resources, funding acquisition, investigation, project administration, writing–review and editing. A. Venook: Resources, funding acquisition, investigation, project administration, writing–review and editing. C. Blanke: Resources, funding acquisition, investigation, project administration, writing–review and editing. A.F. Shields: Resources, funding acquisition, investigation, project administration, writing–review and editing. E.M. O'Reilly: Resources, funding acquisition, investigation, project administration, writing–review and editing. J.A. Meyerhardt: Conceptualization, resources, supervision, funding acquisition, investigation, project administration, writing–review and editing. K. Ng: Conceptualization, resources, supervision, funding acquisition, investigation, writing–review and editing.

We would like to thank the participants and staff of the clinical trials. The authors assume full responsibility for the analyses and interpretation of these data. Research reported in this publication was supported by the NCI of the NIH under award numbers U10CA180821, U10CA180882, and U24 CA196171 (to the Alliance for Clinical Trials in Oncology), UG1CA189858, UG1CA189960, UG1CA233163 (SWOG), UG1CA233180, UG1CA233290, UG1 CA233320, UG1CA233337, UG1CA233339; https://acknowledgments.alliancefound.org; U10CA180863 and CCS Grant 707213 (CCTG); UG1CA233234 and U10CA180820 (ECOG-ACRIN); U10CA180868 (NRG); U10CA180888; R01CA205406 (to K. Ng); Department of Defense (R33CA160344, to K. Ng), Project P Fund (J.A. Meyerhardt and K. Ng), Douglas Gray Woodruff Chair Fund (J.A. Meyerhardt), Anonymous Family Fund for Innovations in Colorectal Cancer (J.A. Meyerhardt), and Cancer Center Support Grant/Core Grant P30 CA008748 (E.M. O'Reilly). Also supported in part by Pfizer. The content is solely the responsibility of the authors and does not necessarily represent the official views of the NIH.

The publication costs of this article were defrayed in part by the payment of publication fees. Therefore, and solely to indicate this fact, this article is hereby marked “advertisement” in accordance with 18 USC section 1734.