Colorectal Cancer Fought by Vitamin D – 25th meta-analysis

Circulating vitamin D status and prognosis in colorectal cancer: a systematic review and meta-analysis with exploratory evidence on vitamin D receptor polymorphisms - April 2026

BMC Cancer (2026). https://doi.org/10.1186/s12885-026-16026-x

Background: Circulating 25-hydroxyvitamin D [25(OH)D] levels and genetic polymorphisms in the vitamin D receptor (VDR) have been explored as potential prognostic factors in colorectal cancer (CRC). This study aimed to investigate the association between circulating 25(OH)D levels and CRC prognostic outcomes, with a narrative evaluation of VDR polymorphisms.

Methods: We performed a systematic literature search in the Cochrane Library, PubMed, ScienceDirect, and Scopus. The primary outcomes were CRC-specific survival, overall survival (OS) and disease-free survival (DFS). Hazard ratios (HRs) with 95% confidence intervals (CIs) were pooled using the random-effects model with inverse variance weighting, comparing high versus low 25(OH)D levels. Due to heterogeneity in genetic models and limited available data, VDR polymorphisms were synthesized using a narrative approach.

Results: Of the 61 studies included in the systematic review, 35 studies were included in the meta-analysis, which showed that higher 25(OH)D levels were associated with a lower risk of mortality, including a 26% lower CRC-specific mortality (HR 0.74; 95% CI 0.69–0.80; I2 = 0.01%), a 32% lower overall mortality (HR 0.68; 95% CI 0.64–0.72; I2 = 7.6%), and improved DFS (HR 0.71; 95% CI 0.61–0.83; I2 = 36.8%). The prognostic roles of VDR polymorphisms, particularly rs7975232 (ApaI), rs1544410 (BsmI), rs2228570 / rs10735810 (FokI), and rs731236 (TaqI) with CRC outcomes, including CRC-specific survival, OS, and DFS were reported across 18 studies.

Conclusion: This study provides quantitative evidence supporting the potential prognostic relevance of circulating vitamin D levels in CRC. The role of VDR genetic polymorphisms remains inconclusive, warranting further investigations.

PDF

Vitamin D and VDR reduces the risk of getting colorectal cancer - Claude AI deep research

12 minutes, 318 sources

Bottom line: Prospective cohorts converge on ~75–100 nmol/L (30–40 ng/mL) as the serum 25(OH)D "sweet spot" for lowest CRC incidence, with ~20–30% risk reduction versus lower-sufficient levels and a plateau above ~100 nmol/L. The four major RCTs (VITAL, WHI, ViDA, D-Health) are null for CRC incidence, and Mendelian randomization corroborates that null — yet post-hoc VITAL analyses, daily-dosing meta-analyses, and mechanistic/VDR biology still support a real biological effect on cancer progression/mortality. The likeliest reconciliation: observational inverse association is partly confounded (adiposity, inflammation, activity), but RCTs were also handicapped by sub-therapeutic dosing, replete baseline populations, contamination, and follow-up shorter than CRC latency.

1. Serum 25(OH)D thresholds and dose-response

McCullough et al., JNCI 2019;111:158–169 — the definitive pooled analysis (17 cohorts, 5,706 CRC cases / 7,107 controls, assays calibrated to a single Heartland/DiaSorin LIAISON platform, season-standardized). Referent 50–<62.5 nmol/L: - <30 nmol/L (<12 ng/mL): RR 1.31 (1.05–1.62) - 75–<87.5 nmol/L: RR 0.81 (0.67–0.99) - 87.5–<100 nmol/L: RR 0.73 (0.59–0.91) - ≥100 nmol/L (≥40 ng/mL): RR 0.91 (0.67–1.24) — plateau, only 3.5% of controls - Combined 75–<100 vs 50–<62.5: overall 0.78 (0.67–0.92); women 0.67 (0.54–0.83); men 0.90 (0.68–1.18), P_het by sex = 0.008 - Per 25 nmol/L increment: women RR 0.81 (0.75–0.87); men 0.93 (0.86–1.00)

The authors explicitly argued optimal 25(OH)D for CRC prevention is 75–100 nmol/L (30–40 ng/mL) — above the IOM bone-health targets.

Garland/Gorham lineage argues higher. Gorham et al., Am J Prev Med 2007;32:210 pooled 5 serum studies: quintile ORs 1.00/0.82/0.66/0.59/0.46 (p_trend <0.0001), Q5 vs Q1 OR 0.49 (0.35–0.68); ≥33 ng/mL (82 nmol/L) vs ≤12 ng/mL ≈ 50% lower risk. Garland & Gorham, J Steroid Biochem Mol Biol 2017;168:1 (15 studies) modeled continued linear decline, proposing target ≥50 ng/mL (125 nmol/L). This extrapolation is not supported by the McCullough dataset where absolute categorization shows a plateau; most cohorts simply lack power above 100 nmol/L.

Newer data (2020–2026) largely reinforce McCullough: - UK Biobank (Sha et al., 2023; ~411,000; 12.7 y FU): deficiency <25 nmol/L vs sufficient — CRC-specific mortality HR 1.27 (1.07–1.50); insufficiency 25–<50 HR 1.14 (1.00–1.30).

  • UK Biobank incidence (He et al., Int J Cancer 2021; 360,061; 2,509 CRC): Q4 vs Q1 HR 0.87 (0.77–0.98); per 1-SD HR 0.95 (0.91–0.99).
  • Asian dose-response meta-analysis (Hong et al., BMJ Open 2020): Qhigh vs Qlow OR 0.75 (0.58–0.97).
  • Gibbs/Fedirko, JNCI Cancer Spectr 2020: effect modification by GC (DBP) rs4588 isoform — in DBP2 carriers ≥50 vs <30 nmol/L RR 0.47 (0.33–0.67); non-carriers null (P_het=0.01).
  • Survival meta-analysis (Wu et al., Biosci Rep 2020; 17,770 CRC patients): highest vs lowest 25(OH)D overall-survival HR 0.64 (0.55–0.72); CRC-specific mortality HR 0.65 (0.56–0.73).

Convergent threshold: ~75–100 nmol/L (30–40 ng/mL) for incidence; survival analyses suggest continued benefit up to similar levels. Women show roughly 2–3× stronger inverse association per unit than men.

2. RCT vs observational discrepancy

VITAL (Manson et al., NEJM 2019;380:33) — N=25,871; 2000 IU/d D3; median FU 5.3 y; baseline mean 25(OH)D 30.8 ng/mL. Primary invasive cancer HR 0.96 (0.88–1.06); CRC-specific HR 1.09 (0.73–1.62) (only ~110 cases — underpowered). Secondary/post-hoc signals:

  • Total cancer mortality excluding first 2 y: HR 0.75 (0.59–0.96)
  • BMI <25 total cancer HR 0.76 (0.63–0.90) vs BMI ≥25 HR 1.02 (p-interaction=0.002)
  • African American total cancer HR 0.77 (0.59–1.01)
  • Chandler et al., JAMA Netw Open 2020;3:e2025850 — advanced/metastatic or fatal cancer: HR 0.83 (0.69–0.99); in BMI <25 HR 0.62 (0.45–0.86), p-interaction by BMI = 0.03.

WHI CaD (Wactawski-Wende et al., NEJM 2006;354:684) — 36,282 postmenopausal women, 400 IU D3 + 1000 mg Ca; CRC HR 1.08 (0.86–1.34). Dose inadequate to meaningfully shift 25(OH)D; heavy drop-in supplementation in placebo arm.

ViDA (Scragg et al., JAMA Oncol 2018;4:e182178) — 5,110 NZ adults, 100,000 IU monthly bolus (~3,300 IU/d equivalent), FU 3.3 y. Total cancer HR 1.01 (0.81–1.25); null even in baseline-deficient stratum.

D-Health (Neale et al., Lancet Diab Endocrinol 2022;10:120) — 21,315 Australians ≥60 y, 60,000 IU monthly (~2,000 IU/d); FU 5.7 y. Cancer mortality HR 1.15 (0.96–1.39); excluding first 2 y HR 1.24 (1.01–1.54), p=0.05 — signal toward harm with monthly bolus in a replete population. Updated D-Health cancer incidence (Neale 2025) total cancer HR 1.02 (0.95–1.10); CRC null.

Meta-analyses of RCTs (Keum et al., Ann Oncol 2019;30:733).

10 RCTs (incidence): RR 0.98 (0.93–1.03).

5 RCTs (mortality, 1,591 deaths): RR 0.87 (0.79–0.96)

~13% reduction, confined to daily dosing (RR 0.87) not bolus. Keum & Giovannucci 2022 update incorporating D-Health: daily-dosing subgroup mortality RR 0.88 (0.78–0.98); CRC-specific incidence remained null.

Mendelian randomization — uniformly null for CRC incidence: - Dimitrakopoulou et al., BMJ 2017;359:j4761 (70,563 cases / 84,418 controls): CRC OR 0.92 (0.76–1.10) per 25 nmol/L.

  • Ong et al., Nat Commun 2021;12:246 (74 SNPs, 10 cancers): CRC null.
  • He et al., Int J Cancer 2022;150:303 (110 SNPs, 26,397 CRC cases): null bidirectionally.
  • Jiang et al., Sci Rep 2021: null.

MR argues either genuine null causality for lifelong circulating 25(OH)D on CRC incidence, or that current genetic instruments inadequately capture the deficient-range exposure (they explain <3% of 25(OH)D variance and are dominated by variation in the sufficient range).

Why RCT–cohort divergence? Converging critiques (Grant, Boucher, Heaney, Giovannucci):

  • Sub-therapeutic dosing — 2000 IU/d in VITAL lifted on-trial 25(OH)D only to ~42 ng/mL (delta ~12 ng/mL); WHI 400 IU/d negligible.
  • No baseline deficiency selection — only 12.7% of VITAL <20 ng/mL; D-Health/ViDA largely replete (ceiling effect).
  • Short follow-up vs CRC adenoma-to-carcinoma latency (10–20 y).
  • Drop-in contamination — VITAL allowed ≤800 IU/d personal use; WHI substantial.
  • Monthly bolus dosing (ViDA, D-Health) plausibly biologically inferior (CYP24A1 upregulation, non-physiologic PK; Hollis & Wagner, JCEM 2013).
  • Power — VITAL's ~110 CRC cases yielded a 95% CI (0.73–1.62) too wide to detect a 20–30% cohort-type effect.

3. VDR expression, polymorphisms, and mechanism

Expression gradient in colon. Normal colonic epithelium is among the highest VDR-expressing tissues, with co-expression of CYP27B1 and CYP24A1 enabling local 1,25(OH)₂D₃ generation (Tangpricha, Lancet 2001;357:1673; Bises/Cross, Carcinogenesis 2005;26:1581).

VDR is upregulated in early/well-differentiated adenomas and G1–G2 carcinomas (compensatory/differentiation response) and progressively lost in G3–G4, poorly differentiated, and metastatic CRC (Cross, Virchows Arch 2000;437:501; Matusiak, CEBP 2005;14:2370). In 658-patient metastatic CRC, stromal-fibroblast VDR expression independently predicted improved OS/DFS (Ferrer-Mayorga et al., Gut 2017;66:1449).

Mechanism of VDR silencing in tumors.

The Muñoz/Larriba group established that SNAIL1 binds three E-boxes in the VDR promoter, recruits HDAC/LSD1, and extinguishes VDR and 1,25(OH)₂D₃ responsiveness (Pálmer et al., Nat Med 2004;10:917). SNAIL1 is elevated in ~60% of CRC and anti-correlates with VDR; SNAIL2/SLUG and ZEB1 cooperate (Larriba, Carcinogenesis 2009;30:1459). Adjacent histologically "normal" mucosa in SNAIL1-high tumors already shows VDR/CDH1 loss — a field effect (Peña, Oncogene 2009;28:4375). Additional silencing mechanisms: KRAS/MAPK, TNF-α/NF-κB, promoter hypermethylation.

VDR polymorphisms and CRC risk (modest, heterogeneous):

SNP Strongest meta-analytic signal Note
BsmI (rs1544410) BB vs bb OR 0.79–0.87 (Touvier CEBP 2011 RR 0.57, 0.36–0.89; Bai WJG 2012 OR 0.87, 0.80–0.94; Pan Oncotarget 2018 OR 0.79, 0.64–0.97) Only consistently protective variant; stronger in colon than rectum
FokI (rs2228570) Null overall; borderline F vs f OR 1.03 (0.999–1.06) Modifies vitamin D intake–CRC association (Theodoratou)
TaqI (rs731236) Null in Touvier and Serrano meta-analyses
ApaI (rs7975232) Null LD with BsmI/TaqI (baT haplotype)
Cdx2, polyA Null

Effect sizes for any single SNP are |OR−1| < 0.2; clinical utility limited.

Mechanistic pathways of 1,25(OH)₂D₃/VDR in colonic epithelium:

  • Cell-cycle arrest: VDRE-driven induction of CDKN1A (p21) and post-transcriptional stabilization of CDKN1B (p27); Rb hypophosphorylation; suppression of c-MYC, cyclin D1, CDK2/4/6.
  • Apoptosis: BAX/BAK/BAD up, BCL-2 and survivin down; miR-22 induction (Álvarez-Díaz, Hum Mol Genet 2012).
  • Differentiation/MET: Landmark Pálmer et al., J Cell Biol 2001;154:369 — 1,25(OH)₂D₃ induces E-cadherin, occludin, ZO-1/2, relocates β-catenin from nucleus to adherens junctions.
  • Wnt/β-catenin antagonism — multilevel: (i) ligand-bound VDR physically binds β-catenin via AF-2, competing with TCF-4 and repressing MYC, CCND1, LEF1, AXIN2, CD44 (Pálmer 2001; Shah et al., Mol Cell 2006;21:799); (ii) induction of DKK-1 (Aguilera, Carcinogenesis 2007;28:1877); (iii) repression of pro-invasive DKK-4 (Pendás-Franco, Oncogene 2008;27:4467); (iv) in vivo Apc^min/+^;Vdr^-/-^ mice show more ACF, higher nuclear β-catenin, greater tumor burden (Larriba, PLoS ONE 2011;6:e23524).
  • Bile-acid detoxification: Makishima et al., Science 2002;296:1313 — VDR is a lithocholic-acid sensor (~10× more sensitive than PXR/FXR); LCA-VDR induces intestinal/colonic CYP3A4, SULT2A1, MRP3, detoxifying this secondary-bile-acid enteric carcinogen; intestine-specific Vdr deletion exacerbates LCA toxicity.
  • Anti-inflammatory: Direct VDR–p65/RelA interaction inhibits NF-κB; suppression of COX-2, IL-6, TNF-α; inhibition of STAT1 in tumor-associated macrophages (Ferrer-Mayorga, Gut 2017).
  • Anti-angiogenic: HIF-1α, VEGF, IL-8 repression.

Microbiome–VDR axis (Jun Sun lab). VDR transactivates DEFB4, CAMP (cathelicidin/LL-37), ATG16L1, NOD2, claudins, AXIN1. Vdr^-/-^ and intestinal-specific Vdr^ΔIEC^ mice exhibit dysbiosis — depleted Lactobacillus and Akkermansia muciniphila, enriched Bacteroides/Proteobacteria/H. hepaticus; reduced Paneth-cell defensins; thinned mucus; bacterial translocation; accelerated AOM/DSS tumorigenesis via JAK2/STAT3. Human VDR genotype is a top host determinant of gut microbiome composition. Microbial secondary bile acids (LCA, 3-oxo-LCA, iso-LCA derivatives) and LCA-acetate (EC50 ≈ 0.4 µM) are endogenous VDR ligands — a reciprocal microbiome↔VDR↔bile-acid loop.

Therapeutic implications of VDR silencing. SNAIL1-high/VDR-low tumors are refractory to calcitriol/EB1089/paricalcitol/inecalcitol — likely partially explaining negative vitamin-D-analog trials (e.g., AMATERASU in digestive cancers). Rationales now being pursued: combination with HDAC inhibitors / demethylating agents to reactivate VDR; targeting stromal fibroblast VDR (independent prognostic axis per Ferrer-Mayorga Gut 2017); patient stratification by VDR/SNAIL1 IHC prior to vitamin D trials; probiotic (L. plantarum) VDR upregulation.

Synthesis and caveats

The evidence matrix is now strikingly consistent across methodologies when read carefully:

observational cohorts with calibrated assays place the incidence inflection at ~75–100 nmol/L with a plateau beyond, the daily-dosing RCT meta-analysis supports a ~13% cancer-mortality reduction (Keum 2019), VITAL post-hoc BMI-stratified and fatal/metastatic endpoints are positive (Chandler 2020), and mechanistic biology is robust (VDR as an LCA detoxifier, Wnt antagonist, differentiation inducer; SNAIL1-mediated silencing in progression). The ostensible RCT "failure" for CRC incidence reflects a combination of (a) genuine confounding in observational data (MR is null), (b) RCTs designed around bone-centric hypotheses with doses, baselines, and follow-up durations inappropriate for cancer latency, and (c) the probable truth that vitamin D acts more strongly on progression/fatal outcomes and in lean individuals than on initial adenoma formation in already-replete populations.

Key caveats for an expert reader:

  • (1) the Garland ≥50 ng/mL target is an extrapolation the McCullough absolute-cutpoint data do not support; (
  • (2) sex heterogeneity is real and under-appreciated (women benefit roughly 2–3× more per unit 25(OH)D);
  • (3) BMI and GC isoform are probable effect modifiers that trialists should stratify on;
  • (4) monthly bolus dosing should be abandoned for future cancer trials — daily dosing in baseline-deficient subjects with 10+ year follow-up is the design most likely to yield a definitive answer;
  • (5) the VDR silencing/SNAIL1 axis implies that once CRC is established, systemic 25(OH)D sufficiency may be biologically bypassed in the tumor itself — arguing that vitamin D's main value is in primary and early prevention and in modulating the stromal/microbiome/bile-acid milieu rather than in late-stage treatment.

Related in VitaminDWiki