Have COSMOS investigators published subgroup analyses by baseline nutritional biomarkers (e.g., B12, folate, vitamin D) and their interaction with multivitamin effects?
This fact-check may be outdated. Consider refreshing it to get the most current information.
Executive summary
COSMOS investigators collected baseline and follow-up blood specimens (including vitamin D, B12, and folate in a biomarker subsample) and have signaled intent to examine whether baseline nutritional status modifies treatment effects, but published COSMOS papers to date do not report formal subgroup analyses stratified by baseline B12, folate, or vitamin D interacting with the multivitamin intervention (MVM) [1] [2] [3]. Instead, published subgroup work has focused on clinical or behavioral strata (for example, diet quality, cardiovascular disease history, smoking) and exploratory biomarker measurement papers or ancillary studies remain forthcoming [4] [5] [6].
1. What COSMOS actually measured — biomarker collection but limited published biomarker interaction analyses
The COSMOS protocol and ancillary descriptions make clear that a large biospecimen subcohort was assembled specifically so investigators could assess baseline nutritional biomarkers and how those levels change with intervention; the trial prospectively collected baseline blood and spot urine specimens in thousands of participants and planned follow-up biomarker measures in subsets [7] [8] [2]. COSMOS investigators explicitly measured vitamin D, vitamin B12, and folate in a biomarker sample of 399 participants with baseline and at least one follow-up blood sample [1], but the presence of these measurements in a subset does not by itself mean interaction/subgroup analyses by those biomarkers have been published.
2. What has been published — subgroup analyses that are clinical/behavioral, not biomarker-driven
The major COSMOS outcome publications report secondary and subgroup analyses, but the reported interactions center on demographic, clinical, and behavioral variables: for example, subgroup signals for cocoa effects by smoking history (total CVD subgroup analyses) and cognitive subgroup signals by baseline diet quality or by history of cardiovascular disease for MVM effects [6] [4] [5]. The high-profile COSMOS cognitive MVM paper and meta-analysis reported greater cognitive benefit in some preplanned subgroups (such as those with prior CVD) and noted subgroup findings were hypothesis-generating; these papers also emphasize that biomarker- or -omics-based effect modification was not included and remains a research priority [9] [5].
3. Why no published biomarker-by-treatment interaction yet — sample size, planned timing, and conservative interpretation
COSMOS materials and Q&A state that blood samples are held and often analyzed at study end or for specific ancillary projects and that some lab assays are done only on samples from a subset of participants (limiting power for interaction tests) [2] [7]. The AJCN and related COSMOS reports likewise flag that biomarker and -omics analyses are planned as future work to probe mechanisms and effect heterogeneity, and that secondary/subgroup analyses already published were exploratory and not adjusted for multiple comparisons [10] [3]. The explicit reporting that vitamin D, B12, and folate were measured in 399 participants implies a modest biomarker sample for interaction testing compared with the full trial, which the investigators have acknowledged in describing ancillary study design and limitations [1] [8].
4. Bottom line and where to watch next
Based on currently available COSMOS publications and the trial website, investigators have not yet published formal subgroup analyses that test interaction between baseline nutritional biomarkers (vitamin B12, folate, vitamin D) and the multivitamin’s effects on clinical endpoints; instead, biomarker measurement and ancillary-analysis capacity exist and the team has signaled that biomarker- and -omics-based modifier analyses are planned or underway [1] [2] [3]. Readers should expect future COSMOS ancillary papers to address these questions directly, but until those specific interaction analyses appear in peer‑reviewed reports, claims that COSMOS has demonstrated biomarker-dependent MVM effects are premature [10] [9].