Multivitamin-mineral supplements are commonly used in the United States. National surveys indicate that over half of adults report using at least one dietary supplement during the preceding 30 days, whereas about a third report using a multivitamin-mineral product [1-

4]. Evidence regarding the health effects of multivitamin-mineral supplementation is mixed and outcome-specific. Systematic reviews and large randomized trials have generally found little or no benefit for cardiovascular events or all-cause mortality [5-6]. A small reduction in cancer incidence has been reported [7-8]. Evidence for cognition in older adults is present in some studies

Received: Jun. 30, 2026; Revised: Aug. 25, 2026; Accepted: Sep. 3, 2026; Published: Sep. 16, 2026 Copyright ©2026 Thomas F. Heston.

DOI:

This is an open-access article distributed under a CC BY license (Creative Commons Attribution 4.0 International License)

A previous version of this manuscript was deposited as a preprint on Research Square, DOI: .

but remains dependent on the population and outcome examined [9-11].

The prevailing evidence indicates that routine MVI supplementation provides little or no clinically important benefit for preventing cardiovascular disease, most cancer outcomes, or all-cause mortality in generally healthy adults without known nutritional deficiencies, although small benefits have been reported for overall cancer incidence and cognitive function in older adults [8, 11]. An alternative possibility also merits consideration: t reating “ multivitamin use” as a homogeneous intervention may obscure formulation-specific effects by pooling heterogeneous products [12]. These explanations are not mutually exclusive. Distinguishing them requires first establishing how much MVI formulations actually differ.

In epidemiological studies, self-reported dietary-supplement use may be assessed using structured questionnaires or 24-hour dietary recalls. Structured questionnaires commonly use predefined categories of supplement type and frequency, whereas 24-hour recalls are open-ended and can collect more detailed information when required by the study protocol [13-14]. Some studies nevertheless classify participants using a general multivitamin category or assign a default nutrient composition rather than incorporating the exact product formulation. Because multivitamin products vary substantially in nutrient content, this practice can misclassify formulation-specific exposure and produce inaccurate nutrient-intake estimates [15]. Such measurement error may attenuate formulation-specific associations under some conditions. While non-differential misclassification is expected to bias estimates toward the null, that expectation does not hold reliably when exposure has multiple categories or the underlying error structure is complex [16].

Prior work has established that multivitamin composition is not uniform. Methodologic analyses showed that allowing for variation in multivitamin composition improves nutrient intake estimates [15], that the method of supplement exposure assessment materially changes estimated nutrient intakes [17], and that the category itself lacks a consistent scientific or regulatory definition [12, 18]. Chemical analyses documented substantial between-product variability in measured content, including a wide range of vitamin D relative to the label [19-20]. What remains missing is a direct, current, cross-product quantification of how formulations differ in the nutrients they include and at what labeled

%DV, expressed in the same terms in which exposure enters epidemiologic categorization.

The present study addresses this gap using a current (2026) sample of multivitamins ranked by sales volume in the dominant U.S. online marketplace, rather than a catalog or convenience sample. It quantifies compositional heterogeneity across 25 micronutrients along three explicit axes and assesses whether products sold and used

as “multivitamins” are similar enough in composition to justify a single exposure category, or whether the category is intrinsically heterogeneous.

The 64 multivitamin formulations varied along three distinct dimensions: the number of nutrients each product contained, the labeled amount of each nutrient, and the frequency of megadoses. Each dimension is reported in turn.

Best-selling multivitamins varied markedly along all three axes examined: breadth, ranging from 7 to 24 of 25 nutrients; per-nutrient dose, with vitamin B12 alone spanning 100% to 33,333% of the Daily Value; and megadosing, present in nearly half of products. “Multivitamin use,” as recorded in most dietary research, therefore does not correspond to a single, well-defined exposure.

Prior work has established that multivitamin composition varies, with chemical analyses documenting substantial between-product variability [19-20], and methodological work showing that composition-aware modeling improves intake estimates [12, 15, 18]. The present study extends this literature in three specific ways. It draws on a current sample defined by online sales rank rather than by convenience or catalog listing, targeting formulations people actually buy, albeit within a single retail channel; it quantifies heterogeneity across 25 nutrients simultaneously using label %DV, the exact form in which exposure enters most epidemiologic categorization; and it introduces two simple, reportable indices, composition breadth and megadose prevalence, that make the heterogeneity legible to researchers designing exposure definitions.

Because a single “multivitamin user” category can contain a product supplying 7 nutrients and one supplying 24, or 100% DV of vitamin B12 alongside 33,333% DV, pooling such products under one exposure label plausibly introduces substantial measurement error. The direction of the resulting bias cannot be assumed. Non-differential misclassification is often expected to attenuate

associations toward the null, but that expectation does not hold reliably when exposure has multiple categories or the error structure is complex [16], and correcting such error is a recognized challenge in nutritional epidemiology [25-26]. Whether this mechanism materially contributes to the persistently null findings of major multivitamin trials and cohorts [3, 5-6, 8] cannot be determined from compositional data alone. The present findings support this as a testable hypothesis rather than an established explanation. Distinguishing this from the competing possibility that micronutrient supplementation has little effect on hard endpoints across the entire compositional range will require studies that link specific formulations or composition strata to outcomes. Because the direction of bias is unknown, null findings from studies using a pooled multivitamin category cannot be read as conservative estimates of a true effect.

The megadoses observed were concentrated in water-soluble B-complex vitamins and biotin. Several of these nutrients, including thiamin, riboflavin, vitamin B12, biotin, and pantothenic acid, have no established Tolerable Upper Intake Level because of low toxicity [27]. For these nutrients, extreme label %DV values exaggerate differences in biologically delivered dose, because absorption of large oral boluses is limited; the heterogeneity is real at the level of labeled exposure but attenuated at the level of physiologic exposure. The exceptions are informative. Vitamin E, megadosed in two products at up to 1,787% DV, is fat-soluble and has a defined upper intake level, and chromium is a mineral with accumulation potential; for such nutrients, label heterogeneity tracks biologically meaningful differences more closely. Even among the water-soluble vitamins, safety is not unlimited. For example, regulatory agencies have revisited the upper intake level for vitamin B6 [28], with reports of neuropathy associated with routine vitamin B6 supplement use [29-30]. These reports underscore that megadosing is not uniformly benign, at least for the water-soluble vitamin B6.

Labeled %DV also does not capture nutrient form, which alters bioavailability. Folate, for example, appears as folic acid in some products and as 5-methyltetrahydrofolate in others, and these forms differ in absorption and metabolism; two products listing an identical folate %DV may deliver different physiologic exposures. A study that pools such products mixes not only doses but chemical forms, compounding exposure misclassification beyond what the %DV heterogeneity alone implies.

Exposure misclassification is a recognized source of bias in pharmacoepidemiology, and the magnitude of the resulting distortion depends on which aspect of exposure is misclassified [31]. The same reasoning applies to nutrient exposure: treating compositionally distinct formulations as equivalent obscures dose-response relationships and formulation-specific effects. Accordingly, nutritional research examining multivitamin

supplementation should record specific products and doses, or stratify by composition, rather than rely on a single categorical exposure.

This study has several limitations. First, the analysis relied on labeled %DV rather than laboratory assay; however, labeled composition is precisely the information available to consumers and to the questionnaire-based instruments the study critiques, so it is the operative exposure in epidemiologic research. Independent laboratory analyses have shown that labeled and measured supplement content can diverge substantially [32-35], reinforcing that label-based composition should be interpreted as declared, rather than verified content. Second, the sample was drawn from a single online retailer; however, Amazon’s dominance of U.S. online supplement sales [36] and its aggregation of warehouse-club and direct-sales brands make it a broad and reproducible frame. That frame is nonetheless partial, since e-commerce accounted for an estimated 26.9% of

U.S. supplement sales in 2025 [21]. Best-seller ranking therefore weights the sample toward the products most purchased online rather than across all retail channels, and the findings describe the online segment of the category rather than the whole. Third, the data capture a single time point; however, the date-stamped, sales-ranked frame makes the snapshot exactly reproducible and anchors it to the current market rather than to historical formulations. Fourth, 64 products is a modest number, and the sample is not a census of adult multivitamin formulations. It is a complete enumeration of the qualifying products within the top 100 listings on a single best-seller list on a single date; formulations ranked below that cutoff and those sold only through other channels were not assessed. The direction of this limitation is nonetheless favorable to the study’s aim, because a truncated sample can only understate compositional heterogeneity: adding formulations can widen the observed range of breadth and dose but cannot narrow it. Documenting that the category is heterogeneous, and by how much at minimum, therefore does not require exhaustive coverage. These features make the sample well suited to the study’s aim of characterizing, rather than explaining, compositional heterogeneity.

Best-selling multivitamins vary markedly in how many nutrients they contain, from 7 to 24 of 25; in the labeled amount of each nutrient; and in whether they megadose, with 47% containing at least one nutrient at 1,000% of the Daily Value or more. “Multivitamin use”

is therefore a compositionally heterogeneous exposure rather than a single intervention. Studies of the health effects of multivitamin supplementation should specify exact formulations, record brand and dosage information, or stratify analyses by nutrient composition. Whether this heterogeneity contributes to the inconsistent findings of the existing literature is a hypothesis these data motivate but do not test; examining it will require linking specific formulations to outcomes. For clinicians, broad statements that multivitamins are effective or ineffective are difficult to interpret without specifying which formulations and doses were studied.

Supplementary materials associated with this article are available at the following link: cn/FNDS-604%20Supplementary%20Table%20S1.xlsx.

This study received no external funding.

The full product list, ASINs, dosage forms, screening dispositions, and extracted %DV data are provided as Supplementary Table S1 and are permanently archived in the Zenodo Repository (DOI: 10.5281/zenodo.18050899). No other data were generated or analyzed.

The author declares no conflicts of interest.

Large language models were used for language editing, tabulation, and formatting assistance; the author reviewed, verified, and is fully responsible for all content.

2347. doi:10.1001/jama.2021.15650.

analysis of 3 cognitive studies within COSMOS. The American Journal of Clinical Nutrition. 2024; 119(3): 692-701. doi:10.1016/j.ajcnut.2023.12.011.

680-687. doi:10.1093/ije/dyi060.

jamainternmed.2013.8759.

become top sales channel for dietary supplements. New Hope Network. 2026. Available from: https:// e-commerce-poised-to-become-top-sales-channel-for-dietary-supplements [Accessed 23rd July 2026].

I. Quantity of melatonin and CBD in melatonin gummies sold in the US. JAMA. 2023; 329(16): 1401. doi:10.1001/jama.2023.2296.

doi:10.1371/journal.pone.0343600.