PLOS One has retracted a widely cited 2022 study linking pre-infection Vitamin D (25-hydroxyvitamin D3) levels to COVID-19 severity, citing inadequate analyses and reliance on 25(OH)D results taken 14–730 days before infection. Announced on September 8, 2025, the retraction emphasized that concerns centered on the study’s ability to test its core association, and editors did not allege misconduct by the authors [1].

The PubMed record now marks the 2022 article as retracted (original PMID 35113901) and hosts the retraction notice under PMID 40920624. The index entry recaps methodological objections, especially the temporal mismatch between pre-infection Vitamin D measurements and the clinical course of COVID-19, and links out to the full text of the retraction notice [2].

Key Takeaways

– shows PLOS One retracted the 2022 Vitamin D–COVID study on Sept 8, 2025, citing inadequate analyses and 14–730 day pre-infection measurement windows. – reveals the original cohort had 1,176 hospital admissions, but only 253 patients had pre-infection 25(OH)D tests recorded 14–730 days earlier. – demonstrates authors reported vitamin D deficiency (<20 ng/mL) increased severe/critical COVID-19 odds 14-fold versus ≥40 ng/mL, with 95% CI ranging 4–51. - indicates deficiency prevalence was 87.4% among severe/critical cases compared with 34.3% in mild/moderate cases, a highly significant difference reported with p<0.001. - suggests controversy persists: authors disputed the retraction and critics called their defense “deeply bizarre,” as debate over Vitamin D evidence quality intensifies.

Why PLOS One Retracted the Vitamin D COVID-19 Study

Retractions are blunt instruments, so editors typically reserve them for cases where fundamental validity is in doubt. Here, editors concluded that the analytical approach was not adequate to test whether pre-infection Vitamin D status predicted COVID-19 severity. A central concern was the timing: many 25(OH)D results were drawn months—and in some cases years—before SARS-CoV-2 infection, spanning 14 to 730 days. That gap can disconnect the measured exposure from the relevant disease window and may confound severity comparisons if levels changed between testing and infection [1].

The retraction notice also clarified that this was not an authorship or misconduct finding. Rather, it was a methodological determination about whether the data and analyses could support such a strong clinical inference. In observational research, timing and confounding are often decisive. In this case, both the long pre-infection window and questions about model specification undermined the editors’ confidence in the original quantitative claims about risk gradients tied to Vitamin D [1].

What the Original Vitamin D Findings Claimed

The paper—first circulated as a medRxiv preprint—assembled a retrospective cohort of 1,176 hospitalized COVID-19 patients between April 7, 2020 and February 4, 2021. Of those, 253 had pre-infection 25(OH)D results on file from 14–730 days prior. The authors reported that Vitamin D deficiency, defined as less than 20 ng/mL, was associated with markedly higher illness severity. Compared with levels at or above 40 ng/mL, deficiency was linked to a 14-fold higher odds of severe/critical disease (odds ratio 14; 95% CI 4–51), an estimate that drew widespread attention [4].

A post-publication summary of the peer-reviewed version highlighted stark group differences: deficiency was present in 87.4% of severe/critical cases versus 34.3% of mild/moderate cases (p<0.001). It also noted that the study used a cosinor model to adjust for seasonal variation in 25(OH)D and acknowledged limitations typical of retrospective analyses, including potential confounding and selection bias in who had prior Vitamin D tests [5].

Methodological Fault Lines: Timing, Models, and Bias

The 14–730 day gap between Vitamin D testing and infection is not just a technical footnote; it is a key source of potential exposure misclassification. Vitamin D levels fluctuate with season, sun exposure, diet, supplementation, and changes in health status. A measurement taken a full year before infection may not reflect a patient’s actual pre-infection status. Misclassification of exposure usually biases estimates, and depending on its structure, can either attenuate or exaggerate observed associations.

Seasonality adjustment with a cosinor model is a reasonable tool, but such models rest on assumptions about periodicity and amplitude that must fit the local data. If patients tested at different times of year also differ systematically in comorbidities, care-seeking, or other risk factors, seasonal adjustment alone cannot eliminate those biases. The study’s reliance on patients who happened to have prior 25(OH)D tests also introduces selection concerns, since those tested may differ from those not tested in clinically relevant ways.

In retrospective cohorts, strong effect sizes can flag real signals—or residual confounding. The reported 14-fold odds ratio contrasts the tails of the distribution (deficiency <20 ng/mL versus ≥40 ng/mL). If unmeasured or inadequately modeled covariates correlate with both Vitamin D status and adverse outcomes (for example, frailty or socioeconomic factors), the estimate can be inflated. Even sophisticated multivariable adjustments cannot correct for missing or poorly measured confounders.

The editors’ conclusion that the analyses were “inadequate to test the association” is a high bar: it suggests that the data structure, variable timing, and model choices, taken together, do not justify causal or even robust prognostic claims at the reported magnitude. That determination forces a reassessment of confidence in the effect size and the study’s clinical implications for risk stratification by Vitamin D status [1].

Authors Push Back and Community Reaction

The authors have publicly disputed the retraction, arguing that they explicitly examined pre-infection—not at-infection—Vitamin D status and offering additional analyses in their defense. They criticized aspects of the retraction process and contended that some editorial criticisms were erroneous. A prominent critic, however, labeled the authors’ defense “deeply bizarre,” underscoring a sharp divide between the research team and external commentators over what the data can reliably prove. The episode highlights opaque boundaries between methodological debate and editorial action in fast-moving pandemic science [3].

Interpreting Vitamin D Evidence after the Retraction

This retraction does not resolve whether Vitamin D plays a biological role in COVID-19 outcomes; it specifically undercuts confidence in one study’s methods and conclusions. The strongest evidence for causal effects typically comes from randomized controlled trials and large, prospective cohorts with precise exposure timing. Retrospective, convenience-sampled data—especially with long lags between exposure measurement and outcome—are inherently limited for estimating prognostic or causal effects.

For clinicians and public health readers, the practical implication is caution: do not overinterpret large observational effect sizes when exposure timing is misaligned. A 14-fold odds ratio is eye-catching, but if derived from a subset with older lab tests, the estimate may reflect selection and measurement biases. Severity differences, such as the 87.4% versus 34.3% deficiency split across clinical strata, warrant hypothesis generation, not clinical guidance, until replicated with tighter designs and prespecified analytic plans.

Policy discourse around Vitamin D often mixes well-supported facts—such as widespread deficiency in some populations and known roles in bone health—with contested claims about infectious disease outcomes. After this retraction, evidence syntheses should down-weight the retracted findings and prioritize trials and prospective cohorts that measure 25(OH)D close to infection, track supplementation, and systematically adjust for confounders like age, comorbidities, obesity, and deprivation.

What This Retraction Means for Vitamin D Research

For researchers, the message is methodological: exposure timing matters as much as sample size. Re-analyses should test the sensitivity of results to shorter measurement windows (for example, ≤90 days pre-infection), employ rigorous missing-data strategies, and pre-register modeling choices. Where feasible, measuring 25(OH)D at or near diagnosis can clarify prognostic value independent of disease-driven changes in nutrient status.

For readers and journalists, the lesson is interpretive discipline. Extraordinary effect sizes in observational studies deserve extra skepticism, especially when derived from subgroups with idiosyncratic data availability. Future Vitamin D studies should report the distribution of measurement lags, stratified estimates by time windows, and robustness checks that do not rely on a single seasonal model or dichotomization threshold.

For the broader public, today’s bottom line is unchanged: maintain evidence-based Vitamin D intake for established indications, and look to diversified, higher-quality evidence when assessing claims about respiratory infections. This retraction narrows the evidentiary base for a strong Vitamin D–COVID-19 severity link, and it should prompt better-designed studies rather than overgeneralization from convenience data.

The Paper’s Long Tail: Citations, Preprints, and Post-Publication Review

The study’s trajectory—from a 2021 preprint to a 2022 peer-reviewed publication and a 2025 retraction—illustrates pandemic-era dynamics. Preprints accelerated access to findings, but also allowed headline numbers to proliferate before exhaustive vetting. Post-publication platforms aggregated critiques and context, while journals navigated between author defenses and external methodological concerns.

As with other COVID-19 controversies, public attention focused on a single statistic—the 14-fold odds—at the expense of caveats about timing and selection. The ensuing retraction re-centers those caveats. Going forward, editors and reviewers may emphasize mandatory sensitivity analyses around exposure timing, particularly for biomarkers like 25(OH)D that vary with season and behavior. That shift could improve the reliability and reproducibility of Vitamin D epidemiology.

Data Transparency and Next Steps for Stakeholders

Transparency will be critical to rebuild confidence. Sharing de-identified data and code, clarifying assay methods, and documenting measurement intervals can allow independent teams to stress-test findings. Journals can publish structured post-publication exchanges that map critiques to specific analyses, making the decision to correct, update, or retract more legible.

Funding agencies and guideline bodies should incentivize trials that randomize supplementation with adequate dosing, monitor achieved 25(OH)D levels, and predefine clinical endpoints and timing. Such designs can quantify effect sizes (if any) without the timing ambiguities that proved pivotal here. For the Vitamin D community, precision on “when” the exposure is measured may be as decisive as “how much” and “in whom.”

Sourcing the Record

The factual record is anchored by the journal’s retraction notice, indexing entries, preprint data, post-publication summaries, and independent reporting. Together, these materials document the key dates, sample sizes, thresholds, and effect estimates—and the reasons editors judged the analyses insufficient to support the headline conclusion that pre-infection Vitamin D strongly predicts COVID-19 severity. By engaging these primary sources directly, readers can trace how statistical choices and timing windows can govern the strength of scientific claims [1].

Sources:

[1] PLOS One (Retraction Notice) – Retraction: Pre-infection 25-hydroxyvitamin D3 levels and association with severity of COVID-19 illness: https://journals.plos.org/plosone/article?id=10.1371/journal.pone.0331693

[2] PubMed (Retraction entry) – Retraction: Pre-infection 25-hydroxyvitamin D3 levels and association with severity of COVID-19 illness: https://pubmed.ncbi.nlm.nih.gov/40920624/ [3] Retraction Watch – Authors defend retracted paper on vitamin D and COVID-19 called ‘deeply bizarre’ by critic: https://retractionwatch.com/2025/09/25/authors-defend-retracted-paper-on-vitamin-d-and-covid-19-critic-called-deeply-bizarre/

[4] medRxiv (preprint) – Pre-infection 25-hydroxyvitamin D3 levels and association with severity of COVID-19 illness (preprint): www.medrxiv.org/content/10.1101/2021.06.04.21258358v1″ target=”_blank” rel=”nofollow noopener noreferrer”>https://www.medrxiv.org/content/10.1101/2021.06.04.21258358v1 [5] Sciety – Pre-infection 25-hydroxyvitamin D3 levels and association with severity of COVID-19 illness: https://sciety.org/articles/activity/10.1371/journal.pone.0263069