Figure 2 shows a visual overview of the results, with details in Table 1 and Table 2. Figure 3, 4, and 5 show forest plots for a random effects meta-analysis of all studies with pooled effects, mortality results, and cases.

To avoid bias in the selection of studies, we analyze all non-retracted studies. Here we show the results after excluding studies with major issues likely to alter results, non-standard studies, and studies where very minimal detail is currently available. Our bias evaluation is based on analysis of each study and identifying when there is a significant chance that limitations will substantially change the outcome of the study. We believe this can be more valuable than checklist-based approaches such as Cochrane GRADE, which may underemphasize serious issues not captured in the checklists, overemphasize issues unlikely to alter outcomes in specific cases (for example, lack of blinding for an objective mortality outcome, or certain specifics of randomization with a very large effect size), or be easily influenced by potential bias. However, they can also be very high quality.

The studies excluded are as below. Figure 6 shows a forest plot for random effects meta-analysis of all studies after exclusions.

[Magaña], unadjusted results with no group details.

People with healthier diets have reduced risk for COVID-19. Statistically significant improvement is seen for cases. 5 studies from 5 independent teams in 3 different countries show statistically significant improvements in isolation. Meta analysis using the most serious outcome reported shows 54% [27‑71%] improvement. Results are similar after exclusions.

Studies analyze diet quality before infection, and use different definitions of diet quality.

Ahmadi

[Ahmadi] Retrospective 468,569 adults in the UK, showing significantly lower COVID-19 mortality with physical activity.

Submit Corrections or Updates.

Kim

[Kim] Retrospective healthcare workers in six countries with exposure to COVID-19 patients, showing lower risk of moderate/severe COVID-19 with plant-based diets.

Submit Corrections or Updates.

Magaña

[Magaña] Retrospective 89 COVID-19 patients in Spain, showing lower mortality with adherence to the Mediterranean diet.

Submit Corrections or Updates.

Merino

[Merino] Retrospective 592,571 participants in the UK and USA with 31,815 COVID-19 cases, showing lower risk or COVID-19 cases and severity for higher healthful plant-based diet scores. Notably, the assocation was less evident with higher levels of physical activity.

Submit Corrections or Updates.

Moludi

[Moludi] Retrospective 60 COVID-19 hospitalized patients and 60 controls in Iran, showing pro-inflammatory diets associated with COVID-19 cases and severity. IR.KUMS.REC.1399·444, IR.TBZMED.REC.1399·225.

Submit Corrections or Updates.

Nguyen

[Nguyen] Analysis of 3,947 participants in Vietnam, showing significantly lower risk of COVID-19-like symptoms with physical activity and with a healthy diet. The combination of being physically active and eating healthy reduced risk further compared to either alone. The analyzed period was Feb 14 to Mar 2, 2020, which may have been before testing was widely available.

Submit Corrections or Updates.

Perez-Araluce

[Perez-Araluce] Retrospective 5,194 participants in Spain with 382 COVID-19 cases, showing lower risk of COVID-19 with high adherence to a Mediterranean diet, with statistical significance only when excluding healthcare professionals.

Submit Corrections or Updates.

We performed ongoing searches of PubMed, medRxiv, ClinicalTrials.gov, The Cochrane Library, Google Scholar, Collabovid, Research Square, ScienceDirect, Oxford University Press, the reference lists of other studies and meta-analyses, and submissions to the site c19early.com. Search terms were diet AND COVID-19. Automated searches are performed every few hours with notification of new matches. All studies regarding the use of diet for COVID-19 that report a comparison with a control group are included in the main analysis. Sensitivity analysis is performed, excluding studies with major issues, epidemiological studies, and studies with minimal available information. This is a living analysis and is updated regularly.

We extracted effect sizes and associated data from all studies. If studies report multiple kinds of effects then the most serious outcome is used in pooled analysis, while other outcomes are included in the outcome specific analyses. For example, if effects for mortality and cases are both reported, the effect for mortality is used, this may be different to the effect that a study focused on. If symptomatic results are reported at multiple times, we used the latest time, for example if mortality results are provided at 14 days and 28 days, the results at 28 days are used. Mortality alone is preferred over combined outcomes. Outcomes with zero events in both arms were not used (the next most serious outcome is used — no studies were excluded). For example, in low-risk populations with no mortality, a reduction in mortality with treatment is not possible, however a reduction in hospitalization, for example, is still valuable. Clinical outcome is considered more important than PCR testing status. When basically all patients recover in both treatment and control groups, preference for viral clearance and recovery is given to results mid-recovery where available (after most or all patients have recovered there is no room for an effective treatment to do better). If only individual symptom data is available, the most serious symptom has priority, for example difficulty breathing or low SpO₂ is more important than cough. When results provide an odds ratio, we computed the relative risk when possible, or converted to a relative risk according to [Zhang]. Reported confidence intervals and p-values were used when available, using adjusted values when provided. If multiple types of adjustments are reported including propensity score matching (PSM), the PSM results are used. Adjusted primary outcome results have preference over unadjusted results for a more serious outcome when the adjustments significantly alter results. When needed, conversion between reported p-values and confidence intervals followed [Altman, Altman (B)], and Fisher's exact test was used to calculate p-values for event data. If continuity correction for zero values is required, we use the reciprocal of the opposite arm with the sum of the correction factors equal to 1 [Sweeting]. Results are expressed with RR < 1.0 favoring treatment, and using the risk of a negative outcome when applicable (for example, the risk of death rather than the risk of survival). If studies only report relative continuous values such as relative times, the ratio of the time for the treatment group versus the time for the control group is used. Calculations are done in Python (3.9.12) with scipy (1.8.0), pythonmeta (1.26), numpy (1.22.2), statsmodels (0.14.0), and plotly (5.6.0).

Forest plots are computed using PythonMeta [Deng] with the DerSimonian and Laird random effects model (the fixed effect assumption is not plausible in this case) and inverse variance weighting. Mixed-effects meta-regression results are computed with R (4.1.2) using the metafor (3.0-2) and rms (6.2-0) packages, and using the most serious sufficiently powered outcome.

We received no funding, this research is done in our spare time. We have no affiliations with any pharmaceutical companies or political parties.

We have classified studies as early treatment if most patients are not already at a severe stage at the time of treatment (for example based on oxygen status or lung involvement), and treatment started within 5 days of the onset of symptoms. If studies contain a mix of early treatment and late treatment patients, we consider the treatment time of patients contributing most to the events (for example, consider a study where most patients are treated early but late treatment patients are included, and all mortality events were observed with late treatment patients). We note that a shorter time may be preferable. Antivirals are typically only considered effective when used within a shorter timeframe, for example 0-36 or 0-48 hours for oseltamivir, with longer delays not being effective [McLean, Treanor].

Prophylaxis

Effect extraction follows pre-specified rules as detailed above and gives priority to more serious outcomes. For pooled analyses, the first (most serious) outcome is used, which may differ from the effect a paper focuses on. Other outcomes are used in outcome specific analyses.

Supplementary Data