Male sex identified by global COVID-19 meta-analysis as a risk factor for death and ITU admission

Reports throughout the COVID-19 pandemic indicated that men may experience more severe disease than women, but the extent and universality of this bias and whether it reflects higher infection rates or more severe outcomes remained unclear. The study aims to determine, using a large-scale global meta-analysis, whether male sex is associated with differential risk of SARS-CoV-2 infection, intensive treatment unit admission, and mortality, and to assess whether any observed sex bias is a global phenomenon. Establishing sex as an independent risk factor has important implications for public health, clinical triage, and mitigation strategies.

Previous outbreaks of coronaviruses have shown similar sex differences in outcomes. During the 2003 SARS-CoV-1 epidemic in Hong Kong, the age-adjusted relative mortality risk for males was 1.62, and in Singapore male sex was associated with increased odds of ITU admission and death. For MERS (2013–2014), case fatality rates were substantially higher in men (52%) compared to women (23%). Beyond coronaviruses, extensive literature documents sex differences in immune responses: females typically have higher CD4+ T cell counts, greater B-cell-driven antibody production, and stronger humoral responses to vaccination, often achieving protective titers at lower doses. Females also produce more type I interferon in response to viral RNA via TLR7, influenced by sex hormones and X-linked immune genes with potential biallelic expression in female immune cells. Estrogens can augment T-cell responses, antibody maturation, and innate cytokine responses, whereas testosterone tends to be immunosuppressive. These biological differences may underlie observed male susceptibility to severe COVID-19.

Search and inclusion: Reports from January 1, 2020 to June 1, 2020 were collected from global sources including national and regional health dashboards, the BMJ Global Health blog, and published articles. Reports were eligible if they provided sex-disaggregated data for total infections, ITU admissions, or mortality. Exclusions included missing total infections by sex, sample size <5, possible duplication (particularly among early reports from China), and sources with large proportions of unknown sex. For the USA, sex-disaggregated data were only available for some states; these were analyzed as separate regions. Data extraction: Researchers entered data into a structured online tool. When reports provided only percentages by sex, counts were derived by multiplying percentages by totals and rounding to the nearest integer, assuming reported percentages reflected true proportions. Cases of unknown sex were included where reported to avoid bias, as their proportion was small. Data represent country or regional aggregates and not individual-level data. Final dataset: 90 reports describing 3,111,714 infected cases across 46 countries and 44 US states; 8 reports (341,257 cases) with 12,067 ITU admissions by sex; and 70 reports (after excluding one potentially overlapping Chinese report) covering 2,751,155 cases and 214,361 deaths with sex-disaggregated mortality. Meta-analysis: Proportions and odds ratios (OR) were estimated using random-effects models. For infection proportions, the classic inverse variance method with logit-transformed proportions was used and sensitivity checked via generalized linear mixed models (GLMM), yielding identical pooled estimates. For ITU admission and mortality, ORs with 95% confidence intervals were estimated using Mantel-Haenszel (crude) and DerSimonian-Laird random-effects methods, with inverse variance weighting reflecting study size and precision. Two-sided tests assessed departure from null hypotheses (proportion = 0.5, OR = 1). Forest plots visualized study-specific and pooled estimates. Sensitivity analyses evaluated potential reporting bias. Code and data availability were provided (GitHub and DOI links).

- Infection risk: The pooled proportion of male COVID-19 cases was 0.50 (95% CI 0.48–0.51; p = 0.56; n = 3,111,714), indicating no sex difference in infection rates. - ITU admissions: Male sex was associated with substantially increased odds of ITU admission with a pooled OR of 2.84 (95% CI 2.06–3.92; highly significant). Across sources totaling 341,257 cases and 12,067 ITU admissions, men had almost threefold higher odds of ITU admission than women. - Mortality: Male sex was associated with higher odds of death with a pooled OR of 1.39 (95% CI 1.31–1.47; highly significant) across 70 reports covering 2,751,155 cases and 214,361 deaths. - Global consistency: With few exceptions, the sex bias towards worse outcomes in men was observed worldwide. - Bias assessment: Forest plots and sensitivity analyses suggested minimal influence of reporting bias for infection proportions and ITU OR estimates; the pooled mortality OR may be conservative (potential underestimate).

This meta-analysis demonstrates that male sex does not confer a higher risk of SARS-CoV-2 infection but is strongly associated with more severe clinical outcomes, including ITU admission and mortality. The consistency across diverse regions suggests that biological factors, in addition to socio-behavioral and health system factors, contribute importantly to sex differences in COVID-19 severity. Prior coronavirus epidemics (SARS, MERS) showed similar male vulnerability, aligning with a broader immunological literature where females generally mount stronger innate and adaptive responses, including higher type I interferon production and more robust humoral responses. Estrogens can enhance immune activation and antibody maturation, while testosterone may suppress immune responses, providing plausible mechanistic underpinnings for the observed disparities. These findings underscore the need to incorporate sex as a key variable in public health strategies, clinical risk stratification, and therapeutic and vaccine development, including consideration of sex-specific dosing or response monitoring where appropriate.

In a global meta-analysis of over 3.1 million COVID-19 cases, males and females had equal infection rates; however, males faced markedly higher risks of ITU admission and death. The sex bias is widespread and likely reflects, at least in part, biological differences in immune responses between sexes. These results support sex-disaggregated surveillance and the integration of sex as a critical factor in clinical management, resource allocation, and vaccine and therapeutic strategies. Future research should elucidate the mechanistic roles of sex hormones, X-linked immune genes, and age–sex interactions, and assess how comorbidities, health behaviors, and care access intersect with biological sex to influence outcomes.

Analyses were based on aggregated country- and state-level reports rather than individual patient data, limiting adjustment for confounders (e.g., age, comorbidities, testing methods, care setting). Some sources reported percentages rather than counts, necessitating back-calculation of counts and introducing rounding error. A minority of cases and deaths had unknown sex; these were included to avoid selection bias, but may introduce slight discrepancies. Heterogeneity in reporting practices and data completeness across regions (including limited sex-disaggregated data in parts of the USA) may impact estimates. Early reports from China risked duplication; efforts were made to exclude overlapping datasets. Sensitivity analyses suggested that mortality ORs could be underestimated due to reporting biases. The time window (to June 1, 2020) captures early pandemic dynamics and may not reflect later phases, treatments, or variants.