arising from T. Takahashi et al. Nature https://doi.org/10.1038/s41586-020-2700-3 (2020)
The sex disparity in COVID-19 mortality varies widely and is of uncertain origin. In their recent Article, Takahashi et al.1 assess immune phenotype in a sample of patients with COVID-19 and conclude that the “immune landscape in COVID-19 patients is considerably different between the sexes”, warranting different vaccine and therapeutic regimes for men and women—a claim that was disseminated widely following the publication2. Here we argue that these inferences are not supported by their findings and that the study does not demonstrate that biological sex explains COVID-19 outcomes among patients. The study overstates its findings and factors beyond innate sex are treated superficially in analysing the causes of gender or sex disparities in COVID-19 disease outcomes.
Takahashi et al. measured more than 100 immune markers in a sample of patients with COVID-19 and uninfected healthcare workers (HCW). They compared male and female patients and HCW both at baseline and longitudinally over the disease course. These comparative analyses, both within sex and between sex, across patients and HCW, at baseline and over time, yielded more than 500 findings1. Most of the findings in the paper are presented as raw data, unadjusted for possible covariates. Among the more than 200 findings from adjusted analyses, 13 (6%) remained statistically significant after controlling for covariates (primarily age and body mass index (BMI)). This count excludes analyses on antibodies and viral load, as well as comparisons of female HCW (F_HCW) versus male HCW (M_HCW), female patients (F_Pt) versus female HCW and male patients (M_Pt) versus male HCW.
There is considerable mismatch between the claims made in the paper and the results presented in the data tables, making it challenging to understand the basis of many of these claims. The discussion section focuses on claims related to ten immune markers, positing a variety of sex differences across diverse analyses (reconstructed in Table 1). The expanded data tables demonstrate that nine of these claims are based on raw data and do not hold true in adjusted analyses. For example, interleukin-18 (IL-18) and IL-8, emphasized in the abstract and discussion as higher in male patients, show a sex difference only in baseline-unadjusted analyses of the smaller cohort. This indicates that these reported sex differences in immunological response are better explained by factors other than biological sex.
Similarly, attempting to address the potential role of these markers in disparate outcomes between men and women, Takahashi et al. associate lower levels of activated T cells at baseline with poorer outcomes among men, but not among women, in a subsample of 12 patients who deteriorated during the course of the disease (6 male and 6 female). However, as fig. 4 demonstrates, deteriorated male patients are older1. After adjusting for age, there are no sex differences in activated T cells among the patient samples.
Although statistical significance is not the only consideration when evaluating study results, the authors use statistical significance to summarize their own results and imply that the central findings remain statistically significant after adjustment. Particularly considering the sweeping scope of the study’s conclusions, combined with the study’s limited sample size, large confidence intervals, few repeat measures for many participants in the longitudinal cohort, and lack of clinical discussion of effect sizes, statistical significance remains an important guidepost for contextualizing the study’s findings.
Three findings that are described as sex differences1 are actually differences within sexes that do not correspond with between-sex differences (Table 1). For example, CCL5 differs at baseline between female patients who would later deteriorate (F_deteriorated) and those who remained stable (F_stabilized) (n = 5 F_deteriorated; 14 F_stabilized, adjusted difference: 0.39, 95% confidence interval (0.03, 0.74), P = 0.03), with no such difference among male patients who deteriorated and those who remained stable (n = 6 M_deteriorated; 10 M_stabilized, adjusted difference: 0.16, 95% confidence interval (−0.23, 0.54), P = 0.70). However, comparing the difference-in-difference, there is no evidence that the change in CCL5 between deteriorated and stabilized patients differs between the sexes (adjusted difference: 0.23, 95% confidence interval (−0.18, 0.64), P = 0.25). Such within-sex differences without accompanying between-sex differences cannot be interpreted as indicating sex-specific disease progression between men and women.
Overall, Takahashi et al. present three findings that are significant after adjustment and can properly be conceptualized as sex differences1: at baseline, numbers of non-classical monocytes (ncMono) were higher in male patients (n = 21 female and 16 male) and activated CD8 T cell numbers were higher in female patients (n = 21 female and 16 male), and male patients had higher levels of CCL5 in longitudinal analysis (n = 48 female and 43 male) (Table 1).
There are also three findings of a greater difference-in-difference that maintain significance after adjustment: at baseline, IL-8 was higher in both male and female patients compared with HCW, but the increase in IL-8 in male patients relative to male HCW was greater than the increase in female patients relative to female HCW (n = 19 F_Pt, 28 F_HCW, 16 M_Pt and 15 M_HCW); at baseline, CXCL-10 was higher in both male and female patients compared to HCW, but the increase in male patients relative to male HCW was greater than the increase in female patients relative to female HCW (n = 19 F_Pt, 28 F_HCW, 16 M_Pt and 15 M_HCW); and, in longitudinal analyses, CCL5 increased in male patients compared with male HCW, but did not differ between female patients and female HCW (n = 48 F_Pt, 28 F_HCW, 43 M_Pt and 15 M_HCW) (Table 1).
However, none of these findings of sex differences appear robust across the conducted analyses. For instance, while baseline levels of ncMono and CD8 T cells differ in the direct comparison between female and male patients, the sex difference disappears in the corresponding difference-in-differences analysis. In addition, none of the markers that do show sex differences in cohorts A and B emerge as predictive variables of interest in analyses comparing stable with deteriorated patients. While we fully recognize that immune differences would not necessarily be expected to be consistent across analyses, the lack of consistency, illustrated in Table 1, is part of a triangulating web of observations suggesting that the sex difference findings do not show a strong signal and may be artefactual.
Biological sex differences are the only causal model considered in the study. While it is plausible that sex-related biological variables may have a role in explaining sex disparities in COVID-19, strong evidence not cited by the researchers suggests a large role for social and other variables in producing the sex differences they seek to explain. For example, research demonstrates substantial variation in the magnitude and direction of the COVID-19 sex disparity across geographical localities, amongst racial and ethnic groups, and over time; these patterns are better explained by contextual factors than biological sex differences3,4,5,6. Previous research also predicts that occupational sex segregation7 and comorbidities are likely to largely explain COVID-19 sex disparities, as observed in recent SARS-CoV-1 and Middle East respiratory syndrome (MERS) epidemics8,9,10. Other studies document gender differences in conformity to COVID-19 public health guidelines11. Further research raises questions about whether aggregate patterns of higher COVID-19 mortality in men constitute a COVID-19-specific sex disparity, given men’s pre-existing higher aggregate mortality rates before the pandemic12.
Gender influences both exposure to the virus and susceptibility to severe outcomes. Occupational work segregation or adherence to behaviours such as mask wearing mediate viral load and therefore disease severity13. Chronic diseases, which are differentially distributed across men and women due to both gender- and sex-related factors, are also important contributors to COVID-19 progression and outcomes14. Notably, immune function is modified during the progression of many chronic diseases15. This is one avenue by which observed differences in immune markers may reflect gendered chronic conditions and associated immune responses rather than sex-specific biological mechanisms in response to the SARS-CoV-2 virus.
In these ways, the claims1 that sex differences in immune factors underlie COVID-19 sex disparities and merit “sex-dependent approaches to prognosis, prevention, care, and therapy for patients with COVID-19” are not only unsupported by the data, they are also not appropriately contextualized within the empirical literature on the primary role of social factors as causes of sex disparities in respiratory infectious disease epidemics.
The study by Takahashi et al.1 should be characterized as an exploratory study of possible associations between immunological variables and sex disparities in COVID-19 outcomes. The study presents largely null findings that support an assessment of male–female similarities in immune response to the SARS-CoV-2 virus. We stress that in no way does this study provide a foundation for clinical practice or for public health strategies to ameliorate COVID-19 sex disparities.
Takahashi, T. et al. Sex differences in immune responses that underlie COVID-19 disease outcomes. Nature 588, 315–320 (2020).
Mandavilli, A. Why does the coronavirus hit men harder? A new clue. The New York Times https://www.nytimes.com/2020/08/26/health/coronavirus-men-immune.html (26 August 2020).
US gender/sex COVID-19 data tracker. Gender Sci Lab https://www.genderscilab.org/gender-and-sex-in-covid19 (2020).
Islam, N., Khunti, K., Dambha-Miller, H., Kawachi, I. & Marmot, M. COVID-19 mortality: a complex interplay of sex, gender and ethnicity. Eur. J. Public Health 30, 847–848 (2020).
Dehingia, N. & Raj, A. Sex differences in COVID-19 case fatality: do we know enough? Lancet 9, e14–e15 (2021).
Rushovich, T. et al. Sex disparities in COVID-19 mortality vary across US racial groups. J. Gen. Intern. Med. 36, 1696–1701 (2021).
Adams, R. B. Gender equality in work and COVID-19 deaths. Covid Economics (11 May 2020).
Chan-Yeung, M. & Xu, R. H. SARS: epidemiology. Respirology 8 (Suppl), S9–S14 (2003).
Jia, N. et al. Case fatality of SARS in mainland China and associated risk factors. Trop. Med. Int. Health 14 (Suppl 1), 21–27 (2009).
Yang, Y.-M. et al. Impact of comorbidity on fatality rate of patients with Middle East respiratory syndrome. Sci. Rep. 7, 11307–11309 (2017).
Galasso, V., et al. Gender Difference in COVID-19 Related Attitudes and Behavior: Evidence from a Panel Survey in Eight OECD Countries; https://www.nber.org/papers/w27359 (National Bureau of Economic Research, 2020).
Krieger, N., Chen, J. T. & Waterman, P. D. Excess mortality in men and women in Massachusetts during the COVID-19 pandemic. Lancet 395, 1829 (2020).
Gandhi, M., Beyrer, C. & Goosby, E. Masks do more than protect others during COVID-19: reducing the inoculum of SARS-CoV-2 to protect the wearer. J. Gen. Intern. Med. 35, 3063–3066 (2020).
Zhang, J. et al. Risk factors for disease severity, unimprovement, and mortality in COVID-19 patients in Wuhan, China. Clin. Microbiol. Infect. 26, 767–772 (2020).
Hotamisligil, G. S. Inflammation and metabolic disorders. Nature 444, 860–867 (2006).
The authors declare no competing interests.
Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
About this article
Cite this article
Shattuck-Heidorn, H., Danielsen, A.C., Gompers, A. et al. A finding of sex similarities rather than differences in COVID-19 outcomes.
Nature 597, E7–E9 (2021). https://doi.org/10.1038/s41586-021-03644-7