Characteristics and quality of studies pertaining to masks published in the Morbidity and Mortality Weekly Report

ImportanceBecause the MMWR has substantial influence on United States public health policy and is not externally peer-reviewed, it is critical to understand the scientific process within the journal. Mask policies during COVID is one topic that has been highly influenced by data published in the MMWR. ObjectiveTo describe and evaluate the nature and methodology of the reports and appropriateness of conclusions in MMWR pertaining to masks. Design, Setting and ParticipantsRetrospective cross-sectional study of MMWR publications pertaining to masks from 1978 to 2023. Main outcome measuresStudy date, design, disease focus, setting, population and location. Whether the study was able to assess mask effectiveness, if results were statistically significant, if masks were concluded to be effective, if randomized evidence and/or conflicting data was mentioned or cited, if causal statements were made about effectiveness, and, if so, whether they were appropriate. Results77 studies, all published after 2019, met our inclusion criteria. 75/77 (97.4%) studies were from the United States alone. All geographic regions and age groups were represented. The most common study design was observational without a comparator group 22/77 (28.6%). The most common setting was the community (35/77;45.5%). 0/77 were randomized studies. 23/77 (29.9%) assessed mask effectiveness, with 11/77 (14.3%) being statistically significant, but 58/77 (75.3%) stated masks were effective. Of these, 41/58 (70.7%) used causal language. Only one mannequin study used causal language appropriately (1.3%). 72/77 (93.5%) pertained to SARS-CoV-2 alone. None cited randomized data. 1/77 (1.3%) cited conflicting evidence. Conclusions and RelevanceMMWR publications pertaining to masks drew positive conclusions about mask effectiveness over 75% of the time despite only 30% testing masks and <15% having statistically significant results. No studies were randomized, yet over half drew causal conclusions. The level of evidence generated was low and the conclusions drawn were most often unsupported by the data. Our findings raise concern about the reliability of the journal for informing health policy.