Study population and setting
This was an ecological study at the U.S. national level correlating the ratio of COVID-19 deaths to hospitalizations (case fatality rate) with the ratio of the number of COVID-19 convalescent plasma (CCP) units shipped to the number of hospitalizations (utilization rate) from August 3, 2020 to February 22, 2021. The case fatality rate was calculated from publicly available national-level data on COVID-19 admissions and deaths. The number of units of CCP shipped to US hospitals was recorded by Blood Centers of America; its use as a proxy for CCP usage was checked by correlating CCP units dispensed to those administered through the FDA’s Expanded Access Program. Mortality data was shifted by a varying number of weeks to account for a lag between hospitalization and death. The association between CCP utilization and the COVID-19 case fatality rate was assessed with Pearson’s correlation coefficient and linear regression. Confounders (e.g., mean age of hospitalized patients and total number of hospitalized patients) were considered but not included in analyses due to lack of associations with exposure or outcome. The authors then used results from linear regression to estimate deaths under counterfactual scenarios of CCP usage.
Summary of Main Findings
During the study period, there was a strong negative correlation between weekly CCP utilization and the case fatality rate (R = -0.52; p = 0.002). COVID-19 mortality was 22.3% in the lowest quintile of CCP utilization (12-24%) and decreased to 18% in the highest quintile of CCP use (41-52%). Associations between: a) the case fatality rate and the percent of admissions in those over 65 years old; and b) the case fatality rate and the number of hospitalizations were not statistically significant; these variables were not included in analyses. If the regression coefficient for CCP utilization were applied to a counterfactual scenario of CCP utilization (i.e., peak utilization from August-October 2020), the model implies that an estimated 29,018 fewer deaths would have resulted.
The authors conducted some limited sensitivity analyses (e.g., changing the lag between hospitalization and mortality, truncating the study period).
This was an ecological study, and the analysis thus did not link individual mortality with CCP administration. The study was conducted at the national level without considering county, state, or regional-level differences over time in hospitalization, mortality, and CCP utilization. Changes over time in the patient population affected by COVID-19, and in the care received by these patients, were not adequately accounted for by this analysis. Indeed, no potential confounding variables were included in the analysis and the results are based on a crude linear association. Important potential confounding variables (e.g., the underlying mortality risk profile of the patient population, hospital case burden, other treatments administered, the changing prevalence of SARS-CoV-2 variants, type/location of care facility) were unsatisfactorily considered and/or omitted.
This study does not add useful evidence to the question of whether COVID-19 convalescent plasma is effective in reducing mortality among hospitalized patients.
This review was posted on: 1 September 2021