Limitations of Safety Update on Convalescent Plasma Transfusion in COVID-19 Patients

To the Editor:

Joyner et al¹ have reported a safety update on the use of convalescent plasma transfusion in a convenience sample of 20,000 hospitalized patients with coronavirus disease 2019 (COVID-19). The authors should be applauded for undertaking this monumental effort in providing the medical community with this valuable data in a timely manner amid the pandemic. Upon reflecting on the methods and results, I would like to express reservation on some of the methods used and how the findings were interpreted.

First, Table 1 offers important insights on the significant variability in the temporal trends of key variables. For example, a proportion of patients who had “current severe or life-threatening COVID-19” decreased by ∼22% from April to June 2020. The proportion of patients who had “high risk of severe or life-threatening COVID-19” increased by ∼85% in the same time interval. This may lead us to conclude that pooling the mortality proportion without stratifying by calendar month may not have been appropriate. This is also evident from Figure 2 which shows that the mortality decreased by nearly 70% from weeks 1 to 7.

Second, it seems that the reported cumulative incidence proportion of mortality is crude and was not adjusted for potential confounders. This is probably the most serious flaw in the analysis and interpretation of these data. Some of these confounders include age groups, sex, comorbidities (cardiovascular disease, hypertension, diabetes, chronic kidney disease, and lung disease), laboratory data (hematology and liver and renal function), clinical status, clinical symptoms, time since hospitalization, and calendar month (as described earlier). Furthermore, these data were collected from all over the United States where lockdown measures and stay-at-home orders varied significantly during that time interval. Such variables should have been considered as potential confounders. In addition, numerous empirical medical practices were, and many are, practiced; for example, the use of hydroxychloroquine, or experimental drugs such as remdesivir. Such data were not collected/reported in this study which makes it impossible for the authors to decisively attribute, or not, whatever they observe regarding convalescent plasma.

Third, given the purely descriptive nature of this study, this is a large case series with no comparison group; it is not appropriate to make any inferential statements. This includes the authors’ strong statements on no increased risk of adverse cardiac or thrombotic/thromboembolic events or low mortality. With a group of hospitalized COVID-19 patients who were not exposed to convalescent plasma transfusion, the authors may have been able to test a safety profile benefit or lack thereof.

Fourth, the explanation to the declining mortality rate over the study period can also be confounded, positively or negatively, by other variables, including the decreasing mean age of COVID-19 patients² who have been shown to have a substantially lower risk of mortality.³ Although it is true from their data that there were more critically ill patients later in the study than earlier, there was also a higher proportion of patients who were at higher risk of severe of life-threatening COVID-19, as we described earlier. Attributing the decline in mortality to the “expeditious” availability of convalescent plasma is invalid and may not be relevant to the study findings, once again, given the absence of a comparison group.

Fifth, time since first symptom, hospitalization, intensive care unit admission, and mechanical ventilation to convalescent plasma transfusion should have been reported to take into account their potential impact on the timing of transfusion on mortality.^{4 , 5}

Sixth, the authors state that “[this study] …support[s] the notion that earlier administration of plasma within the clinical course of COVID-19 is more likely to reduce mortality.” No data are provided to directly or indirectly support this. No comparison group, no temporal data on the natural progression of the disease in relation to convalescent plasma transfusion, no statistical adjustment for potential confounders, and no adjusted subgroup or sensitivity analyses.

Finally, comparing these 20,000 patients to a group of patients who did not receive convalescent plasma transfusion would significantly contribute to the body of evidence in this area, not only in terms of safety but also effectiveness to reduce mortality, or even better, a randomized controlled trial to assess efficacy to reduce mortality.