A total of 832 articles were retrieved, of which 830 were retrieved in electronic data from PubMed, Embase, Web of SCI, and Cochrane Library, and 2 papers were retrieved in a previously published meta-analysis[14, 15].476 records were retrieved after removal of duplicates. By reading the title and abstract, we excluded the unrelated literature on the convalescent plasma treatment of COVID-19. We read the full text and selected studies that met the inclusion criteria, excluded studies that met the exclusion criteria, and studies that data cannot be extracted. A total of 7 randomized controlled trials and 1363 patients were included in the meta-analysis [2, 7, 9, 14–17](Fig. 1).
Among the included studies, there were conducted in China, India, Argentina, Bahrain, Iraq, the United States, and Brazil[7, 9]. There only were severe COVID-19 patients in three studies[15–17]. Three were severe and critically ill COVID-19 patients in three studies[2, 7, 9, 14]. Data of Severe patients and critical patients could be extracted separately in two trials[2, 7]. Convalescent plasma was compared to stand treatment in six studies[2, 9, 14–17]. Convalescent plasma was compared to control plasma. anti-SARS-CoV-2 antibody titers are tested in all studies[2, 7, 9, 14–17]. Characteristics of the included studies were shown in Table 1. Figure 2 and Fig. 3 showed the details of the RoB of all studies. Table S2 shows the certainty of results.
Compared to patients of the control group, there was no difference in clinical improvement (Four studies, RR 1.06, 95% CI 0.96 to 1.17, p = 0.22, moderate certainty) (Fig. 4) and mortality (seven studies, RR 0.86, 95% CI 0.66 to 1.11, p = 0.48, moderate certainty) (Fig. 5) for patients of convalescent plasma therapy group. We performed a sequential analysis of mortality. A total of 1,363 patients were included in our study, with an actual sample size of 3,330 required (Fig. 6). Outcome estimates were based on the following statistical indicators: the probability of type I error (α = 0.05), probability of type II error (b = 0.2), relative risk reduction (RRR = 30%), and 15% event rate in the control group. The TSA results showed that the cumulative Z value did not cross the traditional cut-off, nor did it cross the TSA cut-off, nor did it reach the required patient sample size. This indicates that further validation is still needed to verify a difference in safety between the two groups.
Time of respiratory support
The trial of Agarwal A and the trial of O'Donnell MR reported on the duration of respiratory support[7, 16]. But data on the time of respiratory support could not be converted to mean and SD. Therefore, we were unable to conduct a meta-analysis of the timing of respiratory support. Compared to control group patients, time of respiratory support of convalescent plasma group patients was no different in the trial of Agarwal A (median 9 days, IQR: 6 to 13 vs. 10 days, IQR:6 to 13, p = 0.7) and the trial of O'Donnell MR (median 6 days, IQR: 3 to 16 vs. 7 days, IQR:3 to 11, p = 0.508). The trial of Sekine L reported on the duration of time without respiratory support, and there was no difference in the duration of time without respiratory support between patients in the convalescent plasma treatment group and control patients.
Time to hospital discharge
Six trials reported on time to hospital discharge[2, 7, 9, 15–17]. The data of
time to hospital discharge could not be extracted in the trial of Li and Simonovich VA[2, 17]. In the trial of O'Donnell MR, time to hospital discharge could not be converted to mean and SD. The meta-analysis showed no difference in time to hospital discharge (three studies, MD -1.21day, 95% CI -4.78 to 2.36, p = 0.51, very low certainty) (Fig. S1) between patients in the convalescent plasma treatment group and those in the control group.
COVID-19 Nucleic Acid Negative Rate
Three trials reported negative nucleic acid rates for COVID-19 [2, 9, 16]. Two trials reported the rate of COVID-19 nucleic acid negative within 72h[2, 16]. Two trials reported COVID-19 nucleic acid negative rates within 7day[9, 16]. There was no difference in COVID-19 nucleic acid negative rate at 72h (Two studies, RR 1.62, 95% CI 0.83 to 3.16, p = 0.16, very low certainty) (Fig. S2) and 7d (Two studies, RR 1.19, 95% CI 1.00 to 1.40, p = 0.05, low certainty) (Fig. S3) between convalescent plasma therapy group patients and control group patients.
Four trials reported discharge rates at 28 or 30 days [2, 7, 9, 17]. There was no difference in discharge rate (Four studies, RR 1.06, 95% CI 0.96 to 1.17, p = 0.43, moderate certainty) (Fig. S4) between patients in the convalescent plasma therapy group and those in the control group.
We performed a subgroup analysis to investigate convalescent plasma therapy on clinical improvement and mortality rate in severe COVID-19 patients and critical COVID-19 patients. In the critical COVID-19 patients, convalescent plasma did not increase rate of clinical improvement (two trials, RR 0.97, 95% CI 0.49 to 1.92, p = 0.93, very low certainty) (Fig. 4) or decrease mortality rate (two trials, RR 0.69, 95% CI 0.40 to 1.19, p = 0.18, low certainty) (Fig. 5). In severe COVID-19 patients, convalescent plasma also did not increase the rate of clinical improvement (Tree trials, RR 1.08, 95% CI 0.98 to 1.19, p = 0.14, moderate certainty) (Fig. 4) or reduce mortality (Five trials, RR 0.89, 95% CI 0.64 to 1.24, p = 0.49, moderate certainty) (Fig. 5).