Effect of convalescent plasma transfusion on outcomes of coronavirus disease 2019: a meta-analysis with trial sequential analysis

As of 18 December 2022, the severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) has been reported to be responsible for approximately 650 million confirmed cases and more than 6.6 million deaths worldwide [1]. Despite optimistic expectations, specific therapeutic options are still limited and efforts at mass vaccination have fallen short of offering long-lasting immunity or providing effective coverage of the entire population, especially in underdeveloped countries [2, 3]. Disappointingly, latest data from the World Health Organization (WHO) indicate that over 3.7 million new cases are still being reported and over 10,000 victims succumb to COVID-19 each week. Besides, there is good evidence that currently reported incidence rates do underestimate the actual incidence owing to the progressive relaxation in COVID-19 testing strategies adopted worldwide resulting in fewer tests being performed and subsequently fewer cases being detected [1].

Transfusion of convalescent plasma (CP) obtained from donors who recovered from a recent SARS-CoV-2 infection has been suggested as a treatment option for coronavirus disease 2019 (COVID-19). Providing passive immunity against the virus via transfusion of high titers of neutralizing antibodies contained in CP is the rationale underlying its use in this context [4].

Convalescent plasma transfusion (CPT) has been previously employed to combat similar viral outbreaks with promising results. Evidence from these trials suggests that CPT could be more effective in patients with high viral load and in those with more progressive or more severe forms of the disease. Moreover, it has been suggested that early administration of CP could improve outcome under such circumstances when neutralizing antibodies have not been produced in adequate amount by the host and the viral load is yet high [5]. On the other hand, the risk of transfusion-related reactions does not seem to be significantly higher in this setting than that associated with plasma transfusion for other indications [3]. Consequently, the concept of utilizing CPT has been extended to the current SARS-CoV-2 problem but with inconsistent results, even at the highest level of evidence as obtained from meta-analyses [6‐8].

The present meta-analysis updates the results of an antecedent meta-analysis published over 1 year ago [9]. As more evidence is being added from newly appearing publications and in absence of convincing indications that the problem of COVID-19 has vanished, we deemed it pertinent to update our evidence on the benefit or futility of CPT in management of COVID-19. Our prime aim was to re-examine the effect of using CPT on mortality and need of invasive mechanical ventilation (IMV) in patients with COVID-19 as seen in the light of currently accumulating evidence.

Searching the literature returned 4,451 relevant titles. After exclusion of duplicate records (n = 817), 3,634 reports were examined by title and abstract, out of which 3,525 titles were excluded for ineligibility. The remaining 109 records were examined by full text. Eighty-Three full-text articles were excluded because they were non-randomized studies of intervention (NRSI) (n = 40), case reports or case series (n = 16), single-arm clinical studies with no comparator group receiving standard of care only (n = 24), interim analysis for a study published later (n = 1) or were published in languages other than English with only the abstract provided in the English language (n = 2). Twenty-Six RCT including 19,816 patients were eligible. All 26 trials provided valid information on mortality that was utilized for quantitative synthesis (meta-analysis) regarding this outcome measure [16‐41]. Endpoint of mortality varied among studies. Thirteen trials examined 28-day mortality [16‐21, 26, 29, 32, 33, 37, 38, 40], six trials examined 30-day mortality [22, 23, 31, 36, 39, 41], three trials examined 60-day mortality [25, 28, 34] and a single trial examined 90-day mortality [24]. Three trials broadly reported in-hospital overall mortality [27, 30, 35]. As regards need of IMV, 17 trials (16,083 patients) provided valid information for quantitative synthesis. Search results are illustrated in Fig. 1 and characteristics of included studies are shown in Online Resource 1. Risk of bias assessment for individual trials is shown in Online Resource 2 and overall risk of bias in included trials is shown in Online Resource 3. The majority of included studies (22/26) were qualified as high-quality [16‐19, 26, 28‐34, 37‐41], fulfilling all five criteria of the RoB 2 Tool [11]. One or more criteria of the RoB 2 Tool [11] were either unclearly reported or completely missing in four studies [20, 27, 35, 36], which were regarded as low-quality.

Convalescent plasma transfusion has been advocated as a promising therapy for COVID-19. The results of clinical trials and reviews, however, have been inconsistent, and usually inconclusive. The present review showed that information is currently adequate to draw out clear conclusions with acceptably high certainty regarding the benefit or futility of CPT for patients with COVID-19. Firstly, the present evidence suggests that adding CPT to standard treatment is not associated with reduced mortality or need for IMV compared with standard treatment alone. Secondly, certainty for lack of benefit of adding CPT to standard treatment is high. Thirdly, cumulated information size has been large enough and is opting for futility of CPT. Consequently, further conduct of RCT to seek a possible benefit of CPT may not be justified.

CP has been previously utilized for management of other coronavirus-induced infections such as severe acute respiratory syndrome (SARS) and Middle East respiratory syndrome (MERS). The rationale underlying use of CP in these settings is to chelate the culprit viruses by neutralizing antibodies present in high titer in plasma of subjects who have just recovered from a recent coronavirus infection [5]. Use of CPT has been extended to the setting of COVID-19 with rather inconsistent conclusions [6‐8, 43].

The present meta-analysis updates the results of a previous meta-analysis published over one year ago [9]. According to the Cochrane guidelines, the need to update a meta-analysis is determined by two fundamental considerations, whether the health problem is still relevant, and whether a sufficient number of new studies have been made available [44, 45]. The latter issue is particularly pertinent since the primary aim of a meta-analysis is to provide an answer to the question of interest as obtained from best available evidence. As a significantly large size of new information is made available, previously obtained evidence, though once valid, may become invalidated. From this perspective, the whole process is basically a function of how much new information has accumulated and how much the problem is still thriving rather than how long has elapsed since a meta-analysis was made available to readers [44, 45]. Since the review of Ling and coworkers was published [9], eight more eligible RCT including 3,044 patients have been added to the published database [23, 26, 27, 31, 32, 37, 40, 41]. This fact, combined with ongoing reports from authentic organizations about resurgence of new cases of COVID-19 and the global underestimation of the real incidence of such new cases [1], has compelled updating our evidence on this issue.

Although the present review corroborates the findings of that of Ling and colleagues [9], it has got some notable merits. Besides the more robust information size of 26 studies encompassing approximately 20,000 patients, the present review has targeted a larger effect size corresponding to an RRR of 20%. Based on an assumed mortality rate of 20% in patients receiving standard of care (SOC) only, this is translated to an absolute reduction in mortality of 4%, a clinically more relevant effect size compared with the rather meager RRR of 10% targeted by previous meta-analysts which is equivalent to an absolute reduction in mortality of as small as 2% by CPT [9]. The implication of this contrast is that the demonstrated futility of CPT by the present meta-analysis is even more decisive in view of the improved chances to demonstrate a benefit to CPT, if any, which has been achieved by the enhanced power of the analysis as a function of the larger information size as well as the bigger target effect size that is more clinically meaningful.

To the present authors’ knowledge, the meta-analysis of Ling and coworkers [9] has been preceded by four previous meta-analyses that provided inconsistent conclusions [6‐8, 43]. In one meta-analysis including RCT and NRSI, the authors concluded that their results favored the efficacy of CP as a therapeutic for COVID-19 [6]. However, a deliberate examination of their results may warrant some reconsideration of these conclusions. The authors of that review reported that pooling of NRSI alone showed clear benefit of CPT on mortality while RCT showed no statistically significant benefit. The authors conducted sensitivity analysis after they excluded a large RCT [17] accounting for over one-third of statistical weight that had a directionally different effect size. This elimination redirected the effect size toward favoring CPT over SOC alone. The pretext for excluding that study from the analysis was the low level of SARS-CoV-2 antibodies in approximately two-thirds of patients in the interventional arm. However, that exclusion may have been inappropriate for the following reasons. Firstly, the authors of the excluded trial did conduct a modified intention-to-treat analysis where they compared the outcomes of patients receiving CPT with non-detectable neutralizing antibody titers, detectable neutralizing antibody titers or neutralizing antibody titers of 1/80 or higher versus controls receiving SOC only and found no statistically significant differences among the four subgroups [17]. Secondly, most trials investigating the effect of CP have not provided adequate information regarding the antibody titers in donated CP. Thirdly, there is little consensus on what should be considered as the minimum titer of SARS-CoV-2 antibodies in this clinical context. Thus, taking into consideration the potential bias induced by exclusion of an influential RCT together with reliance on evidence from NRSI, the conclusions of that meta-analysis would be questionable. Similarly, the meta-analysis by Kloypan and colleagues [8] showed that the benefit of CPT varied with the study design, the statistically significant benefit for mortality observed by pooling clinical and observational trials together being downgraded to just a trend for better outcome when only RCT or double-blinded RCT were analyzed. By the same token, an earlier meta-analysis [7] found low-certainty non-conclusive evidence from eight RCT, and low-certainty evidence from 13 cohort studies for reduction in mortality at 28 days. Another meta-analysis by Bansal and coworkers [43] provided somewhat similar results; pooling all studies showed a benefit of CPT on reducing mortality, and CPT was still beneficial when prospective or retrospective studies were pooled separately. Nonetheless, pooling RCT separately showed no statistically significant reduction in mortality. Interestingly, those authors emulated the same methodology of previous reviewers [6] and conducted sensitivity analysis excluding the study of Agarwal and colleagues [17].

The present review has got some strengths. Besides the considerably larger information size than previously published meta-analyses both in terms of number of patients and in terms of number of events, there has been little possibility of methodological bias in included RCT, the vast majority of which (22 out of 26 trials) fulfilled all five criteria of the RoB 2 Tool [11] and were qualified as at low risk of bias. Secondly, the present authors made use of the GRADE system [15] to determine objectively the level of certainty in obtained evidence. Inclusion of high-quality RCT exhibiting evident consistency and precision of estimated effect sizes has enhanced the level of certainty in the present meta-analysis and allowed the authors to grade the evidence as high. Thirdly, the present meta-analysis has made use of TSA [46] and has targeted more practical parameters for this type of analysis in order to draw empirically meaningful conclusions.

On the other hand, the present meta-analysis has got some limitations. Firstly, we included only studies published in English. Therefore, it is likely that a large number of relevant studies have been trimmed off by this restrictive criterion. Secondly, we limited our review to studying only two outcomes, mortality and need of IMV. Although most trials were concerned with other measures of clinical or biochemical improvement, the present authors found most of these outcome measures difficult to define or to quantify objectively from available data. Besides, from a clinician’s viewpoint, this couple of outcomes is almost unanimously held as the most decisive measure of success or failure of COVID-19 management. Thirdly, for understandable reasons, studies included in the present meta-analysis did not adopt a uniform time-frame for assessment of their main end-points of interest. So, while some authors limited their interest to in-hospital mortality [27, 30, 35], others extended surveillance to as long as 90 days from inclusion [24]. A related issue is the evident diversity among trials regarding the standard treatment offered to patients. This is foreseeable in view of the inevitable time lag that had to elapse between emergence of the pandemic and accumulation of enough information that enabled epidemiologists and clinicians to formulate justifiable frameworks for patient management. Besides, developed guidelines have been subject to a dynamic process of continuous reappraisal and update as newer pieces of information are being made available. This applies to internationally [47] as well as to regionally developed guidelines [48]. For instance, the National Institute of Health (NIH) issued the first practice guidelines for COVID-19 in April 2020. Since then, the guidelines have undergone over 60 updates pragmatically translating newer evidence into clinical practice. Though missing from the original guidelines, the NIH added a statement for corticosteroids in the June 2020 updates, then added statements for CPT and ivermectin in the September 2020 and January 2021 updates, respectively [48]. This incessant tuning up of recommendations and practice guidelines has certainly contributed to the diversity in what investigators had to offer to their patients as standard of care. Likewise, studies varied widely regarding the severity of disease at inclusion which ranged from just mild disease [30, 40] to critical illness [24, 35]. Most of the studies, however, included patients with either severe [18‐21, 23, 28, 29, 32‐34, 36, 37] or moderate and severe disease [22, 25, 26, 31, 41], while few studies exclusively recruited patients with moderate disease [17] or with disease of any degree of severity [16, 27]. These methodological inconsistencies may render interpretation of present results rather intricate. Nonetheless, despite this apparent variability, it is noteworthy that tests of heterogeneity did demonstrate remarkable consistency across studies as regards the estimated effect size, while sensitivity analyses did reveal practically unchanged size and direction of the overall effect as studies were sequentially eliminated. In fact, such reproducibility or consistency of effect is regarded by the GRADE system as a criterion enhancing the strength of evidence [15]. Fourthly, as a possible source of confounding, we detected evidence of publication bias in favor of studies displaying effectiveness of CPT in reducing mortality or need of IMV. Although this could be a reason of concern, at least theoretically, we employed the trim-and-fill method [14] to impute missing studies and adjust point estimates accordingly. In effect, difference between adjusted and unadjusted point estimates was too trivial to be of any clinical value and we opted not to downgrade the level of evidence obtained from the GRADE system [15].

Availability of new information combined with resurgence of the coronavirus disease 2019 problem has impelled updating our evidence on using convalescent plasma transfusion as a possible treatment. The size of available information is adequate to conclude with high level of certainty that convalescent plasma therapy added to standard treatment of COVID-19 is not associated with reduced mortality or need of invasive mechanical ventilation compared with standard treatment alone. In view of these findings, further trials on efficacy of convalescent plasma therapy in coronavirus disease 2019 patients are probably not needed.