Demographic characteristics of COVID-19 patients
The study enrollment period was from the 8th of April to the 14th of June 2020. During that period, a total of 204 patients from 14 Dutch hospitals with a RT-PCR confirmed SARS-CoV-2 infection and admitted for a moderate, severe or life-threatening COVID-19 infection were screened for eligibility. All had symptomatic COVID-19 disease as evaluated by study physicians according to the guidelines set by the Dutch National Institute for Public Health and Environment. (25) The most common reason for patients to decline participation was fear of adverse events (Supplementary figure 1). A total of 86 COVID-19 patients were enrolled and randomized to standard of care (n=43) or treatment with ConvP (n=43) (Table 1).
Overall, 72% of the patients were male, and median age was 63 years (IQR 56 – 74). At inclusion, they had COVID-19 related symptoms for a median of 10 days (IQR 6 – 15) and had been admitted to the hospital for 2 days (IQR 1 – 3) in line with what was previously reported as median duration of symptoms at hospitalization. (26, 27) A total of 13 patients were admitted to the ICU and mechanically ventilated, all for no longer than 96 hours at inclusion. Patients randomized to standard of care (SoC) experienced symptoms for a median of 2 more days at inclusion and more people had WHO disease severity scores ≥4 (98% versus 84%) or ≥6 (ICU admission (19% versus 12%). Four out of 5 blood biomarkers associated with an unfavorable COVID-19 disease course (CRP, ferritin, LDH, lymphocyte count) were slightly more disadvantageous in the SoC group. The total number of comorbidities were 56 and 57 respectively. From 66 of the patients, blood samples for additional immunological evaluations could be retrieved at baseline. Of these, 34 were in the ConvP arm and their baseline characteristics were balanced with the overall group (Supplementary table 1).
Clinical outcomes
Of the 43 patients randomized to ConvP, 6 (14%) died while 11 of the 43 (26%) SoC patients died leading to an unadjusted odds ratio of 0.47 (95% CI: 0.15 – 1.38) for death. While numerically higher, the predefined primary endpoint in the study protocol was an adjusted analysis of the overall mortality at day 60 after enrollment for patients treated with ConvP of which the odds ratio was 0.95 (95%CI: 0.20 – 4.67, p=0.95) (Supplementary table 2). This adjusted analysis of mortality accounts for consistently reported predictors of death in COVID-19 patients. (28-32) By adjusting for the same confounders in a proportional odds ordinal logistic regression model we identified age specific probabilities for being scored in specific categories of the 8-point WHO COVID-19 disease severity score at day 15 (Figure 1A) and day 30 (Figure 1B) after randomization. Notably, these probabilities were comparable between both study arms for all scores at day 15 (p=0.58) and day 30 (p=0.67) throughout the study (Supplementary table 3 and 4). An identical number of 25 (58%) patients in the ConvP and 25 (58%) in the SoC group had improved by day 15 on the WHO COVID-19 disease severity score (adjusted odds ratio 1.30, 95%CI: 0.52 – 3.32, p=0.58). Treatment with ConvP was also not associated with earlier discharge in Cox Regression analysis (adjusted hazard ratio 0.88, 95%CI: 0.49 – 1.60, p=0.68), also when cumulative incidences were corrected for the competing risk of death (Figure 1C and Supplementary figure 2 and Supplementary table 5). The median duration of admission was 8 days (IQR 3 – 21) and 8 days (IQR 4 – 19) for ConvP versus SoC. Patients who died were 69 years (IQR 63 – 84) and 12 (71%) were men. Baseline data of patients who survived or died in each of the study arms are available in the supplementary table 6. No serious adverse events possible related to ConvP were observed.
Overall, no difference was observed between the highest measurements of CRP or ferritin (the 2 biomarkers included in the electronic case report form) 7 and 14 days after enrollment between groups (Figure 1D and Supplementary table 7). In a subset of 34 patients with data available, we noticed that the median absolute lymphocyte counts in the peripheral blood at these time points were comparable and followed similar trends in recovery in each treatment group within 2 weeks after enrolment.
Donor characteristics
Collectively, 3200 recovered COVID-19 donors volunteered in April 2020 as ConvP donors. The first 115 patients were selected who had RT-PCR proven COVID-19 disease, fulfilled the other inclusion criteria for ConvP collection, had a determination of anti-SARS-CoV-2 neutralizing antibody responses (Supplementary table 8) and completed the online questionnaire regarding donor characteristics. Hundred-five of the 115 donors were male, the median age was 43 years and they had been symptomatic for a median of 12 days (IQR 8 – 18). Their disease course had been generally mild reflected by a 12% admission rate for COVID-19. Overall, we detected virus specific total Ig and IgM antibodies against SARS-CoV-2 receptor binding domain (RBD) by ELISA in serum samples of 114 of 115 (99%) donors at median 34 days in their convalescent phase. The median total Ig and IgM optical density ratios in all donors were respectively 15.08 (IQR 8.60 – 18.41) and 4.03 (IQR 0.96 – 14.33). Although optical density ratios from ELISAs correlate with neutralization capacity against SARS-CoV-2 substantial outliers with lower than expected neutralizing antibody titers are often observed and the correlation plateaus at increasing antibody levels leading to a loss of discriminative capacity to detect plasma with very high neutralizing antibody titers. (33) Therefore, a plaque reduction neutralization test (PRNT) using the whole SARS-CoV-2 virus was used for the selection of donors for the study. In 110 of 115 donors (96%) tested, neutralizing antibodies could be detected. The median PRNT50 titer was 1:160 (IQR 1:80 – 1:640) with 78% and 43% having a PRNT50 of at least 1:80 or 1:320 respectively. PRNT50 titers of 1:80 and 1:320 were previously shown to predict a <5% and <1% chance of demonstrating replication competent virus in the upper respiratory airway of COVID-19 patients. (34) A titer above 1:80 was defined as the minimum neutralizing capacity required for a donor to be eligible for ConvP donation. Of the 19 donors of whom ConvP was eventually used, all but 2 had a PRNT50 titer of at least 1:320. These 19 selected individuals all had mild disease without hospitalization and a more recent resolution of symptoms (20 days) than other donors.
Immunological analyses in COVID-19 patients
Serum was available from 66 subjects for PRNT50 and ELISA testing at inclusion. Logistical issues at the peak of the pandemic prevented the collection of serum from the remaining 20 patients. At inclusion, 80% tested positive with the total Ig anti-SARS-CoV-2 RBD antibody test with optical density ratios at 14.80 (IQR 1.84 – 18.41) while IgM anti-SARS-CoV-2 RBD antibodies were present in 77% with optical density ratios at 5.98 (IQR 0.86 – 25.19) (Figure 2A). Interestingly, the 19 selected donors had levels of anti-SARS-CoV-2 RBD total Ig (optical density ratio 18.39, IQR 14.19 – 18.39) that were comparable (p=0.78) with the baseline levels in hospitalized patients who had been symptomatic for a median of 10 days while anti-SARS-CoV-2 RBD IgM tended to be higher (optical density ratio 25.13, IQR 7.55 – 33.32; p=0.02). The optical density ratio of the total Ig (p=0.71) and IgM (p=0.83) anti-SARS-CoV-2 RBD antibody between patients in the ConvP and SoC group were however not meaningfully different (Figure 2B).
We further confirmed the presence of anti-SARS-CoV-2 antibodies in the serum of COVID-19 patients by measuring anti-nucleocapsid (N protein) IgM and IgG antibodies and comparing COVID-19 patients with a group of convalescent plasma donors. Fifty-one patients had anti-nucleocapsid IgM and IgG antibodies with a median of 5.65 units (IQR 2.59 – 11.73) and 16.19 units (IQR 6.91 – 30.48), respectively (Figure 2C). Donors (n=54) had anti-nucleocapsid IgM and IgG antibodies with a median of 3.62 units (IQR 2.20 – 6.28) and 20.25 units (IQR 13.74 – 30.69), respectively. Using a cut-off of 11 units, 14 patients were positive for IgM antibodies (27.5%) and 31 for IgG antibodies (60.8%) while from the donors 11 were positive for IgM antibodies (20.4%) and 45 for IgG antibodies (80.3%). None of 9 healthy controls had IgM or IgG antibodies against SARS-CoV-2 nucleocapsid above the cut-off.
Next, to explore the functionality of detected antibodies, we used viral neutralization tests with SARS-CoV-2 from the same serum samples of patients and ConvP donors in whom the presence of anti-SARS-CoV-2 RBD specific antibodies were determined. This was possible in 56 of 66 enrolled patients due to limited serum availability after antibody testing in 10 patients. To our surprise, in 44 (79%) patients, neutralizing antibodies at PRNT50 of ≥1:20 were detected at median 1:160 (IQR 1:20 – 1:1280) (Figure 3). As expected, this titer correlated (r2=0.36, p=0.07) with the duration of symptoms (Supplementary figure 3), and was comparable to the median titer observed in all 115 donors tested (p=0.4). The median PRNT50 of the 19 selected donors (1:640, IQR 1:320 – 1:1280) was however higher p=0.011, with 90% having a PRNT50 titer ≥1:320 compared to 46% (p=0.001) in patients. Only 2 of the 19 selected donors had titers <1:320 and these were administered to 2 patients with PRTN50 titers at >1:2560 and 1:640 respectively at baseline. These patients were admitted for 12 and 13 days and both survived throughout day 60.
To independently confirm the high virus neutralizing anti-SARS-CoV-2 antibody levels in the included patients, we additionally tested serum from 37 RT-PCR confirmed COVID-19 patients from the month preceding the start of the study from whom serum samples from <72h after hospital admission to a non-ICU ward were available at the Erasmus MC university medical center. With a median age of 65 years (IQR 56 – 74), 60% males and symptom duration of 9 days (IQR 4 – 13) these patients were comparable to the study population, as were their biomarkers that predict disease outcomes (Supplementary table 9). We found that 26/37 (70%) of these patients had anti-SARS-CoV-2 Ig antibodies including 23/37 (62%) at a ratio >10, indicating the presence of neutralization capacity based on our previous observation (Supplementary figure 4). (34)
Finally, to assess whether ConvP treatment had a more indirect effect on the COVID-19 disease course by potentially dampening the inflammatory response, we measured serum pro-inflammatory cytokines IL-6, TNFα, IFNγ, IL-1β, IL-2, IL-4, IL-10 and IL-12p70. We investigated 9 patients that received ConvP and 10 SoC for which we had a complete set of serum samples for the first 2 weeks post inclusion in the study. For these patients, we compared the cytokines levels at enrollment (day 1), and at days 7 and day 14 after enrollment. There was no difference between the treatment arms on day 1 for IL-6, TNFα and IFNγ. Importantly, the decrease in cytokine levels on day 7 and 14 were comparable between patients receiving ConvP and standard of care both for concentration (Figure 4A) and fold change from day 1 (Figure 4B). No differences were seen between groups for IL-1β, IL-2, IL-4, IL-10 and IL-12p70 (Supplementary table 10). The above provides further evidence that plasma therapy had no effect on the inflammation and course of COVID-19.
Virological analyses in COVID-19 patients
Nasopharyngeal swabs were taken of 51, 54, 53, 15 and 45 patients at enrollment (day 1) and day 3, day 7, day 10 and day 14 after enrollment respectively. In unadjusted analyses, the proportion of samples where SARS-CoV-2 genome was detectable by RT-PCR was higher in the SoC group at day 1 (82% vs 67%), day 3 (79% vs 46%), and day 14 (21% vs 9%) compared to patients in the ConvP group. The calculated median SARS-CoV-2 viral loads were higher in the SoC group at inclusion (5,7x103 copies/ml, IQR: 7,3x102 – 4,7x104 vs 1.4x103 copies/ml, IQR: 0 – 1.7x104) and day 3 (1,1x103 copies/ml, IQR: 0 – 8,2x103 vs 0 copies/ml, IQR: 0 – 8.7x101). Thus, the apparent lower virus shedding in the upper airway was already present before ConvP initiation and this remained so over time in the treatment group without major appreciable influence of ConvP therapy (Figure 5A). Indeed, after adjustment in a mixed model by covariates associated with COVID-19 disease severity as we had pre-defined in the study protocol (Supplementary table 11), the slope of the viral load decay from day 1 to 14 was estimated to be less steep in the ConvP group at -0.4log copies/mL (95%CI -0.7 – -0.1) overall (Figure 5B) and comparable when considering only patients with detectable SARS-CoV-2 in nasopharyngeal swabs at enrollment (Figure 5C).
With regard to virus viability, we had 12 patients (6 in each arm, median 8 days of symptoms at inclusion) where we were able to do culture samples for SARS-CoV-2 replication, obtained after median 2 days following inclusion. Although a systematic collection of cultures before inclusion was not required per protocol and current knowledge indicates that viral cultures tend to become negative around 8 days of symptoms, (34, 35) none of the 6 patients from the SoC arm and 1 patient on ConvP had cytopathic effects in culture after 7 days of incubation, again signaling no added value of ConvP. This patient with a sample containing replication competent virus had 6 days of symptoms at inclusion, without antibodies detected then, and died during follow-up.