2.1 Study Design

A single-center, open label, Phase I/II prospective non-randomized trial (Trial NL9379) was conducted at the Erasmus Medical Center (Rotterdam, the Netherlands). The protocol was approved by the Dutch competent authority (CCMO) and the institutional review board (METC) at Erasmus MC. Written informed consent was obtained from all patients.

Based on the availability of ConvP and later of COVIg, the inclusion of a total of 104 patients was planned across groups of different doses and products. In supplementary material S1, Figure S1, the study design is presented in detail. Treatment allocation (ConvP/COVIg) was open-label. Eight study arms were created with different volumes and concentrations. We decided to first recruit patients in the COVIg arms since no ABO compatibility is required. The batches with high neutralizing antibody (Nab) titers were tested first because this minimized the risk of underdosing. Also, when a rapid clearance of Nab in patients receiving the high-titer COVIg would be observed, testing with the low Nab-titer batches would no longer be required. Patients in the ConvP group were allocated to a predefined volume and Nab-titer based on ABO compatibility. Patients included in the COVIg arm could participate a second time in a ConvP arm of the study at the time they had become SARS-CoV-2 spike antibody negative.

2.2 Study Products

Convalescent plasma was provided by the Dutch blood bank (Sanquin Blood Supply). Donors met the standard plasma donor criteria, had a history of symptomatic COVID-19, and had recovered for at least 14 days. Hyperimmune globulin was manufactured by Prothya Biosolutions and provided by the Dutch Ministry of Health, Welfare and Sport. These particular batches were derived from pooled plasma from at least 1000 donors, including ConvP donations. Both products were produced while the ancestral variant (Wuhan-1) type was dominant in the Netherlands and, therefore, the Nab-titer against this strain was measured. The Nab-titer was measured by using a virus neutralization test, methods of this test are described in the supplementary material (S2). Antibody treatment was administered intravenously. Since ConvP was collected before anti-SARS-CoV-2 vaccines had become available, plasma with very high Nab-titers was rare and the majority of the ConvP from these non-vaccinated donors had a Nab-titer ranging between 270 and 500 IU mL−1. In this study, ConvP with a Nab-titer of 500 and 910 IUmL−1 was used and is referred to as intermediate-titer and high-titer ConvP, respectively. By pooling regular plasma with ConvP, Prothya was able to produce two batches of COVIg with an increased Nab-titer of 270 and 910 IU mL−1. These products will be referred to as low-titer and high-titer COVIg, respectively. The Nab-titer, given in IUmL−1, is a unit of antibody neutralization of the ancestral SARS-COV-2 variant, as described by Nguyen et al. It facilitates the comparison of Nab-titers between a broad range of in-house virus neutralization tests [20].

Because IgG titers correlated well with neutralization assays of the ancestral virus, the LIAISON® SARS-CoV-2 TrimericS IgG assay (DiaSorin) was used for measuring anti-SARS-CoV-2 IgG antibodies [21]. This strategy avoided potential capacity issues in the laboratory since virus neutralization tests are much more labor intensive. In addition, the turn-over time of virus neutralization tests is much longer. The LIAISON® SARS-CoV-2 Trimerics IgG assay comprised a chemiluminescence immunoassay (CLIA) determining the anti-trimeric spike protein-specific IgG antibodies. The assays manual and the performance characteristics can be found in supplementary material S3 Results of the CLIA are reported in binding antibody units/mL (BAUmL−1), which is the preferred unit for binding capacity by the WHO [21,22,23]. Throughout this manuscript, we will report results from this test as the CLIA antibody test (CAT) titer in BAUmL−1. Chemiluminescence immunoassay was also performed on 11 of the 13 administered ConvP units from which median binding capacities were 3230 BAU mL−1 and 3070 BAU mL−1 for the intermediate and high Nab-titer, respectively. Chemiluminescence immunoassay was performed ten times on the high Nab-titer COVIg batch (910 IU mL−1), from which a median of 3985 BAU mL−1 was obtained. For a rough estimate of the Nab-titer in IU mL−1, the result of the LIAISON® SARS-CoV-2 TrimericS IgG assay can be divided by 4, more details can be found in supplementary material S4.

2.3 Patient Selection

Patients were aged at least 18 years and had no anti-SARS-CoV-2 antibodies at baseline. First, patients who had received B-cell-depleting therapy were included but after the start of the vaccination campaign, all patients lacking anti-SARS-CoV-2 antibodies at least two weeks after full vaccination (two mRNA vaccines, two adenovirus vector vaccines [ChAdOx1-S], or one adenovirus vector vaccine [Ad26.COV2.S]) could participate in the study as well. Patients were screened with a point-of-care antibody test (Roche SARS-CoV-2 Rapid Antibody Test®). Negative test results were verified by the DiaSorin CLIA test and were deemed negative if CAT titers were < 33.8 BAU mL−1 according to the manufacturer’s instructions [21]. Patients had no symptoms of SARS-CoV-2 infection and tested negative with a qPCR test at the time of screening for the study.

2.4 Clinical and Biochemical Monitoring

SARS-CoV-2 spike antibody measurement was performed using CLIA at baseline and, subsequently, after 24 and 48 hours and after 1, 2, 4, 6, 8, 12, 18, and 24 weeks or until the CAT titer had become negative (< 33.8 BAU mL−1) again. Blood sampling was also halted if the patient received another anti-SARS-CoV-2 vaccination during follow-up or had a breakthrough infection.

2.5 Primary Endpoints2.5.1 Population Pharmacokinetic Analysis

To perform a population PK analysis, the measured CAT titers versus time curves from ConvP and COVIg were described using non-linear mixed-effect modeling with NONMEM v7.4 (ICON Development Solutions, Ellicott City, MD, USA), which was guided using PsN v4.9.0. Pirana v2.9.9 was used for model management and R v4.2.1 (R Core Team, 2020) with Xpose v4.5.3 were used for graphical model diagnostics [24,25,26]. For obtaining the model parameters, first-order conditional estimation (FOCE) was applied with epsilon-eta interaction.

2.5.2 Model Development

Model development commenced with evaluating the most parsimonious compartment model to describe the CAT titer versus time data with initial parameter values obtained from the literature [27, 28]. For constructing the statistical model, the residual unexplained variability (RUV) was evaluated using an additional, proportional, or mixed (additive and proportional) error model. Moreover, inter-individual variability (IIV) was evaluated for each parameter separately using a log-normal distribution. Inter-occasion variability (IOV) was not estimated, as only data from one dosing event were collected. Model building was conducted using a stepwise approach, as the addition of parameters to the model was evaluated one by one.

2.5.3 Covariate Analysis

In the covariate analysis, the patient and treatment characteristics were used to explain the obtained IIV for the model PK parameters. Selection of the covariate relationships was based on biological and clinical plausibility. For evaluating the dichotomous covariate relationships, the following model was applied:

$$_},i}=_}}\times _}}}^}},$$

in which θTV,i is the typical value for the individual patient i, θpop is the population PK parameter value, θcov the parameter describing the covariate effect and COV is the covariate value being 1 if present and 0 otherwise. For the continuous covariate relationships, the following relationships (linear, power, and exponential) were applied:

$$_}}=_}}\times \left(1+_}}\times \left(}-}}_}}\right)\right),$$

$$_}}=_}}\times }}}}_}}}\right)}^_}}},$$

$$_}}=_}}\times }}^_}}\times \left(}-}}_}}\right)\right)},$$

in which COVmed is the median for the covariate value. Before applying a covariate model, the plausibility of that relationship was first evaluated using graphical exploration. The covariate model was constructed using a standard forward inclusion (p = 0.05, df = 1) and backward elimination (p = 0.01, df = 1) procedure. For a list of the evaluated covariate relationships, see Supplementary material S5.

2.5.4 Model Evaluation

The ability of the model to predict the CAT titer measurements was described using an objective function value (OFV). As the OFV is χ2 distributed, the difference between the OFVs (dOFV) from two hierarchical models was used for model selection and dOFV values of 3.84 and 6.64 indicated a significant difference of p < 0.05 and p < 0.01 for one degree of freedom, respectively.

Model evaluation and selection were also based on graphical exploration using goodness-of-fit (GOF) plots, and prediction-corrected visual predictive checks (pcVPCs). Furthermore, the robustness of the parameter estimation from the final model was evaluated by non-parametric bootstrap analysis with 2000 replications.

2.5.5 Dosing Regimen Simulation

To explore which dosing regimen would result in predefined CAT titer targets, Monte Carlo simulations of different dosing regimens were conducted using the final model. The cut-offs were based on prior research by different research groups. Feng et al showed that a titer of 264 BAU mL−1 was associated with 80% vaccine efficacy whereas Goldblatt et al and Dimeglio et al reported 150 BAU mL−1 as sufficient for offering protection [29,30,31]. Based on this, we examined simulations for a cut-off of 100, 150, 250 and 300 BAU mL−1. For the covariate relationships of the final model, values were taken randomly from the study cohort. Dosing regimens were rounded to the nearest practical volume.

2.5.6 Secondary Endpoints

To evaluate the protective effect of ConvP and COVIg, patients were instructed to undergo PCR testing when they would become symptomatic in order to detect potential breakthrough SARS-CoV-2 infections. If possible, the viral strain was sequenced for patients admitted to the hospital. For investigating the safety of ConvP and COVIg, serious adverse events (SAEs) were assessed.