Characteristics of the study populations
Of 6,108 participants in the ANRS0001S COV-POPART cohort study, 3,693 were included, 2,570 in the specific populations, and 1,123 in the control group (Fig. 1). Most of the participants in the specific populations ended their primary vaccine course between 1st May and 31st June 2021 (71.8%) and between 1st June and 31st July 2021 (71.3%) for the control group. Participants in the specific populations were older (52.5 years interquartile range (IQR): [41.9; 61.0] vs 47.4 [37.0; 57.8]) compared to the control group. Mainly prescribed primary vaccine course was two doses of BNT162b2 (81.8% and 84.4%, respectively). Of participants in specific populations, 33.3% were with HIV, 29.1% were obese non-diabetic, and 16.6% were diabetic, 7.5% received 3 doses in the primary vaccine course (Table 1).
Description of antibody responses
One month after the last dose received during the primary vaccine course the geometric means of anti-spike IgG, anti-RBD IgG, and neutralizing antibodies were 836.8 [789.7; 886.7] BAU/mL, 354.2 [330.8; 379.2] IU/mL and 139.4 [130.9; 148.5], respectively in specific populations. In the control group, geometric mean concentrations were higher 1415.7 [1347.4; 1487.4] BAU/mL, 585.2 [555.8; 616.2] IU/mL and 317.1 [297.5; 337.9], respectively 22 (Supplementary Table 1 and Supplementary Fig. 1).
Description of SARS-CoV-2 infections
Over a median follow-up of 5.0 months [4.4; 5.8], 257 SARS-CoV-2 infections were observed, 143 (105 identified by antigenic test or PCR, 32 positive anti-NCP, and 6 increase in anti-Spike IgG antibody levels) in specific populations and 114 (99 identified by antigenic test or PCR, 13 positive anti-NCP, and 2 increase in anti-Spike IgG antibody levels) in the control group. During follow-up, 24 deaths were observed, all in participants from the specific populations. The majority of SARS-CoV-2 infections occurred in December 2021 (18.9% and 21.9%) or in January 2022 (39.2% and 50.0%), respectively in specific populations and in the control group. The cumulative probabilities at five months were 6.0% specific populations (95% confidence interval (CI): [5.0; 7.0]) and 10.1% in the control group (CI: 8.3; 11.9) (Supplementary table 2 and Supplementary Fig. 2).
Anti-SARS-CoV-2 antibody measurements and SARS-CoV-2 infection
The risk of SARS-CoV-2 infection was similar according to terciles of antibody markers in specific populations and from the control group (Fig. 2).
In the control group, after adjustment, higher levels of anti-Spike IgG antibody were associated with a lower risk of SARS-CoV-2 infection. The decrease in risk was more pronounced for lower values, but the risk continued to decrease for higher values of anti-Spike IgG (Fig. 3). Participants with a value of 264 BAU/mL of anti-Spike IgG antibodies had a lower risk of SARS-CoV-2 infection (HR: 0.4 [0.0; 0.9]) compared to a participant with a value of 15 BAU/mL. Risk of SARS-CoV-2 infection was not differentiated by neutralizing and RBD IgG antibodies (Fig. 3).
When SARS-CoV-2 infection was defined using only antigenic test and PCR, similar results were estimated with a decreased risk for participants with higher anti-Spike IgG value in the control group, and no association for other markers in both populations (Supplementary Table 3).
Predictive performances
In specific populations, the predictive performances of the three antibodies were moderate, with better performances for anti-Spike IgG (AUROC: 73.2%), then anti-RBD IgG (70.9%), and neutralizing antibodies (68.5%). Adding variables to the model, slightly increased the predictive performances (74.5%, 74.9, and 72.4%, respectively for anti-Spike IgG, RBD IgG, and neutralizing antibodies), which were similar and superior to the one of a model without antibodies (67.0%). In the control group, the performances of the antibody were superior both in univariable (78.3%, 77.0%, and 78.8%, respectively for anti-Spike IgG, RBD IgG, and neutralizing antibodies) and in multivariable (82.0%, 81.2%, and 81.4%, respectively for anti-Spike IgG, RBD IgG, and neutralizing antibodies). However, the gain induced by the addition of the antibodies to the model was low (81.2%) (Table 2).
When defining SARS-CoV-2 infection based on antigenic test and RT-PCR only, the predictive abilities of the antibody values decreased and were similar in both populations (Anti-Spike IgG: 71.0% and 71.1%, anti-RBD IgG: 70.1% and 67.8%, Neutralizing antibodies: 66.7% and 68.8%, respectively in specific population and the control group) (Supplementary table 5).
In specific populations who received two doses during the primary vaccine course, diabetic or obese, and other specific populations than diabetic, obese, and persons with HIV, the results were similar to the overall specific population. Among participants with HIV, the predictive performances were low both in univariable and multivariable analysis (Supplementary Table 6).
Factors associated with the risk of SARS-CoV-2 infections
After adjustment on IgG anti-Spike value, older age (HR: 0.8 [0.7; 0.9] and 0.7 [0.6; 0.8] for an increase of 10 years, respectively in specific populations and the control group) and being a woman (0.7 [0.5; 1.0] and 0.7 [0.5; 1.0], respectively) were significantly associated with a decreased risk of SARS-CoV-2 infection, in both specific populations and the control group. In specific populations, higher BMI (1.1 [1.0; 1.3] for an increase of 5 kg/m2) was associated with an increased risk of SARS-CoV-2 infection (Supplementary Table 3). The same factors were identified when modifying the definition of SARS-CoV-2 infection (Supplementary Table 4).