Appropriate Sampling and Longer Follow-Up Are Required to Rigorously Evaluate Longevity of Humoral Memory After Vaccination

Abstract

One of the goals of vaccination is to induce long-lived immunity against the infection and/or disease. Many studies have followed the generation of humoral immunity to SARS-CoV-2 after vaccination; however, such studies typically varied by the duration of the follow-up and the number of time points at which immune response measurements were done. How these parameters (the number of time points and the overall duration of the follow-up) impact estimates of immunity longevity remain largely unknown. Several studies, including one by Arunachalam et al. (2023. J. Clin. Invest. 133: e167955), evaluated the humoral immune response in individuals receiving either a third or fourth dose of mRNA COVID-19 vaccine; by measuring Ab levels at three time points (prior to vaccination and at 1 and 6 mo), Arunachalam et al. found similar half-life times for serum Abs in the two groups and thus suggested that additional boosting is unnecessary to prolong immunity to SARS-CoV-2. I demonstrate that measuring Ab levels at these three time points and only for 6 mo does not allow one to accurately evaluate the long-term half-life of vaccine-induced Abs. By using the data from a cohort of blood donors followed for several years, I show that after revaccination with vaccinia virus, vaccinia virus-specific Abs decay biphasically, and even the late decay rate exceeds the true slow loss rate of humoral memory observed years prior to the boosting. Mathematical models of Ab response kinetics, parameterized using preliminary data, should be used for power analysis to determine the most appropriate timing and duration of sampling to rigorously determine the duration of humoral immunity after vaccination.

FIGURE 1.

Biphasic decline of VV-specific Abs after revaccination. (A) We analyzed the kinetics of Ab titers in four long-term blood donors from subjects 1–4 shown in Fig. 2 of Amanna et al. (8). These individuals were followed up for 20+ years during which they had been revaccinated with VV. (B) We fitted a mathematical model of humoral immune response (Eq. 1) to subsets of the data that include revaccination and estimated the rate of Ab expansion (ρ), the proportion of Ab conversion into a long-lived population (f_m), and the half-life of the humoral immunity (T_1/2, B and see Table I for estimated model parameters). Supplemental Fig. 1C–1F shows dynamics of individual subpopulations of ASCs predicted by the model. (C) Kinetics of Ab response following VV revaccination in one volunteer (donor 97) suggests infinite half-life of the long-term memory [see main text for best fit parameters; these data were published previously (10)]. In (B) and (C), markers denote the data, and lines are predictions of the mathematical model. (D) Sparse measurements of Ab titers after vaccination allow alternative mathematical models (Eq. 1 with different parameter sets) with drastically different predicted longevities of Abs. Here, the markers are average Ab titers from Fig. 1Ciii of Arunachalam et al. (4), and lines are predictions of two alternative mathematical models with different assumed subpopulations of ASCs.

FIGURE 2.

The approach to stable Ab levels takes months to years after revaccination with VV or tetanus vaccine. We fitted several alternative mathematical models for Ab decay (Eq. 2) to data by varying the number of subpopulations for four volunteers revaccinated with VV (A) or tetanus (B) vaccines. In the analysis, we included only the data from the peak Ab response at the first instance after an apparent revaccination (noted by the horizontal dashed line). In both cases, the models with two subpopulations fitted the data better than the model with a single subpopulation, but a larger model (n = 3) did not improve the model fit quality but did provide longer half-life times for persisting Abs. For each fit, we provide the estimated half-life time (T_1/2) based on the slowest decay rate. In both panels, markers denote the data, and lines are predictions of the mathematical model. Parameters of the best fit models are shown in Table II. Note the different scales of the follow-up for boosting with two different vaccines. EU, ELISA units.