The real seroprevalence of SARS-CoV-2 in France and its consequences for virus dynamics.

Chloé Dimeglio, Jean–Michel Loubes, Marcel Miédougé, Fabrice Hérin, Jean‐Marc Soulat, Jacques IzopetResearch 2020 · DOI

Abstract

The SARS-CoV-2 virus has spread world-wide since December 2019, affecting millions of people and killing thousands. The recent use of serological tests to measure the effective seroprevalence of SARS-CoV-2 in the general population has led us to assess the consequences of a much lower than expected seroprevalence on the real dynamics of the virus in the population. Our prediction models assumed a 4.2% seroprevalence in Occitania and it has turned out to be only 1.3% (CI 95%: 0.6% - 2.6%, binomial method). This implies a lower than expected virus transmission rate, which may be due to infectivity varying with the patient’s symptoms or to a constraint due to an uneven population geographical distribution.

Introduction

During the initial wave of the COVID-19 pandemic, mathematical modeling served as a primary tool for public health policy, predicting virus spread and the necessary stringency of lockdown measures. However, these models often relied on early estimates of the basic reproduction number ($R_0$) and assumptions about the infectiousness of asymptomatic carriers. In France, specifically the Occitania region, a significant discrepancy emerged between these theoretical predictions and clinical observations. This paper investigates the gap between predicted and actual seroprevalence to refine our understanding of SARS-CoV-2 dynamics.

Methodology

The study utilized a dual approach combining empirical serological data with a discretized Susceptible-Infectious-Recovered (SIR) model.

Serological Analysis

Plasma samples from 613 individuals in Occitania were analyzed using an ELISA kit targeting the spike protein receptor-binding domain (RBD). This method provided a high-fidelity snapshot of the population's antibody status, with the cohort designed to represent the general population's exposure levels.

Mathematical Modeling

The researchers developed a discretized SIR model that moved beyond standard continuous assumptions. This model incorporated undetected contagious carriers ($P_n$), detected carriers ($Q_n$), and a variable contagion coefficient. It also integrated parameters for lockdown ($c$) and quarantine ($q$) effectiveness. The core of the analysis involved comparing the real-world seroprevalence ($I_{N,R}$) against the model's estimates ($I_{N,E}$), as shown in the following inequality:

Comparison of real-world (R) vs. estimated (E) seroprevalence mathematical framework.

Results

The most significant finding was the stark overestimation of virus spread by existing models. While the statistical model predicted a seroprevalence of 4.2% for Occitania (and up to 17.5% nationally), the actual measured seroprevalence was only 1.3% (95% CI: 0.6% – 2.6%).

This discrepancy suggests that the effective transmission rate was considerably lower than the initial WHO-estimated average of 2.2. The mathematical refinement of the model, accounting for variable contagion parameters and population distribution, is represented in the expanded inequality below:

Refined SIR model inequality accounting for variable contagion parameters and population dynamics.

Discussion

The authors identify two primary factors contributing to this "overestimation gap":

  1. Asymptomatic Infectivity: The assumption that asymptomatic individuals—comprising roughly half the infected population—are as contagious as symptomatic patients appears incorrect. Lower transmission efficiency in asymptomatic cases would significantly reduce overall seroprevalence.
  2. Geographical Clustering: Standard models assume "perfect mixing," where any individual can infect any other. In reality, the virus spreads in clusters with limited inter-group communication. This uneven population density acts as a natural brake on virus dynamics.

The study dismisses the idea that the lockdown was simply "more effective" than modeled, as the models already assumed a high 80% effectiveness rate, which aligned with observed societal behavior.

Conclusion

The low real-world seroprevalence indicates that the French population remained highly vulnerable following the first lockdown. The findings suggest that relying on the slow accumulation of natural herd immunity is an insufficient strategy. Instead, the authors advocate for aggressive testing, contact tracing, and isolation to manage the virus's true, albeit slower-than-modeled, dynamics.

Summary updated

Abstract

Abstract The SARS-CoV-2 virus has spread world-wide since December 2019, affecting millions of people and killing thousands. The recent use of serological tests to measure the effective seroprevalence of SARS-CoV-2 in the general population has led us to assess the consequences of a much lower than expected seroprevalence on the real dynamics of the virus in the population. Our prediction models assumed a 4.2% seroprevalence in Occitania and it has turned out to be only 1.3% (CI 95%: 0.6% - 2.6%, binomial method). This implies a lower than expected virus transmission rate , which may be due to infectivity varying with the patient’s symptoms or to a constraint due to an uneven population geographical distribution.
BibTeX
@inproceedings{dimeglio2020real,
  author = {Chloé Dimeglio and Jean–Michel Loubes and Marcel Miédougé and Fabrice Hérin and Jean‐Marc Soulat and Jacques Izopet},
  title = {{The real seroprevalence of SARS-CoV-2 in France and its consequences for virus dynamics.}},
  booktitle = {Research},
  year = {2020},
  doi = {10.21203/rs.3.rs-40015/v1},
}
Figure