Download report as PDF
< Coronavirus Disease 2019 (COVID-19) Resources at ISPM
Contact: christian.althaus@ispm.unibe.ch
Three recently detected variants (B.1.1.7, B.1.351, P.1) of SARS-CoV-2 have been detected in various countries worldwide including Switzerland. We aim at tracking the spread of these variants in the cantons of Geneva, Bern, and Zurich, and in Switzerland overall. As a comparison, we also provide an analysis of the spread of B.1.1.7 in Denmark. Finally, we estimate the current proportion of the variants for a given region and their increased transmissibility.
Figure 1. Increase in the proportion of SARS-CoV-2 variants among positive samples in Switzerland and Denmark. Note that the projected trajectories for Zurich and Switzerland are overlapping. Error bars and shaded areas correspond to 95% confidence intervals of the data (blue) and model (red), respectively.
The proportion of the variants has increased rapidly in all regions (Figure 1). Fitting a logistic growth model to the data allows to quantify and project the increase in the proportion of the new variants (see Methods) (Figure 2, Table 1). Note that due to the delay from infection to sample collection, the shown increase and the estimated proportions of the new variants reflect the epidemic situation from around one week ago.
Figure 2. Estimated proportion of the variant (left) and increase in transmissibility (right). The increase in transmissibility is estimated assuming the effective reproduction number of the wild-type \(R_w \approx 1\) and an exponentially (blue) and delta (green) distributed generation time of 5.2 days (Ganyani et al.). Error bars correspond to 95% confidence intervals.
Table 1. Estimated proportion of the variant and increase in transmissibility. The increase in transmissibility is estimated assuming the effective reproduction number of the wild-type \(R_w \approx 1\) and an exponentially (blue) and delta (green) distributed generation time of 5.2 days (Ganyani et al.).
Region | Variant | Proportion at 5 Mar 2021 | Increase in transmissibility |
---|---|---|---|
Geneva | 501Y | 91% (89%-93%) | 38% (34%-43%) - 47% (40%-53%) |
Bern | 501Y | 74% (66%-80%) | 41% (35%-48%) - 51% (42%-62%) |
Zurich | 501Y | 78% (72%-82%) | 40% (36%-45%) - 50% (43%-57%) |
Switzerland | B.1.1.7 | 78% (75%-81%) | 45% (42%-47%) - 56% (53%-60%) |
Denmark | B.1.1.7 | 86% (85%-87%) | 46% (45%-47%) - 58% (56%-60%) |
The estimates for the increase in transmissibility are in good agreement with earlier findings for B.1.1.7 in the UK and other countries (Volz et al., Davies et al., Leung et al.). Taken together, these findings underline the importance of efficient control measures in order to prevent an increase in SARS-CoV-2 incidence in Switzerland.
We use data sets from the following five regions for our analysis (download here):
Competitive growth between variant and non-variant (‘wild-type’) strains of SARS-CoV-2 can be described by the following two ordinary differential equations: \[\begin{align} \frac{dW}{dt} &= \beta S W - \gamma W, \\ \frac{dV}{dt} &= \beta (1 + \tau) S V - \gamma V, \end{align}\] where \(W\) and \(V\) are individuals infected with wild-type and variant, respectively, and \(S\) the population of susceptibles. For the variant, the transmission rate \(\beta\) is increased by the factor \(1 + \tau\) and both strains are cleared at rate \(\gamma\). For \(\tau > 0\), one can easily show that the proportion of the new variant among all infections increases according to logistic growth: \[\begin{align} p(t) &= \frac{V(t)}{W(t) + V(t)} = \frac{1}{1 + \kappa e^{- \rho t}}, \end{align}\] where \(\kappa = W(0)/V0\) and \(\rho = \beta \tau S\), i.e., the difference in the net growth rates between the variant and the wild-type. \(\rho\) can be directly estimated by fitting a logistic growth model (binomial regression) to the proportion of the new variant. The increased transmissibility of the variant is then given by \(\tau = \rho/(\beta S)\). Since the effective reproduction number of the wild-type is \(R_w = \beta S/\gamma\) and the generation time is \(D = 1/\gamma\), we obtain \(\tau = \rho D/R_w\). For a situation where \(R_w \approx 1\), we can simplify to \(\tau \approx \rho D\). This derivation assumes an exponentially distributed generation time. For a delta distributed generation time, one finds that \(\tau \approx e^{\rho D} - 1\).