Underdetected dispersal and extensive local transmission drove the 2022 mpox epidemic

Abstract

The World Health Organization declared mpox a public health emergency of international concern in July 2022. To investigate global mpox transmission and population-level changes associated with controlling spread, we built phylogeographic and phylodynamic models to analyze MPXV genomes from five global regions together with air traffic and epidemiological data. Our models reveal community transmission prior to detection, changes in case reporting throughout the epidemic, and a large degree of transmission heterogeneity. We find that viral introductions played a limited role in prolonging spread after initial dissemination, suggesting that travel bans would have had only a minor impact. We find that mpox transmission in North America began declining before more than 10% of high-risk individuals in the USA had vaccine-induced immunity. Our findings highlight the importance of broader routine specimen screening surveillance for emerging infectious diseases and of joint integration of genomic and epidemiological information for early outbreak control.

Keywords: global health; interventions; mpox; phylodynamics; phylogenetics; public health; vaccination campaigns; viral epidemiology; viral evolution.

Figure 1.. Case counts and publicly available sequences by geographic region.

(A, B) Confirmed positive weekly mpox cases by country (A) and global region (B) smoothed using a 7-day rolling average on daily data and then aggregating into weekly counts. Only countries with greater than 5 sequences on GenBank were included. (C, D) Monthly count of publicly-available MPXV genomes found on GenBank by country (C) and global region (D).

Figure 2.. Phylogeographical estimates of MPXV spread in 4 global regions.

(A) The maximum clade credibility tree summary of the Bayesian inference conducted using asymmetric discrete trait analysis and Skygrid prior on 1004 sequences. Colors correspond to the regions in the legend. Ancestral nodes with greater than 50% posterior support are highlighted with a white circle overlaid. Inset histogram on bottom left corner shows 95% interval for the time to most recent common ancestor (TMRCA)(B-D) Estimated number of introductions (B), exports (C), and average time of local persistence in days (D) for each global region. Horizontal black line denotes median estimates.

Figure 3.. Phylodynamic investigation reveals underdetection of mpox.

(A) Regional specific introductions and the resulting outbreak clusters extracted from the MCC tree summary of the Bayesian inference conducted using MASCOT-GLM on 587 sequences. Colors correspond to the regions in the legend. Ancestral nodes with greater than 50% posterior support are highlighted with a white circle overlaid. (B) Estimates of effective population sizes (Neτ in years) from April 2022 through December 2024. The coalescent time scale depends on both effective population size Ne (number of effective individuals) and on generation time τ (years per generation), resulting in Neτ being a measure of coalescent time scale in years . (C) Regional prevalence (in number of cases, interpreted as census population size N) estimated independently using publicly-available case counts, and (D) Estimates of model predictor coefficients for Ne estimation. All of the predictors displayed on the x-axis were included in the analytic model. Dark line represents median estimates, light bands represent 95% HPD.

Figure 4.. Phylodynamic estimates of MPXV transmission dynamics in 5 global regions.

(A) Relationship between estimated date of introduction and persistence time. Each circle represents a single viral introduction with greater than 50% posterior support into the region denoted by the color (i.e. a green point represents an introduction into Western Europe). The size of each point is proportional to the size of the outbreak cluster resulting from each introduction with larger circles representing more resulting downstream tips. Blue dashed line represents the linear best fit line using Pearson’s correlation. Blue shaded region denotes the variability of the line and the resulting estimates from Pearson’s correlation are shown in text above the shaded region. (B) Estimates of model predictor coefficients for migration rate estimation. Error bars denote 95% HPD interval for the magnitude of predictor coefficient (C) Percentages of new cases due to introductions were estimated as the relative contribution of introductions to the overall number of infections in the region. The inner area denotes the 50% HPD interval and the outer area denotes the 95% HPD interval. Estimates were smoothed using a 14 day rolling average.

Figure 5:. Estimates of time-varying reproductive number (Rt) in five global regions.

Estimates of Rt from April 2022 through December 2023 via MASCOT-GLM using 587 sequences subsampled equally throughout time. The inner area denotes the 50% HPD interval and the outer area denotes the 95% HPD interval. Dashed line highlights an Rt value of 1 above which denotes an exponentially growing viral epidemic. Estimates were smoothed using a 14 day rolling average.

Figure 6.. North American MPXV local transmission dynamics.

(A) North American Rt estimated via phylodynamics (solid bands). Dashed orange line indicates the cumulative percentage of high-risk individuals with vaccine-induced immunity in the US. (B) Scatter plot comparing mean Rt calculated via MASCOT-GLM for North America vs cumulative percentage of high-risk individuals with vaccine-induced immunity in the United States. Vaccine-induced immunity was estimated via a two-week lag since the date of vaccination. Red line indicates the best fit spline for scattered points. Dashed gray line indicates expected linear decrease in Rt with increasing vaccine-immunity assuming SIR dynamics. Over each point are the dates that correspond to the mean Rt and percent of immunity at that moment.

Figure 7.. Transmission heterogeneity estimates obtained from clusters of identical mpox sequences.

(A) Mean size of clusters of identical sequences for different geographical regions by month of first cluster detection. (B) Probability to observe a cluster of size 118 among 2624 clusters as a function of the reproduction number R and the dispersion parameter k assuming 5.5% of infections are sequenced. Estimates of (C) the reproduction number R by geographical unit and (D) the dispersion parameter k across geographical units from August 2022 exploring different assumptions regarding the proportion of infections detected. In B, the point corresponds to estimates obtained by Blumberg and Lloyd-Smith (43) from the analysis of epidemiological clusters during previous outbreaks. The segments correspond to the associated 95% confidence intervals. In C-D, points correspond to maximum likelihood estimates and vertical segments to 95% likelihood profile confidence intervals. The horizontal dotted line and the shaded area correspond to estimates obtained by Blumberg and Lloyd-Smith (43) from the analysis of epidemiological clusters during previous outbreaks. In B, the dotted white lines correspond to contour lines for probabilities of 10⁻⁴, 10⁻² and 10⁻¹.