Risk of Clade 1 Mpox Outbreaks Among Gay, Bisexual, and Other Men Who Have Sex With Men in the United States

We developed an agent-based model to simulate sexual mpox transmission among MSM. We adapted a previous model that assessed clade II transmission in MSM networks, adding new data on U.S. sexual network structures and exploring transmission parameters to represent clade I mpox.

We produced simulations for 13 counties among 50 jurisdictions in the Ending the HIV Epidemic (EHE) Initiative. The 50 EHE jurisdictions account for more than half of all new HIV diagnoses, and many represent urban areas across the United States with large MSM populations. We chose 13 counties that represent a range of mpox population immunity levels (11%-88%) across the 50 non-state EHE jurisdictions, prioritizing counties that are hubs for international travel and large events. For each of the 13 counties, we established baseline population-level immunity based on vaccination and case reporting to DCIPHER as of February 2024. We assumed that prior infection with clade II mpox provides full protection against both mpox clades, and that vaccination with the JYNNEOS vaccine provides partial but strong immunity (75.2% and 85.9% for one dose and two doses, respectively).

We then generated sexual networks with sizes equivalent to the estimated MSM population for each county. The sexual behavior within the networks was estimated from an online survey of cisgender MSM from across the United States. We assumed that the distribution of type and frequency of sexual behavior is the same across all counties as data does not exist to estimate these parameters at a local level. Finally, we also modeled short-term behavioral adaptations as a reduction in the frequency of spontaneous or one-time sexual encounters, assuming a level of behavioral adaptation in our analyses similar to what occurred in the District of Columbia in 2022 based on previous modeling work and documented across the United States.

The cumulative number of infections and proportion of simulation runs with infections remaining one year after introduction of the virus were compared across three transmission scenarios. Transmission scenarios included the following: 1) baseline, parameterized for clade II (74.5%); 2) clade I, assumed 10% more transmissible than clade II (81.9%); and 3) clade I, assumed 20% more transmissible than clade II (89.4%). For each simulation, we assumed that five MSM with the highest levels of sexual activity (defined as having one or more spontaneous/one-time sexual partners per week in addition to any main or casual partners) were exposed to mpox and ran the simulation for one year. We summarized results across 1,000 simulation runs for each county and scenario combination.

How did this model differ from previous modeling during the clade II outbreak?

The updated model was fit to sexual network data that was collected more recently (2017-2019), representing MSM across the United States rather than a single geographic region. Furthermore, we added data on oral sex partnerships in addition to anal sex partnerships and recalibrated sexual activity group strata to better characterize the range of sexual activity reported in the data. We also calibrated the clade II transmissibility parameter using clade II mpox case data from early in the 2022 outbreak. This parameter has a calibrated distribution of β(4.24, 1.45), with mean equal to 74.5% probability of transmission per contact.

We also added additional transmission parameters to explore possible clade I scenarios. While there are no studies that estimate the exact difference in per-contact transmissibility between the globally circulating clade II virus and clade I in humans, there is evidence that rash intensity and detectable viral loads are greater for clade I relative to clade II in traditional zoonotic and household transmission settings, and a small mammal model demonstrated that virulence of clade I is greater than clade II. We generalized these lines of evidence, assuming a 10% increase relative to clade II (81.9% per-contact transmissibility) and a 20% increase (89.4% per-contact transmissibility).

Lastly, previous work modeled a range of immunity to mpox using a single population size, where in this work we modeled county-specific MSM population size and composition of population-level immunity. This generates more variance in our results, but both approaches come to similar conclusions about the overall level of population-level immunity that is protective against prolonged transmission of mpox.

What data did you use to estimate county-level immunity coverage? 

Population-level immunity was calculated as the total number of immune people in each county divided by the estimated size of the MSM population with increased risk of mpox exposure in that county. Total number of immune people included everyone reported to have received one or two doses of the JYNNEOS vaccine through January 2024, all people with diagnosed mpox through March 2024, and an estimate of the number of undiagnosed mpox infections based on previous modeling work. We estimated the size of the MSM population with increased risk of mpox exposure in each county using county-level estimates from survey data reduced by 40% to reflect the smaller proportion of MSM considered higher activity based on national survey data.

Limitations 

Our analysis is subject to several limitations, including some that could lead to underestimation of outbreak size and probability of prolonged transmission. We assumed that prior infection with clade II mpox provides full protection against both mpox clades, that vaccination with the JYNNEOS vaccine provides partial immunity, but did not account for waning immunity from either previous infection or vaccination in this model. We also assumed that the JYNNEOS vaccine and prior infection with clade II will provide similar protection against clade I as for clade II, all of which could reduce our estimates of outbreak size.

We also made several assumptions that could have led to overestimates of outbreak size and probability of prolonged transmission. We seeded simulations with a high number of infections in highly connected individuals. In addition, we assumed substantial behavioral adaptation in the face of an outbreak. Our sensitivity analysis for clade II indicated that substantially larger outbreaks could occur if this assumption is violated, though outbreak sizes did not exceed 100 cases on average for counties with >50% immunity among MSM with increased risk of mpox exposure (defined in our model as MSM who are likely to form spontaneous or one-time partnerships in addition to having main and/or casual partners). We also assumed that no vaccination occurred during the simulated year. In counties where analysis indicates that large outbreak sizes are possible, we expect—but did not model—that additional health interventions would be implemented that could potentially reduce outbreak sizes.

This analysis only explicitly modeled 13 counties; therefore, it is not representative of the entire U.S. However, many of the counties from the EHE Initiative that we did not include have similar population immunity to counties considered here and would be expected to have similar probabilities of prolonged transmission.

Finally, we acknowledge substantial uncertainty in the size of the MSM population with increased risk of mpox exposure (which affects coverage estimates) and note that sexual networks may not be accurately reflected for all counties, as the survey data could not be disaggregated to the county level. While unaccounted-for geographic variation in sexual behavior could influence the absolute outbreak size in a given county, because previous modeling work used networks with lower levels of sexual activity and had similar outcomes, we expect that the general conclusions presented here would hold.