Characteristics of the Sexual Networks of Men Who Have Sex With Men in Montréal, Toronto, and Vancouver: Insights from Canada's 2022 Mpox Outbreak

Abstract

Background

The 2022–2023 global mpox outbreak disproportionately affected gay, bisexual, and other men who have sex with men (GBM). We investigated differences in GBM's sexual partner distributions across Canada's 3 largest cities and over time, and how they shaped transmission.

Methods

The Engage Cohort Study (2017–2023) recruited GBM via respondent-driven sampling in Montréal, Toronto, and Vancouver (n = 2449). We compared reported sexual partner distributions across cities and periods: before COVID-19 (2017–2019), pandemic (2020–2021), and after lifting of restrictions (2021–2023). We used Bayesian regression and poststratification to model partner distributions. We estimated mpox's basic reproduction number (R₀) using a risk-stratified compartmental model.

Results

Pre–COVID-19 pandemic distributions were comparable: fitted average partners (past 6 months) were 10.4 (95% credible interval: 9.4–11.5) in Montréal, 13.1 (11.3–15.1) in Toronto, and 10.7 (9.5–12.1) in Vancouver. Sexual activity decreased during the pandemic and increased after lifting of restrictions, but remained below prepandemic levels. Based on reported cases, we estimated R₀ of 2.4 to 2.7 and similar cumulative incidences (0.7%–0.9%) across cities.

Conclusions

Similar sexual partner distributions may explain comparable R₀ and cumulative incidence across cities. With potential for further recovery in sexual activity, mpox vaccination and surveillance strategies should be maintained.

A global outbreak of mpox unfolded from May to October 2022, predominantly affecting gay, bisexual, and other men who have sex with men (GBM). The outbreak was unprecedented in its spread through sexual networks, number of cases generated, and geographical distribution, with most of the nearly 90 000 confirmed cases worldwide (May 2022 to June 2023) occurring among GBM in regions with no previous history of reported transmission [1–3]. These unusual transmission patterns of mpox virus were recognized by the World Health Organization (WHO) as a public health emergency of international concern (lasting from July 2022 to May 2023) [4].

In Canada, 98% of reported cases self-identified as men, nearly all of whom reported sex with other men [5]. At least 70% of all reported cases were concentrated in the country's 3 largest cities: Montréal, Toronto, and Vancouver [6–9]. Mpox cases were identified in hospitals and sexual health clinics, with swift responses from community, clinical, and public health partners [7–9]. After the initial exponential growth in May–June 2022, the number of cases declined rapidly. Since mid-November 2022, sporadic new cases have been reported to the Public Health Agency of Canada [5], and questions remain regarding the future risk of mpox reintroductions [10].

An important factor shaping transmission during the 2022–2023 mpox outbreak was the structure of GBM sexual networks [11, 12]. Studies from Europe and North America attributed local outbreaks to densely clustered sexual networks among GBM with a high number of sexual partners [13–16]. Additionally, earlier case investigations revealed close linkages to international travel and sex-on-premises venues [2, 13, 17]. The concept of core group in sexually transmitted infections posits that a small number of individuals with a high number of sexual partners disproportionally contribute to transmission [18]. It recognizes that heterogeneity in sexual partners is crucial to transmission dynamics. In other words, the average number of sexual partners from a chosen member of the sexual network (ie, degree) is not as informative as the distribution of sexual partner numbers (ie, degree distribution) of that network. Mathematical modeling suggested that the basic reproduction number (R₀) of mpox—the expected number of secondary cases arising from an initial infection in an entirely susceptible population—may be significantly greater than 1, as reported among GBM in the United Kingdom [15]. Estimates of R₀ from other modeling studies based on European and Canadian populations ranged from 1.5 to 4.3 [19].

Although these findings have provided insights into the transmission dynamics of mpox, there remains uncertainty regarding how GBM sexual networks in major Canadian cities shaped transmission. The outbreak occurred in the aftermath of the coronavirus disease 2019 (COVID-19) pandemic, which may have affected usual sexual networks of GBM. For instance, the lifting of travel restrictions and other public health measures may have increased the number and types of sexual partnerships formed and facilitated international dissemination of the virus [13, 17, 20, 21].

Given these uncertainties, we leveraged data from the Engage Cohort Study and mpox surveillance data to improve our understanding of the relationship between sexual networks and mpox transmission during 2022–2023 in Montréal, Toronto, and Vancouver. Specifically, we estimated the distribution of sexual partner numbers among GBM in each city and investigated how these distributions changed over time to assess the influence of the COVID-19 pandemic on sexual behaviors. We also assessed the transmission potential of mpox by estimating the R₀ and the cumulative mpox incidence in each city.

METHODS

Study Setting and Population

The Engage Cohort Study (Engage; 2017 to present) is a prospective, population-based cohort study of GBM in Montréal, Toronto, and Vancouver. Eligible participants were self-identified cis or trans men living in 1 of the 3 cities, aged ≥16 years, who reported sex with another man in the past 6 months, understood English or French, and provided written consent. From February 2017 to August 2019, participants were recruited using respondent-driven sampling (RDS), a method used to sample hard-to-reach populations and estimate representative population characteristics [22]. Initial participants (ie, seeds) were purposively selected to represent diverse characteristics of the GBM community, and all participants were invited to recruit up to 6 peers in their social networks. Participants were followed up every 6 months after the baseline visit (12 months in the first 2 years for Montréal and Toronto). At each visit, participants completed an online questionnaire and underwent laboratory testing for sexually transmitted and blood-borne infections. Details of the cohort, including use of RDS sampling and follow-up visits, have been detailed in previous research [23–26].

To investigate changes in sexual behaviors related to the COVID-19 pandemic, we used data from baseline and follow-up visits. The prepandemic period was defined as the participants’ baseline visit (February 2017 to August 2019). The pandemic period was the earliest follow-up visit that occurred between June 2020 and November 2021, covering successive waves of COVID-19 when physical distancing measures were common. The start of this period was chosen because Engage study visits only resumed 3 months after the COVID-19 pandemic was declared by the WHO (11 March 2020) [27]. The postrestrictions period was defined as the latest follow-up visit that occurred between December 2021 and February 2023. To ensure consistency of the time periods across cities, we defined the start of this period based on the easing of entry requirements for nonessential travel into Canada (3 September 2021) [28]. The start of the period was shifted forward by 3 months to account for the 6-month recall period used in the Engage questionnaire. Additionally, we used travel restrictions given the history of international travel reported from initial mpox case investigations. The end of this period was the latest available data cut from the Engage Cohort Study (February 2023).

Variables

Our primary outcome was the self-reported number of sexual partners in the past 6 months, measured through the question “During the past 6 months, with how many guys have you had any kind of sex (anal, oral, mutual masturbation, rimming, frontal/vaginal, etc.)?”. Informed by epidemiological data on mpox cases [2, 13], the following biobehavioral variables were considered potential correlates of the number of sexual partners in the past 6 months in the analyses (details in Supplementary Table 1): age (16–29, 30–39, 40–49, 50–59, ≥60 years); relationship status and sexual arrangement (no relationship, exclusive relationship, open relationship, unclear); human immunodeficiency virus (HIV) status (determined using fourth-generation testing with a confirmatory assay, or self-reported for the 4% of baseline participants with unavailable testing data); visit to bathhouses and/or sex clubs at least once in the past 6 months (binary); attendance of group sex events at least once in the past 6 months (binary); use of dating apps to find partners at least once in the past 6 months (binary); and participation in transactional sex at least once in the past 6 months (ie, receive money and/or goods in exchange for sex; binary). Missing values for the last 4 variables were handled using the missing indicator method for all analyses [29].

Distribution of Sexual Partner Numbers, Rationale for Weighting, and Computation of Weights

To estimate the distribution of sexual partner numbers in each city, we modeled the observed distributions using a Bayesian framework. Briefly, we first fitted a negative binomial regression model to the number of sexual partners in the past 6 months, using the correlates mentioned above as covariates in the model. We then incorporated sampling and attrition weights via poststratification, by using the fitted posterior sample of the number of sexual partners in the past 6 months for each participant. Lastly, we computed the fitted population distribution of sexual partner numbers based on the poststratified samples.

We chose a regression-based approach over direct distribution fitting to examine individual correlates and easily incorporate survey sampling and attrition weights. Regression models were fitted using Hamiltonian Monte Carlo in Stan, with 6000 iterations (2 chains, 3000 burn-in iterations, no thinning, ensuring that the effective sample size for each parameter was ≥1000), using weakly informative priors and assessing convergence via traceplots and the potential scale reduction factor ($\hat{R}$). We then used the posterior distribution of each participant's outcome to estimate the population distribution of sexual partner numbers, incorporating RDS-II weights and inverse probability of censoring weights (IPCW) via poststratification. For the pre–COVID-19 pandemic time period, only RDS-II weights were used. To adjust for attrition in the pandemic and postrestrictions time periods, we used the product of RDS-II weights and IPCW.

Given that Engage is an RDS sample, we used RDS-II weights to ensure the estimated distributions of sexual partner numbers were representative of the target population (sexually active GBM in each city). With an RDS design, participants with larger social networks have a higher chance of being recruited; RDS-II weights adjust for this oversampling by assigning a weight that is inversely proportional to the self-reported network size [22]. To adjust for attrition at follow-up visits, we used IPCWs to reduce potential biases stemming from correlation between the outcome (number of sexual partners in the past 6 months) and being lost to follow-up [30].

Finally, to compare the distribution of sexual partner numbers between a pair of cities or time periods, we computed the proportion of the iterations (posterior distribution samples) that had greater cumulative density for ≥25 and ≥100 partners. These thresholds were based on a previous modeling study from the Netherlands, which projected that GBM with a mean of 25 partners in the past 6 months are expected to constitute about 20% of mpox cases, and those with a mean of about 100 partners ≥60% of cases [31].

Mpox Basic Reproduction Number (R₀)

The basic reproduction number R₀ is a measure of an infectious agent's transmission potential [32, 33]. To estimate R₀ for mpox, we first calibrated a risk-stratified compartmental model of sexual transmission of mpox to surveillance data of reported cases. We then used the calibrated parameters to compute R₀ using the next-generation matrix (NGM) approach [34, 35].

Specifically, we used the sexual partner distributions of GBM from the postrestrictions period to partition the population into 20 sexual activity groups that effectively capture the small number of GBM with high numbers of sexual partners [15]. Then, we developed, parameterized, and calibrated a deterministic susceptible-exposed-infectious-recovered (SEIR) model of mpox transmission among GBM where both the exposed and infectious period follow Erlang-2 distributions. The model was calibrated in a Bayesian framework, using sampling importance resampling, to the daily number of reported mpox cases [5]. We accounted for reporting delays (average of 2 days) and calibrated the moddel to the early phase of the outbreaks in the 3 provinces (ie, before the scale-up of vaccination) because the overwhelming majority of cases at that time were from these 3 cities [6–9]. We assumed that sexual behaviors were constant over that period. We calibrated 5 model parameters: (1) the transmission probability per effective contact (defined as a sexual partnership); (2) the duration of the effective infectious period; (3) the degree of assortativity by risk groups (ie, mixing parameter); (4) the fraction of all incident mpox cases being reported to the surveillance databases; and (5) the number of imported cases at the start of the epidemic. GBM population sizes for each city were informed by previous estimates [36–39]. Model details are in the Supplementary Methods and parameters are presented in Supplementary Table 2.

Using the model-derived city-specific estimates for the transmission probability and mixing parameter, we estimated R₀ for mpox in each city using the models’ NGM. Briefly, the NGM describes the transmission events generated by one group towards another, and the R₀ can be computed as the largest nonzero eigenvalue of this matrix. We also calculated the effective reproduction number (R_e) accounting for immunity in the top 0.2%, 0.4%, 0.6%, 0.8%, 1%, 2%, and 3%, 4%, and 5% of sexual activity groups. We repeated these procedures using the prepandemic distributions, to evaluate potential increases in mpox transmission if sexual activity recovered to pre–COVID-19 levels.

R₀ and Cumulative Incidence Proportion Based on Reported Mpox Cases

To compare our R₀ estimates to those reported elsewhere, we estimated R₀ from the growth rate of reported mpox cases, using the same data source described above [5]. We used the formula ${\mathcal{R}}_0=(1+\mathrm{\Lambda }D)(1+\mathrm{\Lambda }{D}^{\mathrm{\prime }})$, where the latent period $D^{\mathrm{\prime }}$ was 5.1 days and the infectious duration D was the city-specific value calibrated in the SEIR model, for comparability with our NGM estimates [11, 40–43]. The epidemic growth rate $\mathrm{\Lambda }$ was estimated from the slope of the log cumulative cases over time, using the period of initial exponential growth (first 50 days after the first case in each province) [35]. We finally estimated the cumulative incidence proportion of mpox cases among GBM during the 2022–2023 outbreak in each city using GBM population size estimates from previous studies [36–39].

Sensitivity Analyses

We performed 4 sensitivity analyses. First, to verify the robustness of results regarding the distribution of sexual partner numbers to our weighting approach, we repeated the analyses restricting the analytical sample to participants who had visits at all 3 time points. Second, as age is an important determinant of sexual activity and differences in the age distribution of the participants across the 3 cities have been reported [44], we performed regression-based standardization to the distribution observed in Montréal. Third, given uncertainty regarding mpox transmission probabilities through different types of sex acts [2, 45], we repeated the analyses focusing on the number of anal sex partners as the outcome. Lastly, we fitted zero-inflated regression models to test whether participants reporting zero partners during follow-up significantly influenced our model fit.

All analyses were conducted using R (version 4.2.2), Stan (version 2.26.1), and RStan (version 2.21.7) [46–48]. Additional details on methods can be found in the Supplementary Methods. The code used for analyses is available from GitHub (https://github.com/pop-health-mod/mpox-engage-sex-networks).

Ethics

Ethics approval was obtained from the Research Institute of the McGill University Health Centre and the Research Ethics Office of the Faculty of Medicine and Health Sciences, McGill University (A06-M32-23B), Toronto Metropolitan University (Research Ethics Board [REB] No. 2016-113), the University of Toronto (protocol No. 00033527), St Michael's Hospital (REB No. 17-043), the University of Windsor (REB No. 33443), the University of British Columbia (H16-01226), Providence Health Care (H16-01226), the University of Victoria (H16-01226), and Simon Fraser University (H16-01226).

RESULTS

Study Population

There were 2449 GBM recruited to Engage, with 1179 participants in Montréal, 517 in Toronto, and 753 in Vancouver. The effective sample size (accounting for RDS-II weights) was about half in each city: 516 (Montréal), 324 (Toronto), and 350 (Vancouver). Retention over the study period was slightly higher in Montréal, where 70% and 67% of participants had at least 1 visit during the pandemic and postrestrictions period, respectively, compared to 58% and 56% for Toronto, and 60% and 52% for Vancouver (Table 1 and Supplementary Table 3).

Table 1.

Unadjusted and RDS-II Adjusted Baseline Estimates of the Number of Sexual Partners in the Past 6 Months and its Correlates Among the Engage Cohort Study Participants in Montréal, Toronto, and Vancouver, 2017–2019

	Montréal				Toronto				Vancouver				Overall
	n	(%)	RDS-II Weighted, % (95% CI)		n	(%)	RDS-II Weighted, % (95% CI)		n	(%)	RDS-II Weighted, % (95% CI)		n	(%)	RDS-II Weighted, % (95% CI)
Number of participants	1179				517				753				2449
ESS	516				324				350				1190
Age group, y
16–29	384	(33)	36	(34–39)	222	(43)	51	(46–55)	293	(39)	45	(42–49)	899	(37)	42	(40–44)
30–39	340	(29)	27	(24–29)	183	(35)	24	(20–28)	235	(31)	26	(23–29)	758	(31)	26	(24–28)
40–49	166	(14)	15	(13–17)	57	(11)	8	(6–10)	86	(11)	10	(8–12)	309	(13)	12	(10–13)
50–59	181	(15)	12	(10–14)	39	(8)	11	(8–14)	91	(12)	11	(9–13)	311	(13)	12	(10–13)
60+	108	(9)	10	(8–12)	16	(3)	6	(4–8)	48	(6)	8	(6–10)	172	(7)	9	(8–10)
Relationship status and sexual agreement
Single	671	(57)	56	(53–59)	279	(54)	47	(43–52)	413	(55)	57	(54–61)	1363	(56)	55	(53–57)
Open	330	(28)	24	(22–27)	173	(33)	25	(21–29)	221	(29)	21	(18–24)	724	(30)	23	(22–25)
Exclusive	97	(8)	10	(9–12)	48	(9)	21	(17–24)	83	(11)	15	(13–18)	228	(9)	14	(13–15)
Unclear	81	(7)	9	(8–11)	17	(3)	7	(4–9)	36	(5)	7	(5–8)	134	(5)	8	(7–9)
HIV statusa
Seropositive	215	(18)	14	(12–16)	101	(20)	22	(19–26)	132	(18)	20	(18–23)	448	(18)	18	(16–19)
Seronegative/unknown	964	(82)	86	(84–88)	416	(80)	78	(74–81)	621	(82)	80	(77–82)	2001	(82)	82	(81–84)
Bathhouse/sex club attendance in the P6Mb
Yes	452	(38)	31	(29–34)	273	(53)	39	(35–43)	283	(38)	29	(26–32)	1008	(41)	32	(30–34)
No	716	(61)	67	(65–70)	237	(46)	56	(52–61)	464	(62)	70	(67–73)	1417	(58)	66	(64–68)
Missingc	11	(1)	1	(1–2)	7	(1)	5	(3–7)	6	(1)	1	(0–2)	24	(1)	2	(1–2)
Group sex event attendance in the P6Mb
Yes	273	(23)	16	(14–18)	192	(37)	23	(19–26)	208	(28)	21	(18–24)	673	(27)	19	(17–20)
No	900	(76)	83	(81–85)	316	(61)	72	(68–75)	540	(72)	79	(76–81)	1756	(72)	79	(78–81)
Missingc	6	(1)	1	(0–2)	9	(2)	6	(4–8)	5	(1)	1	(0–1)	20	(1)	2	(1–2)
Dating app use to find partners in the P6Mb
Yes	766	(65)	56	(53–59)	403	(78)	60	(56–64)	556	(74)	65	(61–68)	1725	(70)	59	(57–61)
No	413	(35)	44	(41–47)	114	(22)	40	(36–44)	197	(26)	35	(32–39)	724	(30)	41	(39–43)
Transactional sex in the P6Mb
Yes	88	(7)	6	(5–7)	54	(10)	6	(4–7)	42	(6)	6	(4–8)	184	(8)	6	(5–7)
No	1070	(91)	91	(89–93)	454	(88)	93	(91–95)	703	(93)	93	(91–94)	2227	(91)	92	(91–93)
Missingc	21	(2)	3	(2–4)	9	(2)	1	(0–2)	8	(1)	1	(1–2)	38	(2)	2	(2–3)
Number of sexual partners in the P6M, mean	12.4	(SD 22.1)	8.1	(6.9–9.4)	19	(SD 1.6)	8.7	(7.4–9.9)	12.2	(SD 19)	8	(6.8–9.2)	13.7	(SD 23.8)	8.2	(7.4–9.0)
Number of anal sexual partners in the P6M, mean	7.4	(SD 15.3)	4.9	(3.9–5.8)	12.2	(SD 24.3)	5.9	(4.7–7.1)	8.3	(SD 15.4)	5.2	(4.3–6)	8.7	(SD 17.7)	5.2	(4.6–5.8)

RDS-II weights are inversely proportional to participants’ social network size.

Abbreviations: CI, confidence interval; ESS, effective sample size; HIV, human immunodeficiency virus; P6M, past 6 months; RDS, respondent-driven sampling.

^aHIV status was determined based on fourth-generation testing with a confirmatory assay. If the laboratory test result was unknown, self-reported status was used.

^bAt least once in the P6M.

^cMissing includes “prefer not to answer.”

Accounting for RDS-II weights, participants in Montréal were older on average than in Toronto and Vancouver: 37% aged ≥40 years, versus 25% and 29%, respectively. In all 3 cities, approximately half of participants reported not being in a relationship at baseline. There were fewer participants with HIV in Montréal (14%; 95% confidence interval [CI], 12%–16%), compared to 22% (95% CI, 19%–26%) in Toronto and 20% (95% CI, 18%–23%) in Vancouver. More participants reported attending bathhouses and group sex in Toronto (39% for bathhouses; 95% CI, 35%–43%; 23% for group sex; 95% CI, 19%–26%) than in Montréal (31% for bathhouses; 95% CI, 29%–34%; 16% for group sex; 95% CI, 14%–18%) and Vancouver (29% for bathhouses; 95% CI, 26%–32%; 21% for group sex; 95% CI, 18%–24%). Lastly, the RDS-II–weighted mean number of sexual partners in the past 6 months was 8.7 (95% CI, 7.4–9.9) in Toronto, compared to 8.1 (95% CI, 6.9–9.4) in Montréal and 8.0 (95% CI, 6.8–9.2) in Vancouver (Table 1).

Correlates of the Networks’ Number of Sexual Partners

In all 3 cities, a higher number of sexual partners in the postrestrictions period was strongly associated with attendance of group sex events, with a rate ratio of 3.44 (95% credible interval [CrI], 2.66–4.47) in Montréal, 3.62 (95% CrI, 2.59–5.13) in Toronto, and 3.09 (95% CrI, 2.24–4.34) in Vancouver. Other strong correlates were participation in transactional sex, usage of dating apps, and visit to bathhouses and/or sex clubs (Supplementary Table 4).

Differences in the Distribution of Sexual Partner Numbers by City and Time Period

Overall, the fitted distribution of sexual partner numbers was similar across the 3 cities: the mean number of partners was 10.4 (95% CrI, 9.4–11.5) in Montréal, 13.1 (95% CrI, 11.3–15.1) in Toronto, and 10.7 (95% CrI, 9.5–12.1) in Vancouver. However, prepandemic, sexual networks in Toronto had the heaviest-tailed distribution, with 1.4% (95% CrI, 1.0%–1.9%) of GBM reporting ≥100 partners in the past 6 months, compared with 0.6% (95% CrI, 0.4%–0.8%) in Montréal, and 0.3% (95% CrI, 0.2%–0.5%) in Vancouver. All posterior distribution samples showed that Toronto had a larger cumulative density of ≥100 numbers of sexual partners than Montréal and Vancouver. This result held during the postrestrictions period: 0.6% (95% CrI, 0.3%–0.9%) of GBM in Toronto reported ≥100 partners, 0.3% (95% CrI, 0.2%–0.5%) in Montréal, and 0.5% (95% CrI, 0.2%–0.9%) in Vancouver. After restrictions, Toronto had a larger cumulative density of ≥100 numbers compared to Montréal and Vancouver (95% and 69% of the posterior distribution samples, respectively; Figure 1 and Supplementary Figure 1).

Figure 1.

Cumulative distribution of sexual partner numbers in the past 6 months across Montréal, Toronto, and Vancouver at each time period, weighted by RDS-II and inverse probability of censoring weights, with 95% credible intervals. A, Full distribution (1 time period per panel). B, Selected values (1 time period per panel). C, Full distribution (1 city per panel). Abbreviations: P6M, past 6 months; RDS, respondent-driven sampling.

Compared to the prepandemic period, all 3 cities witnessed a marked reduction in the number of sexual partner numbers during the COVID-19 pandemic. In all 3 cities across all samples from the posterior distributions, the cumulative density of ≥25 sexual partners in the past 6 months were consistently larger for the prepandemic versus pandemic period. Sexual activities appeared to have rebounded after lifting travel restrictions: in Montréal and Toronto, 100% of posterior distribution samples (93% in Vancouver) showed a larger proportion of participants reporting ≥25 sexual partners in the past 6 months as compared to the pandemic period. However, sexual activities have not fully recovered to prepandemic levels: in all 3 cities, 100% of the posterior distribution samples had a greater proportion of GBM with ≥25 sexual partners in the prepandemic than the postrestrictions period (Figure 1).

Model Fit, R₀ From the Next-generation Matrix, and Reported Case Counts

The city-specific SEIR models replicated the daily number of reported cases from surveillance data (Figure 2). The calibrated parameter point estimates across cities were 0.80–0.87 for the transmission probability per effective contact, 3.6–4.2 days for the total duration of the effective infectious period, 0.67–0.78 for the degree of assortativity (mixing parameter), 0.78–0.85 for the reporting fraction, and 2–5 for the number of imported cases (Supplementary Table 2).

Figure 2.

Susceptible-exposed-infectious-recovered (SEIR) model fit to observed mpox incidence data for each city. Data up to mpox vaccination scale-up (dotted vertical lines): 15 June 2022 for Montréal, and 10 July 2022 for Toronto and Vancouver. Shaded area shows the 95% credible intervals.

Using the calibrated parameters from the SEIR model and the NGM method, we estimated R₀ of 2.7 (95% CrI, 2.4–3.7), 2.4 (95% CrI, 2.1–3.2), and 2.4 (95% CrI, 2.0–3.1) in Montréal, Toronto, and Vancouver, respectively (Figure 3A ). These are substantially higher than the R₀ estimated from case counts of 2.0 (Montréal) and 1.9 (Toronto, Vancouver) as these can be biased by early saturation of high sexual activity groups (Supplementary Table 5). We also estimated a cumulative incidence proportion of mpox-diagnosed GBM ranging from 0.7% to 0.9% in all cities (Supplementary Table 6 and Supplementary Figure 2).

Figure 3.

Basic (R₀) and effective (R_e) reproduction number of mpox. R_e = R₀ when the proportion immune is 0. A, R_e estimates from the next-generation matrix, assuming that the top x percent of the population is immune and using the susceptible-exposed-infectious-recovered (SEIR)-calibrated infectious duration, probability of transmission per effective contact, mixing parameter, and reporting fraction. B, Projected R_e to the prepandemic period, assuming prepandemic sexual activity and following the same procedure as for (A). Shaded area shows the 95% credible intervals.

According to the NGM estimates, the mpox R_e estimates were highly sensitive to the contact rates in the highest activity groups. The R_e would have been 1.5–1.6 if there was immunity in the 0.2% of the population with the highest contact rates, and R_e would have gone below 1 (start of epidemic decline) with immunity in the 0.8% highest activity groups. If the distributions of sexual partner numbers had been at prepandemic levels in 2022–2023, the R_e would have declined more slowly, especially in Montréal and Toronto, and R_e < 1 would have only been achieved with >1% immunity (>0.6% in Vancouver; Figure 3B ).

Sensitivity Analyses

When restricting the sample to participants with follow-up at each time period only, the distributions of sexual partner numbers were broadly similar for all cities and periods (Supplementary Figure 3). Similarly, standardizing the age distribution in Vancouver to that observed in Montréal did not substantially change the results. However, in Toronto, the tail of the distribution of sexual partner numbers was slightly lighter after standardization, especially for the prepandemic and postrestrictions time periods (Supplementary Figure 4). When using anal sexual partners as outcome, the comparisons across cities and time periods did not qualitatively change, but the distributions had smaller means and lighter tails (Supplementary Figure 5). Lastly, using a zero-inflated model did not change results (Supplementary Figure 6).

Overall, these sensitivity analyses suggest that the results are relatively robust to our weighting methods, assumptions, and to differences in covariate distributions across the cities.

DISCUSSION

In a large population-based cohort of GBM in Canada's 3 largest cities, there was a marked decrease in the distribution of sexual partner numbers during the COVID-19 pandemic compared to the prepandemic period (2017–2019). Despite a small increase after travel restrictions were lifted (late 2021 to early 2023), GBM sexual activity was below prepandemic levels at the time of the 2022–2023 mpox outbreak. Despite the reductions in sexual partnerships, the R₀ of mpox was 2.4–2.7 in all 3 cities during the 2022–2023 mpox outbreak. This high R₀ was driven largely by contact rates of the small proportion of the GBM population with high number of sexual partners and would be substantially lower if members of these groups were not susceptible to infections through natural immunity or vaccination. These findings support prioritization of mpox vaccination to those at highest risk. Additionally, they suggest that the R₀ for mpox may increase if the population's sexual behaviors further recover to prepandemic levels. Continued public health surveillance and preventative activities —community outreach, vaccination— to mitigate the local impacts of mpox reintroductions are advised.

We found that GBM had substantially fewer sexual partners in all 3 cities during the COVID-19 pandemic, and sexual activity remained lower than prepandemic levels even after restrictions were lifted. These findings are in line with previous research from Canada [49, 50] and Europe [20, 51, 52] which suggest that GBM sexual behaviors were influenced by public health measures and messaging related to the COVID-19 pandemic. The implications of our findings are that, as sexual behaviors are expected to return towards prepandemic levels, future mpox transmission remains possible. This is especially true if there is turnover among sexual activity groups. Infection risks are further amplified with case importations in an interconnected world, and limited availability of mpox vaccines and therapeutics in countries in Africa where mpox has been endemic for decades [53].

We found that attendance of group sex events, participation in transactional sex, usage of dating apps, and visits to bathhouse and/or sex clubs were associated with higher numbers of sexual partners in urban Canadian GBM. Notably, group sex events, use of dating apps, and visits to bathhouse and/or sex clubs were also associated with earlier mpox cases during the 2022–2023 outbreak [2, 13]. In the context of relatively low coverage of second-dose mpox vaccination across the 3 cities, and across other cities in Canada, these venues therefore provide a potentially interesting focus for prioritized and tailored vaccine strategies to increase coverage. For example, pop-up vaccine clinics could be set at bathhouses and sex clubs in partnership with community organizations [54–56]. Additionally, focused communication campaigns to promote mpox vaccination could be rolled out on GBM dating apps [7, 57].

Across Montréal, Toronto, and Vancouver, we estimated an R₀ of 2.4–2.7 based on the estimated postrestrictions distribution of sexual partner numbers. This is comparable to the 2.4 observed in Italy and in a pooled analysis of data from European countries [58, 59]. However, the R_e declined substantially (down to 1.5–1.6) with even just 0.2% immunity in the highest activity groups. The absolute size of this group corresponds to roughly 100 GBM in Montréal, 150 in Toronto, and 50 in Vancouver. As it takes relatively few cases in these groups to reduce transmission potential, the R₀ estimated from the growth rates of the epidemic are lower (R₀ of 1.9–2.0). Moreover, despite the high R₀, we estimated a cumulative incidence proportion of only 0.7%–0.9% of GBM by October 2022, implying that the highest activity groups were quickly depleted. Similarly, Murayama et al [16] found that epidemic growth reached its peak at cumulative incidence proportions of 0.2%–0.5% in various North American and European countries. The high mpox R₀ estimates, contrasted with such low cumulative incidence proportions, further highlight the important role of heterogeneous sexual activity and mixing in the 2022–2023 mpox outbreaks. These underscore that R₀ estimates from case data only should be cautiously interpreted in outbreaks where high levels of heterogeneity in contact rates are suspected, as R_e < 1 can be more readily achieved and maintained if individuals at higher risk are preferentially protected by vaccination and/or prior infection [60].

The results should be interpreted considering 4 main limitations. First, our postrestrictions period overlaps with the time when spread of the severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) Omicron variant took place. Thus, the estimated distribution of sexual partner numbers may have been affected by measures introduced in response to the Omicron SARS-CoV-2 wave. However, these restrictions were relatively short lived, and our definition enabled consistent and comparable time periods across cities [61]. Second, although we used IPCW to address attrition bias, this bias may not have been fully adjusted if the loss to follow-up model was misspecified (ie, not all variables associated with attrition were included). Third, we quantified sexual networks using self-reported sexual partner numbers in the past 6 months, which could be subject to social desirability and recall bias. Lastly, our quantification of mpox transmission potential depends on the level of mixing among sexual activity groups and number of imported cases seeded. Because both these parameters were calibrated in our SEIR model, as they are difficult to measure empirically, their uncertainty was propagated to our results. Regarding mixing, the R₀ could be higher if mixing was more like-with-like (assortative) by sexual activity, or lower if it was proportional.

Our approach to estimating the distribution of sexual partner numbers has several strengths. First, we implemented both RDS-II weights and IPCW to obtain estimates representative of sexually active GBM in the 3 largest Canadian cities. Furthermore, intercity comparisons enabled us to account for potential differences in GBM communities in each city and explore their relative impact on the transmission dynamics of mpox. Finally, Engage's longitudinal population-based data collection allowed us to quantify behavioral changes among GBM from before the COVID-19 pandemic up to February 2023.

CONCLUSION

In Montréal, Toronto, and Vancouver, GBM had fewer sexual partners during the COVID-19 pandemic. Even after travel restrictions were lifted in late 2021, sexual activities among urban Canadian GBM had not fully recovered to prepandemic levels. The overall distribution of sexual partner numbers was similar across cities, potentially explaining the similar observed cumulative fraction of mpox cases diagnosed among GBM in the 3 cities. With sexual activity still below prepandemic levels, public health authorities should maintain vigilance. Improving first- and second-dose vaccination coverage among individuals at risk with high numbers of sexual partners should be prioritized.

Supplementary Data

Supplementary materials are available at The Journal of Infectious Diseases online (http://jid.oxfordjournals.org/). Supplementary materials consist of data provided by the author that are published to benefit the reader. The posted materials are not copyedited. The contents of all supplementary data are the sole responsibility of the authors. Questions or messages regarding errors should be addressed to the author.

Notes

Acknowledgments. The authors thank the Engage study participants, office staff, and community engagement committee members, as well as our community partner agency, RÉZO. We also thank Alain Fourmigue for developing the script to compute IPCWs for Engage data.

Author contributions. J. L. F. A., F. X., and M. M.-G. contributed to the conception and design. J. C., D. G., T. A. H., M. D., and S. M. were involved in the design, data collection, and data management of the Engage Cohort Study. Analyses were performed by J. L. F. A. and F. X., with support from M. M.-G. The SEIR model was developed by M. M.-G., and J. L. F. A. and F. X. assisted. J. K., L. W., O. G., H. S., M. I., and S. M. provided input on preliminary methods discussions. The manuscript was drafted by F. X. and J. L. F. A. All authors contributed to the interpretation of results and reviewed the manuscript for important intellectual content. Overall supervision for this project was provided by M. M.-G. All authors approved the final manuscript.

Software and source code. Analysis code is available on a GitHub repository (https://github.com/pop-health-mod/mpox-engage-sex-networks).

Financial support. This study was supported by the Natural Sciences and Engineering Research Council of Canada's Canadian Network for Modeling Infectious Diseases (to S. M., M. I., and M. M.-G.); the Canadian Institutes of Health Research (CIHR; to M. I., S. M., H. S., and M. M.-G.). M. M.-G.'s research program is funded by a Canada Research Chair (Tier 2) in Population Health Modeling. F. X. acknowledges a studentship from the McGill University Center for Viral Diseases. T. A. H. received support from a Chair in Gay and Bisexual Men’s Health from the Ontario HIV Treatment Network. S. M.'s research program is supported by a Canada Research Chair (Tier 2) in Mathematical Modeling and Program Science.

Engage/Momentum II is supported by the CIHR (grant number TE2-138299); the CIHR Canadian HIV/AIDS Trials Network (grant number CTN300); the Canadian Foundation for AIDS Research (grant Engage); the Ontario HIV Treatment Network (grant number 1051); the Public Health Agency of Canada (grant number 4500370314); Toronto Metropolitan University; Canadian Blood Services (grant number MSM2017LP-OD); and the Ministère de la Santé et des Services Sociaux du Québec.

Supplement sponsorship. This article appears as part of the supplement “Mpox: Challenges and Opportunities Following the Global 2022 Outbreak,” sponsored by the Centers for Disease Control and Prevention (Atlanta, GA).