Spatial Epidemiology of the Syphilis Epidemic in Toronto, Canada

Abstract

Background 

Urban centers across Canada and the United States have battled syphilis epidemics with high rates of human immunodeficiency virus (HIV) coinfection for over a decade. We examined the spatial epidemiology of syphilis over time for Toronto (Canada) with the intention of forming new insights and strategies for restoring low syphilis rates.

Methods 

Syphilis incidence rates, HIV-syphilis coinfection, and sexual risk behavior prevalences were estimated and mapped from primary, secondary, early latent syphilis cases reported to Toronto Public Health between January 1, 2006, and December 31, 2010, using ArcGIS 9.0. Geographic clusters of significantly elevated syphilis incidence rates were identified using SaTScan 9.0. The relationship between syphilis incidence rates and sociocultural factors was modeled using the Besag, York, and Mollie model.

Results 

Between 2006 and 2010, syphilis incidence rates were high in Toronto’s downtown core area, intensified, and spread outward initiating 3 independent outbreak areas. HIV coinfection was high (47%); however, no spatial clustering was identified. Syphilis incidence rates, HIV coinfection, and behavioral risk factors promoting sexually transmitted infection transmission were high outside the core area, suggesting that peripheral sexual networks may be influencing high syphilis infection rates both inside and outside the core.

Conclusions 

Toronto’s syphilis epidemic is mature. Response, resources, and intervention activities should target core and noncore areas.

In Brief

Spatial dynamics of mature syphilis epidemics in Toronto and elsewhere suggest that control measures should target core and noncore areas.

Syphilis outbreaks have been erupting across Canada and the United States since the mid-1990s, starting with outbreaks among heterosexuals and crack cocaine users and shifting to men who have sex with men (MSM) by the early 2000s.^1–5 In Toronto, Ontario, Canada (total population: 2,501,502; male population: 1,204,420), the incidence rate of infectious syphilis (primary, secondary, and early latent) was on a dramatic decline during the 1990s, dropping to a low of 0.8 cases per 100,000 population in 1998 and remaining stable and low (1.1 cases per 100,000 population) through 2001.⁶ However, in late 2002, an outbreak of infectious syphilis emerged (7.6 cases per 100,000 population), escalated (first epidemic peak in 2004: 12.5 new cases per 100,000 population; second peak in 2009: 21 new cases per 100,000),⁶ and is ongoing (2012 incidence rate: 19 cases per 100,000 population; 39 cases per 100,000 men specifically).⁷

Similar to other cities, Toronto’s syphilis epidemic is predominantly among MSM, with 98% of reported cases of infectious syphilis being male and 91% of male cases reporting same sex partners in 2012.^7,8 The 3 most prevalent risk factors among cases included unprotected sex (52%), high rates of partner change (47%), and coinfection with an existing sexually transmitted infection (STI; 15%).⁷ Also similar to other jurisdications,^1,3,9 there has been an increasing concern that human immunodeficiency virus (HIV) transmission might accelerate through HIV-syphilis coinfection, which is high among syphilis cases, with HIV coinfection rates ranging from a low of 30% in 2005 to a high of 57% in 2007. In 2012, 45% of new infectious syphilis cases were coinfected with HIV.⁷ This estimate of coinfection is likely an underestimate of the true rate of coinfection because it does not include anonymous HIV cases, and Toronto has particularly high rates of anonymous HIV testing.

Sexually transmitted infection rates concentrate geographically, and observed spatial patterns have been associated with social and neighborhood environment factors.^10–14 Knowing where an epidemic is concentrated and what behavioral, social, and environmental factors are associated with that epidemic can inform public health response to reduce spread with the intention of ending an epidemic.^12,15–17 The purpose of this study was to gain insights into Toronto’s syphilis epidemic by examining the spatial pattern and spread of syphilis over time, including any spatial relationship with HIV coinfection, sexual risk behaviors, and social or neighborhood factors. The intention was to use the findings to inform a new strategy for responding to the syphilis epidemic.

MATERIALS AND METHODS

Sexually Transmitted Infection Guidelines for Canada

Chlamydia, gonorrhea, chancroid, acute hepatitis B, HIV/AIDS, and syphilis are reportable STIs in Canada.¹⁸ Health care providers conduct a brief interview with patients about their medical history including a brief risk assessment. A focused risk assessment is conducted with any patient whose brief risk assessment signals an STI risk factor. The purposes of the focused risk assessment are to evaluate patient STI risk factors and sexual behaviors and to guide sexual health counseling and STI testing recommendations. Topics covered by the focused risk assessment include sexual relationships, sexual risk behaviors, STI history, reproductive history, substance use, and psychosocial history. Data collected during these interviews are used to complete communicable disease case report cards.

Ontario’s Health Protection and Promotion Act mandates the notification of all confirmed or suspect reportable diseases in Ontario to the Medical Officer of Health for the jurisdiction where the patient resides. Reports rely on a passive surveillance system, and the following have a duty to report and promptly inform Public Health: laboratories (through faxed reports), physicians (telephone call, faxes, etc), other health care providers, and institution administrators.

Syphilis cases are interviewed for partner notifications, and risk factors, including limited demographic information (age, sex), diagnosis data (date, disease, coinfections). A predetermined list of sexual behaviors, are collected via the interviews and entered into and managed by Ontario’s integrated Public Health Information System (iPHIS). Partner notification may be done by the patient, health care provider, or public health authority. Sexual contacts are notified about their exposure to a case by any one, or combination, of self-referral or patient referral, health care provider/public health referral, or contract referral. Laboratory-confirmed HIV cases are also reported to public health and entered into the same provincial reporting database (iPHIS) as all other reportable communicable diseases.

Data

Infectious syphilis incidence rates were estimated from confirmed new cases of primary, secondary, and early latent reported to Toronto Public Health (TPH) from January 1, 2006, to December 31, 2010. Syphilis cases were geolocated using residential address provided at the time of testing and aggregated to the census tract level. Census tract-level syphilis incidence rates were estimated by dividing the total number of primary, secondary, and early latent syphilis cases by the estimated population for a given year. Intercensal population estimates were calculated by linear interpolation and extrapolation assuming a constant age structure based on the 2006 Statistics Canada Census and do not take into account births, deaths, or migration.

The syphilis epidemic overwhelmingly involves the male population, specifically, MSM. Therefore, we limited further analysis to the male population in Toronto. Consequently, age was the only variable remaining for demographic characterization of the epidemic. Age-group estimates were constrained to total male population estimates by year to adjust for the “random rounding” in the Canadian Census.

HIV-syphilis coinfection, or the prevalence of HIV among syphilis cases, was calculated as the total number of new cases of syphilis infected with HIV divided by the total number of new syphilis cases in a given census tract from 2006 to 2010. HIV coinfection was determined in collaboration with Public Health Ontario, a provincial government agency and steward of HIV data for Ontario.

Sexual risk behavior data available through iPHIS were collected inconsistently, meaning there was no standardized way that doctors or other health care providers were instructed to ask sexual risk behavior questions and record responses. The sparseness of data for some sexual risk behaviors suggests that not all risk factors were asked about equally, especially because the response option was “check all that apply.” Consequently, we limited our analysis to behaviors with 5% prevalence or higher, assuming that these sexual risk behavior options were presented most consistently. Sexual risk behavior prevalences were estimated for each census tract by dividing the total number of syphilis cases with a specific risk behavior by the total number of syphilis cases reported.

We hypothesized that social determinants of health might be influencing the spatial distribution of syphilis for Toronto (Appendix I). Census tract-level social determinants were represented using social deprivation measures from the Ontario Marginalization Index (ON-Marg; http://www.torontohealthprofiles.ca/onmarg.php), an Ontario specific version of the Canada Marginalization Index developed to measure marginalization in rural and urban Ontario, both of which use data from the Census of Canada.¹⁹ Four social deprivation dimensions of ON-Marg were mapped at the census tract level to represent (Appendix II) the following: residential instability, material deprivation, dependency, and ethnic concentration.²⁰ The residential instability dimension comprised information on housing, living arrangements, dwelling ownership, and mobility; material deprivation comprised information on education, lone parenting, unemployment, and low-income measures; dependency comprised labor force participation and indicators for the population aged 65-plus years (seniors); and ethnic concentration comprised the proportion of recent immigrants and proportion of people identifying as visible minorities.¹⁹

Mapping

Annual and 5-year maps (2006–2010) were generated at the census tract level for syphilis incidence rates, HIV-syphilis coinfection, and sexual risk behaviors using ArcGIS 9.0 (ESRI, Redlands, CA). Statistics Canada defines census tracts as “...small, relatively stable geographic areas that usually have a population between 2,500 and 8,000 persons. They are located in census metropolitan areas and in census agglomerations that had a core population of 50,000 or more in the previous census” (http://www.statcan.gc.ca/pub/92-195-x/2011001/geo/ct-sr/def-eng.htm).

Cluster Detection

Geographic clusters of significantly elevated syphilis incidence rates were identified using SaTScan 9.0²¹ (www.satscan.org). SaTScan uses scan statistics to “detect and evaluate clusters of cases in either a purely temporal, purely spatial, or space-time setting”.²² Syphilis incidence rates are dependent on transmission in both space and time; therefore, we ran a retrospective space-time discrete Poisson model to scan for areas with high incidence rates for our primary analysis. We also ran a purely spatial sensitivity analysis for each quarter of data to see if the location of clusters changed if temporal dependence was not assumed.

SaTScan identifies statistically significant clusters by “gradually scanning a window across time and/or space, noting the number of observed and expected observations inside the window at each location. The window with the maximum likelihood is the most likely cluster, [or], the cluster least likely to be due to chance.”²² A cylindrical scanning window was built from the census tract centroid, where the circular base of the cylinder was set to the maximum spatial cluster size (5% of the population at risk), and the height of the cylinder was set to the maximum temporal cluster duration (90% of study period with no temporal or spatial adjustments). Syphilis incidence rates were estimated by quarter and assigned to the population-weighted center of each tract to reflect the location of the population at risk. Clusters were assumed to be cylindrical (existing in space and time) with a circular base rather than an elliptical base because the census tract boundaries in Toronto are not elongated along features such as rivers, where an elliptical base may be more appropriate.²³ Expected and observed case counts inside the cylinder were compared with the expected and observed case counts outside the cylinder using a likelihood ratio test. Monte Carlo simulations (999 replications) were run to determine the significance of each cluster.

Statistically significant spatiotemporal clusters of high incidence rates were classified using 2 mutually exclusive categories: core areas for clusters persisting for 5 years or longer, and outbreak areas for clusters persisting for less than 5 years.^2,4,24 Areas with no core or outbreak cluster were classified into a third category: noncore-nonoutbreak areas. In a second analysis, clusters were classified in a binary: core area or noncore area (any areas outside the core including outbreak areas), to both simplify analysis for sustainability and simulate the type of analysis that might occur during routine surveillance.

Epidemic, Incidence Rate, and Prevalence Curves

Epidemic,^4,10 incidence rate, and prevalence curves were constructed for syphilis over time and stratified by demographic and behavioral characteristics of cases for the different cluster classifications (i.e., core, outbreak, noncore-nonoutbreak, and noncore). The spline function in R was used to smooth epidemic curves between time points. Epidemic and prevalence curves were also constructed for HIV-syphilis coinfection. Core, outbreak, noncore-nonoutbreak, and noncore epidemic curves were qualitatively interpreted for each demographic and sexual risk behavior by visually comparing peaks and troughs, amplitude, and noise, to understand changes in the absolute number of syphilis cases taking sexual risks from 2006 to 2010.

Social Determinants of Syphilis Rates

The spatial association between social determinants of health, represented by social deprivation measures—residential instability, material deprivation, dependency, and ethnic concentration—and syphilis incidence rates were quantitatively modeled using Bayesian inference and the Besag, York, and Mollie (BYM) model.^25,26 The BYM model separates area-specific random effects into a correlated component and an uncorrelated component, where the unobserved spatial effect between regions is the correlated component. The unadjusted incidence rate ratio model can be represented mathematically as:

where the observed count (O_i) approximately follows a Poisson distribution with an area-specific expected count (E_i) and incidence rate ratio (r_i). The log-linear area-specific incidence rate ratio (log(r_i)) is modeled with a log-linear expected count (log(E_i)) offset, mean incidence rate ratio for Toronto (α), the correlated component or spatial effect (V_i), and the uncorrelated component or independent random effect (U_i). Here the correlated effect is assumed to follow a conditional autoregressive distribution, where regions are dependent on the values of the neighboring regions.²⁷ The uncorrelated effect is assumed to follow a normal distribution. WinBUGS²⁸ was used for the Markov chain Monte Carlo sampling within the Markov chain. A total of 150,000 iterations were run, with a burn-in of 100,000 iterations and an additional 50,000 simulations for the estimation of posterior distributions. The smoothed incidence rate ratio, or standardized incidence rate ratio, was mapped for Toronto.

Social determinants of health—residential instability, material deprivation, dependency, and ethnic concentration—were included in the BYM model to quantify their association with age-standardized syphilis incidence rates and syphilis-HIV coinfection based on a priori hypotheses of associations (Appendix I).²⁶ Model residuals, namely, the correlated and uncorrelated components (U_i + V_i), were mapped and compared for the adjusted model.

Ethics

The University of Toronto Research Ethics Board reviewed and approved this project.

RESULTS

Spatiotemporal Evolution of Syphilis

Maps of syphilis rates for Toronto show one downtown core cluster and 3 outbreak clusters between 2006 and 2010, as detected by SaTScan (Fig. 1). The relative syphilis rate is 12.2 times greater in the core area than outside the core area, and 2 to 3 times greater inside the outbreak areas compared with outside the outbreak areas (Fig. 1). These results suggest that syphilis rates were high in the original downtown core area (total population: 88,332; male population: 47,035), intensified within the downtown core area, then spread outward initiating 3 outbreaks, one to the north (total population: 28,628; male population: 12,690), one to the west (total population: 104,375; male population: 52,000), and one to the east (total population: 122,145; male population: 59,790) of the downtown core (Fig. 1).

Figure 1: Spatiotemporal evolution of syphilis for Toronto, 2006 to 2010.

Demographic and Behavioral Characteristics of Syphilis Cases

The epidemic curves for core and noncore areas fluctuate in the same rhythm with peaks and troughs occurring at generally the same time, suggesting that core and noncore case dynamics are synchronous. Epidemic curves intertwined between core and noncore areas, although the overall number of cases tended to be higher in noncore areas; patients in the core area were predominantly 30 to 44 years, whereas patients in noncore areas were predominantly younger than 30 and older than 44 years (Fig. 2, column 1). Incidence rate curves indicated that syphilis incidence rates were much higher for the core area than anywhere else and highest in the core area regardless of age, but highest for men 30 to 44 years old (Fig. 2, column 2).

Figure 2: Toronto syphilis epidemic curves and incidence rate curves for men and by age.

From 2006 to 2010, 79% of all male syphilis patients reported having sex with other men. We do not know if the remaining 21% reported “no” to having sex with same sex, or if these data are incomplete because of the way behavior data were collected during interview (simple check all that apply). The most frequently occurring sexual risk behaviors were as follows: not using a condom (56%), having more than 1 sex partner in last 6 months (46%), having a new sexual partner in past 2 months (10%), having a partner with multiple sex partners (10%), and meeting sexual partners at a bath house (6%). Epidemic curves indicate that the number of syphilis cases with risky sexual behaviors is higher for noncore and core areas and lower for outbreak areas (Fig. 3, column 1). Overlapping prevalence curves indicate that the prevalence of sexual risk behaviors is similar for core, outbreak, noncore-nonoutbreak, and noncore areas (Fig. 3, column 2). However, the prevalence of cases not using a condom, having more than 1 sex partner in the past 6 months, having a new sexual contact in the past 2 months, and having a sexual partner with multiple sex partners was generally higher outside the core area than inside the core area (Fig. 3, column 2).

Figure 3: Toronto syphilis epidemic and prevalance curves by sexual risk behavior with more than 5% positive response.

HIV-Syphilis Coinfection

Fourty-seven percent of syphilis cases were coinfected with HIV from 2006 to 2010. No statistically significant spatial clusters of HIV-syphilis coinfection were identified for Toronto (Fig. 4). The prevalence of HIV-syphilis coinfection was higher outside the core area compared with inside the core area when looking at the spatial distribution of coinfection by census tract alone (Fig. 4, map, column 1, row 1). In fact, many neighborhoods outside the core area had low syphilis infection rates, but high HIV-syphilis coinfection. However, when census tracts were aggregated by classification area (i.e., core, outbreak, noncore-nonoutbreak, and noncore areas), the absolute number of HIV-syphilis coinfection cases was comparable for core and noncore areas (Fig. 4, epidemic curve), and the prevalence of HIV-syphilis coinfection was highest for the core and eastern outbreak areas (Fig. 4, prevalence curve).

Figure 4: HIV-syphilis coinfection, 2006 to 2010.

BYM Model Results

Unadjusted posterior probabilities indicate that age-standardized syphilis incidence rates are significantly higher and more concentrated in Toronto’s downtown core area, as well as the northern, western, and eastern outbreak areas, compared with other parts of the city (Fig. 5, left map). These results corroborate the clusters identified using SatScan.

Figure 5: Syphilis and HIV-syphilis coinfection standardized incidence ratio maps for men in Toronto between 2006 to 2010, by 2006 Census Tract.

The unadjusted posterior probabilities for HIV-syphilis coinfection suggests that the age-standardized prevalence of HIV-syphilis coinfection is also higher in Toronto’s downtown core area, although more disperse with the highest prevalence estimates near the lake, and gradually decreasing radially outward from the downtown across the city (Fig. 5, right map). These results are more informative than the SatScan results of no clustering and provide insight into the context of a more generalized HIV-syphilis coinfection pattern for Toronto that may not be statistically significant but is still meaningful.

The syphilis incidence and HIV-syphilis coinfection BYM models adjusted for social determinants quantified that residential instability and ethnic concentration were significantly associated with age-standardized syphilis incidence rates (Table 1). Specifically, increasing residential instability was associated with increasing syphilis incidence rates, whereas increasing ethnic concentration was associated with decreasing syphilis incidence rates (Table 1). No social determinants of health were significantly associated with age-standardized HIV-syphilis coinfection (Table 2).

TABLE 1: Posterior Estimates of Model Coefficients for the Male Syphilis Model

TABLE 2: Posterior Estimates of Model Coefficients for the Male Syphilis-HIV Coinfection Model

DISCUSSION

Syphilis incidence rates increased within Toronto’s downtown core area and spread outward to neighboring environments, similar to what has been observed in New York,¹⁰ Baltimore,² and San Francisco.⁴ Some social determinants of health were associated with the observed spatial pattern of syphilis but not HIV-syphilis coinfection. Specifically, neighborhood level residential instability was associated with increased syphilis incidence rates, whereas ethnic concentration had a protective effect being associated with decreased syphilis incidence rates. HIV-syphilis coinfection was high for the epidemic, but did not exhibit spatial clustering. Therefore, our analysis did not find evidence that the syphilis epidemic was accelerating HIV transmission.

Our focus was on the spatial patterns of syphilis and HIV-syphilis coinfection over time, and the theoretical insights that could be hypothesized about the underlying sexual network, core group dynamics, and transmission dynamics from those patterns. We analyzed the spatial epidemiology of syphilis for Toronto using data reported to the TPH STI surveillance program. Our syphilis incidence rate estimates underestimate the true rates because we were limited to total male population estimates rather than total MSM population estimates for our population denominators.

It has been estimated that 2% of the male population in Canada is gay or bisexual.²⁹ Assuming a uniform distribution of MSM across Toronto, syphilis incidence rates could be as high as 500 to 4900 cases per 10,000 MSM in the core area and 5 to 500 cases per 10,000 MSM outside the core area. However, we know that the spatial distribution of MSM across Toronto is not uniform, given “The Village” (within the core area) and “Queer West” (outside the core area), which are neighborhoods with higher proportions of MSM.³⁰ Therefore, there could be significant variability in the magnitude of the error between neighborhood estimates. Our census tract syphilis incidence rates are underestimated, although it is unclear by how much, and the bias is probably nondifferential; that is, both core and noncore estimates are affected in such a way that we do not expect our estimation or interpretation of the spatial distribution of clusters to change significantly given this bias.

Based on our analysis, Toronto’s syphilis epidemic started with an outbreak in the downtown core area suggesting that the underlying sexual network is sufficient to maintain transmission but, under certain conditions, can increase the rate of infection and transmission. We did not have sexual network data to partner with our analysis to help contextualize and deepen our interpretation of the results; therefore, the mechanism by which these changes are made is unclear. However, transmission could be through introduction of the infection to another population through bridging (comparable to expansion of the core), a change in behavior reflecting a change in the sexual network structure, or both.

Core and noncore case dynamics in Toronto seem to be generally synchronous (Figs. 2 and 3). The epidemic curves indicate that cases in noncore areas fluctuate in the same rhythm as cases in core areas, similar to what was observed in San Francisco.⁴ This pattern suggests that syphilis cases wax and wane throughout entire geographic areas, not just in core or outbreak areas. In San Francisco, we saw that specific demographic characteristics (e.g., heterosexual blacks or white MSM) were common to syphilis cases in core areas and so were able to define core groups.⁴ In Toronto, we had limited demographic and behavioral data, and so our characterization of the core group in the core area is MSM. The prevalence of risky sexual behaviors was higher outside the core area than inside the core area. However, this observation should be interpreted with caution given the incompleteness of the behavioral data available. That said, we expect the quality of sexual behavior data to be better for the core area because that is where both the Hassle Free Clinic (largest sexual health clinic in Canada) and many STI interventions are targeted. If core group characteristics change during times of transition, such as during interepidemic periods,⁴ or when bridging to naïve groups, then sexual risk behaviors are also likely to change, especially as the underlying sexual network expands and potentially changes.

We used SatScan to detect clusters of high syphilis incidence rates, which is one of the more rigorous and widely accepted methods of cluster detection in spatial epidemiology and sexual health currently. SatScan has its limitations. Specifically, clusters within clusters cannot be detected, meaning outbreaks within core areas cannot be determined. In addition, during spatiotemporal analyses, cluster size is fixed so it is not possible to see how clusters change over time. We explored how clusters changed over time by conducting a series of space-only analyses with our data. This analysis enabled us to look for changes in the spatial distribution and size of clusters by year, as well as changes in the relative rates of infection for clusters (results not shown). There was good agreement between the spatiotemporal analysis and the space-only analysis over time. Furthermore, our BYM model results corroborated our SatScan results increasing our confidence in the power and validity of the observed spatial patterns and their interpretation.

The spatial distribution of syphilis for Toronto can be interpreted as one core area expanding over time as it grows to include neighboring communities, subpopulations, and sexual networks, or as one core area with 3 adjacent, independent, outbreak areas. These spatial patterns can be conceptualized to have occurred in 3 ways. The first is a gradual spread by simple diffusion. Sexual partners tend to be located close in space, and the sexual partners of cases in core areas tend to be closer in space than the sexual partners of cases outside core areas.³¹ Therefore, during epidemic conditions, one would expect surrounding environments to become affected before areas further away. That said, our SatScan space-only cluster detections for each year of syphilis data (i.e., no temporal dependence in data; results not shown) continued to identify four independent clusters—the core area and the 3 outbreak areas, suggesting that for Toronto, the core area is not expanding and the outbreak areas are independent of the core area over time.

A second possibility is that cases from core areas have traveled to outbreak areas acting as bridge contacts between discrete sexual networks. Neighborhoods within Toronto are quite distinct, and there is a strong historical east-west divide along the city’s zero line (Yonge Street, which runs north-south and is literally the numeric zero line separating east from west from a city planning perspective). The built environment around this zero, or divide line, has been predominantly nonresidential, further reinforcing a psychosocial divide, which has caused limited social or sexual mixing between east and west. The western edge of the core area ends at this zero line or divide, placing the core area and eastern outbreak area in the east (Fig. 1). The condofication of Toronto’s downtown³² is starting to breakdown the east-west divide such that there is more social mixing and likely a growing overlap or connectivity between east-west sexual networks, so bridging is a possibility. However, the city remains divided socially and, therefore, sexually.³² Residential instability was positively associated with syphilis incidence rates (Table 1), further suggesting that those neighborhoods experiencing intense transition and transformation (condofication) may be good areas to target STI resources and intervention.

A third possibility is that persons in outbreak areas are traveling into the core for sexual relationships. Travel into the core is the most likely form of bridging at work in Toronto’s syphilis epidemic. A large proportion of Toronto’s syphilis core area includes “The Village,” a predominantly gay neighborhood and sanctuary for MSM. Men who have sex with men often seek sexual partners in The Village and so (knowingly or unknowingly) many MSM living outside the core, in outbreak or noncore areas, are traveling directly into the syphilis core area seeking sex. Many of these men may also be “outsiders” or marginalized within the gay community, which may have sexual status and negotiation implications.³³

We were limited in the completeness of the sexual risk behavior data reported to TPH, so we limited our analysis to those behaviors with more prevalent reporting (≥5%). Epidemic and prevalence curves of risky sexual behaviors should be interpreted with caution but do suggest that cases outside the core area may practice riskier sex than cases within the core area (Figs. 2 and 3). Risky sexual behaviors practiced by cases outside core areas traveling to the core area for sex could contribute to bridge transmission of syphilis to naïve areas and populations. It could also explain the observed pockets of high HIV-syphilis coinfection outside the core area (Fig. 4).

Our data reflect the number of new syphilis cases (new infections) and not the number of clients (individuals) who had syphilis; for example, a client who was infected twice (i.e., repeat infection) between 2006 and 2010 will show up as 2 cases. In 2010, 522 syphilis cases were attributed to 509 clients, so repeat infection was low. Of the 509 clients, 241 clients (47%) were coinfected with HIV and 268 clients (53%) were not coinfected with HIV. Clients coinfected with HIV were more likely to have been reinfected with syphilis than clients who were not coinfected with HIV: of the 241 syphilis clients in 2010 with an HIV coinfection, 104 clients (43%) had at least 1 more syphilis infection(s) between 2006 and 2010, and of the 268 syphilis clients in 2010 without an HIV coinfection, 30 clients (11%) had at least 1 more syphilis infection(s) between 2006 and 2010. With these statistics in mind, the observed pockets of high HIV-syphilis coinfection outside the core area may be misleading in that the number of new syphilis cases with HIV coinfection may be high, driving coinfection prevalence up, but the actual number of clients coinfected with HIV and syphilis might be low, especially because HIV-positive clients had higher rates of syphilis reinfection in 2010.

Most cities with syphilis epidemics are likely to target resources and interventions to core areas and core groups because this is the intuitive reaction and standard recommendation^12,15–17 when responding to STI outbreaks and epidemics. However, despite these efforts, syphilis rates have not decreased, as evidenced by rebounding syphilis rates higher than premass treatment rates in Vancouver’s downtown core,³⁴ and the persistence of syphilis epidemics among MSM in Toronto, Vancouver, San Francisco,⁴ and other urban centers for more than a decade.

Toronto’s syphilis infection rates, HIV-syphilis coinfection, and sexual risk behaviors that usually contribute to STI transmission are high outside the core (Fig. 3), suggesting that peripheral sexual networks may be influencing high syphilis infection rates both inside and outside the core.^24,35 Toronto’s epidemic is mature and persistent, and mature epidemics need to target core and noncore areas.¹⁵ Alternatively, high rates of HIV-syphilis coinfection outside the core area, compared with inside the core area, may be driven by a small number of clients with HIV who may be getting reinfected and increasing coinfection rates. If this is true, again, we can inform interventions by targeting the HIV positive MSM population outside the core area. At the same time, reinfection rates were low so this is unlikely.

In response to elevated rates of syphilis in Toronto, periodic promotional campaigns were developed with main messages being threefold: get tested, get treated, and inform your partners. Campaigns were targeted to the MSM community and were evaluated. During the early stages of the outbreak, enhanced surveillance and social networking analysis was undertaken to better inform public health and community partners about the outbreak. A number of online Web sites, bathhouses, and one bar were identified as important in contributing to syphilis transmission and were targeted as part of the response (online banner advertising, expanded outreach, and testing offered onsite at physical locations). An iPhone application and text messaging services were developed as well for online partner notification through the use of InSpot. Syphilis laboratory guidelines and treatment guidelines were developed for and disseminated to community physicians. Partner notification efforts were enhanced.

Going forward, focusing on noncore areas may help address the persistence of the syphilis epidemic. Sexually transmitted infection messaging, services, and resources available in core areas may not reach MSM from noncore areas, resulting in reduced awareness of STI/HIV risk, delayed syphilis/HIV detection and treatment, increased syphilis/HIV transmission in noncore areas, and increased risk of reinfection inside the core, all of which will extend the life of the current syphilis epidemic.

Sexually transmitted infection epidemics evolve through predictable phases with shifts in the subpopulation location of the sexual and social networks that fuel them that often include movement back into core groups as the epidemic declines.^4,15,36,37 We hypothesize that the following observations may be generalizable to syphilis epidemics:

1. Core areas may be able to sustain high rates of syphilis infection better than outbreak and noncore areas.
2. Syphilis epidemics may originate through increasing syphilis rates in core areas that initiate outbreaks outside core areas.
3. Syphilis epidemics may recede back to core areas as they wane.
4. Targeting resources to core areas may prevent an epidemic in the early stages of a core area outbreak.
5. Targeting resources to core and noncore areas may be vital to push mature syphilis epidemics back into core areas.

Furthermore, we hypothesize that Toronto’s syphilis epidemic is driven by MSM and bridge contacts living outside the downtown core area, who seek sex in the downtown core area, and practice risky sexual behaviors, possibly because they are in unstable residential situations, or their ability to negotiate safe sex is compromised, or they are unknowing and underserved, or any combination thereof. Therefore, we recommend that Toronto’s syphilis epidemic response, resources, and intervention activities target MSM in core and noncore areas.