The prevalence and incidence of active syphilis in women in Morocco, 1995-2016: Model-based estimation and implications for STI surveillance

Background Evolving health priorities and resource constraints mean that countries require data on trends in sexually transmitted infections (STI) burden, to inform program planning and resource allocation. We applied the Spectrum STI estimation tool to estimate the prevalence and incidence of active syphilis in adult women in Morocco over 1995 to 2016. The results from the analysis are being used to inform Morocco’s national HIV/STI strategy, target setting and program evaluation. Methods Syphilis prevalence levels and trends were fitted through logistic regression to data from surveys in antenatal clinics, women attending family planning clinics and other general adult populations, as available post-1995. Prevalence data were adjusted for diagnostic test performance, and for the contribution of higher-risk populations not sampled in surveys. Incidence was inferred from prevalence by adjusting for the average duration of infection with active syphilis. Results In 2016, active syphilis prevalence was estimated to be 0.56% in women 15 to 49 years of age (95% confidence interval, CI: 0.3%-1.0%), and around 21,675 (10,612–37,198) new syphilis infections have occurred. The analysis shows a steady decline in prevalence from 1995, when the prevalence was estimated to be 1.8% (1.0–3.5%). The decline was consistent with decreasing prevalences observed in TB patients, fishermen and prisoners followed over 2000–2012 through sentinel surveillance, and with a decline since 2003 in national HIV incidence estimated earlier through independent modelling. Conclusions Periodic population-based surveys allowed Morocco to estimate syphilis prevalence and incidence trends. This first-ever undertaking engaged and focused national stakeholders, and confirmed the still considerable syphilis burden. The latest survey was done in 2012 and so the trends are relatively uncertain after 2012. From 2017 Morocco plans to implement a system to record data from routine antenatal programmatic screening, which should help update and re-calibrate next trend estimations. The Spectrum-STI epidemiological projection tool was used to estimate and project the prevalence of active syphilis in adult women (age 15–49 years) in Morocco between 1995 and 2021 based on statistical fitting of data from surveys conducted in pregnant women attending ANC or family planning services [8]. The Spectrum-STI tool is embedded in a broader suite of demographic and health burden and impact modelling program projection tools used by over 120 countries to estimate the burden of HIV/AIDS and the associated need for HIV/AIDS treatment. The software is available free of charge at: http://avenirhealth.org/software-spectrum.php. The data and assumptions used to generate the Morocco estimates were reviewed, discussed and agreed during two technical workshops, held in May and September 2016. Participants at these meetings included representatives of Morocco’s Ministry of Health and its provincial offices, HIV/AIDS and Maternal and Child Health programs; Morocco’s central reference laboratory, the World Health Organization and UNAIDS, and other partners supporting or implementing the national HIV/STI response. All data collated had been collected and documented earlier, independently from the current study, by Morocco’s HIV/STI program and HIV/STI surveillance unit within the Ministry of Health. Active syphilis infection was defined as a patient positive concurrently on both the Rapid Plasma Reagin (RPR) test and the Treponema pallidum haemagglutination assay (TPHA) test [11]. The details of the model used to generate the syphilis prevalence trends are described in detail in [8]. In brief, syphilis prevalence data from surveys and routine programmatic screening in pregnant women presenting for ANC, women attending family planning (FP) clinics, and any general adult population surveys that met the criteria for representativeness and quality used in the WHO 2012 regional and global estimations [11] were collated. Prevalence estimates were adjusted for the diagnostic test used: against the gold standard of dual RPR and TPHA positivity, which was taken as observed, prevalences based on RPR-positivity regardless of TPHA status were adjusted downward to 0.7 of the observed prevalence for ANC and FP populations; or to 0.6-fold for other general populations; prevalences based on TPHA-positivity regardless of RPR status were adjusted to 0.8-fold the observed value regardless of the population, prevalences based on rapid TPHA-based test in an ANC population to 0.7-fold; and prevalences based on an unknown test to 0.75-fold [8, 12]. Test-adjusted prevalence data were then adjusted upward by 10% to account for under-sampling of high-risk populations in general population surveys, also based on WHO’s 2005, 2008 and 2012 regional and global estimations methodology [11]. Each data point was then assigned a weight, reflecting its national representativeness. For Morocco’s ANC surveys, this weight was calculated by dividing the number of ANC sites sampled in a particular year by 30, the maximum number of ANC sites (in 2009). For example, the 1996 survey, with 9 sites, was given a 30% (= 9/30) weight. Time trends were fitted by logistic regression through all data points combined, assuming no systematic prevalence differences between ANC women, FP women, other general adult women, or general adult men, similar to the approach of the WHO’s regional and global estimations [11, 13, 14]. 95% confidence intervals (CI) were generated by bootstrapping, and the median result across 10,000 bootstrapping iterations taken as the best estimate [8]. Incidence was derived from the prevalence trend estimates. We assumed that the incidence hazard or density (among uninfected people) was constant in each of the consecutive intervals of length 1 year, for t starting from 1995. If the incidence hazard or density, i, and the duration of the STI disease episode, D, are constant in an interval (t0, t0 + 1), then for all t in that interval, the prevalence satisfies the equation: where r = 1/D. We solved this equation for i piecewise every year, after setting t = t0 + 1, for t0 = 1995… to obtain its time trend. From this incidence hazard, the corresponding incidence rate ‘IR’ per capita in the overall national population was calculated as:IR = i(1 − p), where p is the prevalence (see S1 Text). In the default estimation, the average duration of infection with syphilis was assumed to be constant over time and set at 2.42 years. This estimate was based on the values used by the WHO in their 2012 regional and global STI estimations for the Eastern Mediterranean region [11] (S1 Table [11]). In an alternative scenario, the duration of infection was assumed to have progressively decreased over time, from 3.1 years in 1995 to 2.42 years in 2015. This change was made to reflect improvements in access to treatment. In 1995, we assumed 30% of individuals with symptoms were treated, and by 2015 this had increased to 60%, the value also assumed in the WHO 2012 estimations for the Eastern Mediterranean region. The assumed doubling of treatment coverage was in line with our recent Spectrum-STI-based estimation of Morocco’s gonorrhea and chlamydia trends, where we assumed that the treatment coverage of symptomatic gonorrhea and chlamydia in men doubled from 1995 to 2015 [15]. The latter is was based on treatment coverage reported through Knowledge, Attitudes and Practices surveys in youth [16, 17] and among men with urethral discharge [18, 19]. Reasoning that the improved treatment coverage for UD reported in these surveys reflected a general improvement in STI clinical services, we assumed here that treatment coverage also doubled for syphilis. Applying the durations of untreated syphilis and of treated syphilis (S1 Table) to the proportions that are treated and untreated (S1 Table), yielded a weighted average duration of infection that decreased from 3.1 years at 1995 to 2.42 years at 2015. We assumed no disease mortality for adult syphilis. The 95% CIs on incidence bounds reflected the uncertainty (estimated by bootstrapping) in prevalence, and an additional uncertainty on the duration of infection, set at ±50%. For comparison with the Spectrum-estimated trends based on ANC, FP and general population surveys, we compiled syphilis prevalence data from sentinel surveillance and integrated bio-behavioural surveys collected since 1995 in Morocco, as available at the Ministry of Health. To support interpretation of estimated syphilis trends, we assessed time trends in HIV incidence and in risk behaviours underlying HIV incidence, estimated using the HIV transmission model Goals, another module of the Spectrum suite [10]. The representation of Morocco’s HIV epidemic and its drivers in the Goals model was that used in a recent multi-country modelling of the UNAIDS global Fast Track targets [9]. For Morocco, the model representation had been informed by analysis of data from Morocco’s sentinel surveillance and IBBS data, two Modes-of-Transmission studies [4, 20, 21] by the Ministry of Health with the UNAIDS country office, and Morocco’s 2015 round of annual HIV burden estimation that had used the Spectrum module AIDS Incidence Model [22].