Estimating prevalence: chlamydia, gonorrhoea, and trichomoniasis

Prevalence data for chlamydia, gonorrhoea, and trichomoniasis were drawn from the data collected for the 2008 WHO estimates, PubMed literature searches (last search conducted on January 30, 2015) (S1 Text) and requests to the WHO regional STI advisors and other leading experts in the field. In addition, members of the International Union against STIs were asked to identify additional published or unpublished studies that met the study inclusion criteria. Two investigators applied inclusion criteria: sample size of at least 100; specimens collected from 2005 through 2012 (for studies in which no specimen collection date was specified the study had to be published in 2006 or later); population could be considered representative of the general population (study populations included pregnant women, women at delivery, women attending family planning clinics, military recruits, or individuals selected for participation in a Demographic and Health Survey); and study used an internationally recognised diagnostic test with adequate performance characteristics on urine, urethral, or cervicovaginal specimens. Studies conducted among the following groups were excluded because of reasons that are known to bias estimates of general population STI prevalence: patients seeking care for an STI or genital symptoms, women with abnormal Papanicolaou smears, blood donors, women attending gynaecology or sexual health clinics, remote or indigenous populations, men who have sex with men, and commercial sex workers. Duplicate data points were removed and if data were published in more than one paper the paper with most information was included in the database (Fig 1). PRISMA guidance was followed for the manuscript (S2 Text).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 1. PRISMA flowchart.

https://doi.org/10.1371/journal.pone.0143304.g001

Data from studies that met the inclusion criteria (S1 Data) were standardised to approximate the prevalence in the general population of a country by applying adjustment factors for the laboratory diagnostic test used, study location (rural vs. urban) and the age of the study population (S3 Text). In order to reflect the contribution of populations at higher risk of infection, the regional prevalence estimates were increased by 10% for males and females for all four infections. The 10% figure was based on the estimate used in the 2005/08 WHO estimates [6,7] and is a rough approximation of the contribution of these populations.

For some infections, particularly in men, there were fewer than three data points per region for the 2005 to 2012 time period. When there were insufficient data for men, the prevalence in men was assumed to be a proportion of the estimated prevalence in women. This proportion was assumed to be infection-specific, but to be the same across all regions and derived from a uniform distribution with a range of +/- 33.3% of the values used in the 2005 and 2008 estimates (Table 2).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 2. Ratios used to estimate prevalence when data for chlamydia, gonorrhoea, or trichomoniasis were limited.

https://doi.org/10.1371/journal.pone.0143304.t002

For women there were sufficient data points to generate estimates for all regions for chlamydia but not for gonorrhoea or trichomoniasis. In the absence of data, a ratio was assumed between the prevalence of these two infections and chlamydia. The estimated ratio was derived from a uniform distribution within a range of +/- 33.3% (Table 2).

The estimates for the High Income North America region (United States of America and Canada) were based on the United States national chlamydia estimates for 2012 and the gonorrhoea and trichomoniasis estimates for 2008 published by the United States Centers for Disease Control and Prevention (CDC) [11,12]. We assumed that the prevalence of gonorrhoea and of trichomoniasis did not change from 2008 to 2012 in the United States, that the prevalence of each infection in Canada was the same as in the United States, and that their data (calculated for a 15–39 year age range) could be extrapolated to the 15–49 year age range used for the WHO estimates. The geography, age, and high risk adjustments used for other regions in the 2012 WHO estimates process were not applied (S3 Text).

Estimating prevalence: syphilis

Syphilis prevalence estimates were based on the WHO 2012 estimates of maternal and congenital syphilis [18] reported through the GARPR system for 122 of 194 (63%) countries accounting for 76% of individuals aged 15–49 years globally. A PubMed search of the peer reviewed literature conducted in March 2014 identified data for an additional 8 (4.1%) countries missing GARPR data. Where estimates were available from both peer-reviewed literature and GARPR, we used the GARPR estimate because Ministry of Health officials provide the most nationally representative data available. Such national surveillance data are typically not published in the peer-review literature.

Syphilis prevalence estimates were calculated at country level before aggregating for regional and global estimates. All reported data were adjusted for laboratory test to approximate the prevalence of “probable active” infections, defined as having positive serology on both treponemal and non-treponemal tests [19]. We assumed that the prevalence of syphilis in women aged 15–49 years was the same in pregnant and non-pregnant women. In addition, we assumed that the ratio of the prevalence of syphilis in males to females was 1 with a range of +/- 33.3%. As in the 2005 and 2008 estimates, the regional syphilis prevalence estimates for men and women were increased by 10% to reflect the contribution of key high risk populations.

Statistical methods used to pool data

We used a Bayesian meta-analytic approach produce pooled prevalence estimates for the 10 regional groups where there were three or more data points for a particular infection in women or men. The beta-binomial approach for pooling over-dispersed binomial data was used as it provides a robust estimate for the average proportion based on data from heterogeneous studies [20]. The beta-binomial model is appropriate for meta-analysis of prevalence data when the outcome is rare, as for gonorrhoea and syphilis in some regions, and provides estimates of uncertainty that take into account the heterogeneity observed between studies [21].

Meta-analyses were performed with a Markov Chain Monte Carlo (MCMC) algorithm implemented with the BRrugs package [22]. For chlamydia, gonorrhoea, and trichomoniasis, ten thousand samples were generated from the posterior distribution for the expected mean prevalence of each regional grouping separately based on the beta-binomial model and these were used to calculate uncertainty intervals (UI) given by the 2.5 and 97.5 percentiles. Uncertainty, calculated under this approach, reflects both the level of precision within studies (i.e. sample size) and also additional between-study variation arising either from real differences between study populations or differences in study methodology. Global prevalence estimates for chlamydia, gonorrhoea and trichomoniasis were then calculated by weighting the regional prevalence estimates according to population size. The same approach was applied for syphilis, but country level data were used and pooled within each of the Global Burden of Disease project regions [1]. For syphilis, uncertainty was quantified based on analysis of the posterior expected value for country level prevalence and global prevalence calculated by weighting according to country population size. All analyses were carried out using R statistical software [23].

Estimating incidence: chlamydia, gonorrhoea, trichomoniasis, and syphilis

Regional incidence estimates for nine of the 10 regions were estimated from prevalence estimates. The approach was the same as in 2005 and 2008 but we also applied a range of +/- 33.3% around the estimated duration of infection to reflect uncertainty when calculating incidence using the equation: incidence = prevalence/average duration of infection. For the High Income North America region we used incidence estimates for chlamydia, gonorrhoea and trichomoniasis based on the United States national incidence estimates [11,12].

It was assumed that the average duration of infection in each region varies according to the average duration of infection in the absence of treatment for symptomatic and asymptomatic individuals and the probability symptomatic and asymptomatic individuals are treated appropriately. For the 2005 and 2008 estimates countries were classified into one of three treatment groups with a corresponding average duration of infection based on a literature review and expert consultation [6]. A review of the literature in 2012 found insufficient data to merit changing the assumptions for duration of infection used in 2005 and 2008 (S4 Text).

Prevalence and incidence estimates from the 10 regions were aggregated so that results could be presented by WHO regional groupings and by World Bank income classification [24]. United Nations Population data for women and men aged 15–49 years were used to generate numbers of prevalent and incident cases [25].

Summary of findings

This study suggests that, in 2012, there were about 273 million prevalent cases of the four curable STI among adults aged 15–49 years: 128 million cases of chlamydia, 27 million cases of gonorrhoea, 101 million cases of trichomoniasis, and 18 million cases of syphilis. These figures correspond to an estimated incidence of 131 million new chlamydia cases, 78 million new gonorrhoea cases, 143 million new trichomoniasis cases, and 5.6 million new syphilis cases in women and men aged 15–49 years globally. Together the four infections accounted for 357 million new infections in adults that year, or an average of nearly one million new infections each day. Because STI co-infection is common, however, it is unclear exactly how many individuals this represents.

Strengths and limitations

The methods used to generate the 2012 estimates of STI prevalence were based on the methods used for the 2005 and 2008 estimates, with modifications to improve the quality of the estimates based on input from external expert consultations. In particular, the prevalence data were pooled into 10 regions based on epidemiological similarity, geography and data availability rather than by geography alone and greatly improved data for syphilis in women were available at country level through the GARPR system.

The biggest challenges to estimates of this nature are the quantity, representativeness and risk of bias of the available data. Data were included for an eight year time window (2005 through 2012), which means that changes in prevalence over time within this window will adversely affect the accuracy of estimates for 2012, a single year. However, even with an 8 year time window the number of data points for each infection was limited, particularly for men; only two usable data points were identified for trichomoniasis in men. As a result, the authors applied adjustment factors to prevalence estimates derived from available data for all four STIs in men and for gonorrhoea and trichomoniasis in women. The current estimates are limited to urogenital infections. Gonorrhoea and chlamydia can also be rectal and oropharyngeal infections; however, data on their prevalence is even more limited.

Where data were available, estimates of STI prevalence varied substantially. Between-study heterogeneity could represent true variation in infection prevalence between different populations but could also result from methodological sources of bias. For example, the response rate in chlamydia prevalence surveys in high income countries is inversely associated with the estimated prevalence [26]. Pooling data is not ideal under such circumstances but was necessary to generate estimates at regional and global levels. The Bayesian approach, using a beta-binomial model was deemed the most appropriate method for meta-analysis, given the challenges arising from the data collection [20,21]. The over-dispersion parameter in the model made an allowance for the between-study heterogeneity, which is reflected in the uncertainty intervals around the mean prevalence. The estimated incidence rates were derived indirectly, based on the relationship between prevalence and the duration of infection, owing to the absence of good quality incidence data from empirical studies. Assumptions were made about the duration of infection for women and men, levels of symptomatic and asymptomatic infection, all of which vary widely within and between countries. The 2012 duration estimates are unchanged from the 2005 and 2008 estimates although we included a range of values to reflect some of the uncertainty around these estimates.

Comparison with other studies

The Global Burden of Disease 2013 study included morbidity estimates of chlamydia, gonorrhoea, and trichomoniasis, and for tertiary but not primary or secondary syphilis [1]. The Global Burden of Disease study estimated that there were 130 million new chlamydia cases, 74 million new gonorrhoea cases, and 55 million trichomoniasis cases among adults aged 15 to 49 years in 2013. The figures for chlamydia and gonorrhoea were very similar to the WHO estimates and well within the WHO uncertainty intervals. The difference in the trichomoniasis estimates is primarily driven by the high number of new cases in men estimated by WHO (143 million). The high incidence of trichomoniasis in men in the WHO estimates results from the duration of asymptomatic infection, which was assumed to be 1.5 months in men and 18 months in women. The Global Burden of Disease process did not vary presence of symptoms or duration of infection by gender, nor did it include standardization for laboratory test performance. However, for any global estimation process, improved data on trichomoniasis among men will be critical for improving the robustness of the estimates. The estimated prevalence of chlamydia in high income countries is in line with estimates from a published systematic review [26].

Interpretation of the findings

The 2012 global estimates of STI prevalence and incidence are not directly comparable with earlier estimates and are not appropriate for doing trend analyses, owing to differences in the methods used for each set of estimates. In 2012 the estimate for the number of new cases for the four infections combined was 357 million. The corresponding figures for 2005 and 2008 were 448 and 498 million, respectively. The empirical data for syphilis are the most robust and do suggest decreases over time. A recent analysis of the ANC syphilis seropositivity data on which the current estimates were based was able to document a decrease in syphilis seropositivity among consistently reporting countries [18]. For the other infections it is not possible to say how much of the overall decrease reflects changes in the representativeness of the data, changes in the methods used for analysis, or true decreases in infection.

STIs are a global epidemic but there are distinct regional variations. The highest prevalence and incidence rates for syphilis were found in the WHO African Region. Access to antenatal syphilis screening contributes to regional differences in syphilis prevalence. In the African region, the median reported proportion of ANC attendees tested for syphilis was 58%. The median proportions were markedly higher in the other regions and ranged from 83 to 99% [27]. Of note, estimated chlamydia prevalence was relatively low in the African Region and highest in the Region of the Americas and the Western Pacific in women and in the Western Pacific in men. The substantial unexplained heterogeneity between studies of chlamydia prevalence across and within regions deserves further detailed investigation. Both prevalence and incidence of all four infections were lowest in the European and South-East Asian regions. Possible contributing factors are a lower risk of infection due to lower numbers of partners and increased condom use, or better social and economic conditions including clinical STI services. Alternatively, data from the 2005 to 2012 time period might not be representative of these regions.

In all regions, the estimated prevalences of chlamydia in women and men, and of trichomoniasis in women were greater than gonorrhoea or syphilis prevalence. There are several potential factors contributing to these findings. Repeat infection with both trichomoniasis and chlamydia are quite common following currently recommended single dose STI treatment regimens, often because partners are not adequately treated [28]. Additionally, most infections with chlamydia and trichomoniasis in women are asymptomatic and remain untreated [2]. Because most men with gonorrhoea are symptomatic, syndromic management has been effective in decreasing gonorrhoea burden. The higher frequency of asymptomatic STIs in women compared with men, resulting in a longer average duration of infection contributes to the finding that over 69% of the prevalent STI cases but only 48% of the new STI cases were among women.

When looking at incidence by World Bank income classification, 91% of incident infections were among individuals in low, lower-middle, and upper-middle income countries where 84% of the population aged 15 to 49 years live. Over half (56%) of all incident infections were in upper-middle income countries where only 36% of the population aged 15–49 years lives. The prevalence of chlamydia and of gonorrhoea in upper-middle income countries were strikingly high, but it is unclear whether this was an artefact of variable data or whether this was a true epidemiological finding due to differences in sexual risk behaviour or access to care. For syphilis in women, the infection with the most robust available data, the prevalence of infection decreased as average country income increased. This trend also held for gonorrhoea and trichomoniasis, apart from the data points for upper-middle income countries.

Implications of the findings

WHO and its partners are looking at options to improve the quality of future estimates and supporting countries to generate their own national estimates. More data from studies conducted amongst women and men at low risk of infection in the general population are needed to improve estimates of STI prevalence. For example, South Asia, South East and East Asia Regions together had only 10 prevalence data points for chlamydia, gonorrhoea, and trichomoniasis that met the inclusion criteria, yet these three regions account for over half (54%) of the world population aged 15–49 years. Similarly, there were only nine data points for Western, Central, and Eastern Europe and Central Asia, all of which were for chlamydia. For many countries, national STI reference laboratories and surveillance systems need to be strengthened. WHO is also exploring alternative methods for generating STI prevalence and incidence estimates including approaches that use data collected from studies done in key populations, case reports, and data from routine programmes. Robust estimates of the size of key populations such as commercial sex workers and men who have sex with men will also be needed. Statistical methods such as hierarchical Bayesian modelling [20] can provide the tools to combine and explore sources of uncertainty in data from different sources.

In conclusion, the global prevalence and incidence of urogenital chlamydia, gonorrhoea, trichomoniasis, and syphilis in adult men and women aged 15 to 49 years remain high, with nearly one million new cases of curable STI acquired each day. The global threat of antimicrobial resistance, particularly for gonorrhoea [18], makes it essential to improve the monitoring of changes in STI incidence. Robust estimates of STI prevalence and incidence data are important for designing, implementing and evaluating STI interventions and for advocating for funding for program and research efforts to develop improved tools such as more efficacious therapeutics, point-of-care diagnostics, vaccines and microbicides appropriate for global use.

Disclaimer

The authors alone are responsible for the views expressed in this article and they do not necessarily represent the views, decisions or policies of the institutions with which they are affiliated.