Incidence of sexually transmitted infections in men who have sex with men and who are at substantial risk of HIV infection – A meta-analysis of data from trials and observational studies of HIV pre-exposure prophylaxis

Background Men who have sex with men (MSM) and who engage in condomless anal intercourse with casual partners are at high risk of acquiring sexually transmitted infections (STIs), but reliable epidemiological data are scarce. Studies on HIV pre-exposure prophylaxis (PrEP) enrol MSM who indicate that they engage in behaviour that puts them at high risk of acquiring HIV. Because they also screen for STIs at regular intervals, these studies may serve as a valuable source to estimate incidence rates of STIs in this subpopulation of MSM. Methods We systematically searched for trials and observational studies of PrEP in MSM that reported data on the incidence of STIs during the study period. Incidence rates were calculated as events per 100 person-years (py) with 95% confidence intervals (CI). Data from individual studies were pooled building subgroups along study types and geography. We performed sensitivity analyses, including data only from studies that met pre-defined quality criteria. Results Twenty-four publications on 20 studies were included. The majority of studies reported that sexual behaviour and/or STI incidence remained stable or decreased during the study period. For syphilis, incidence rates ranged from 1.8/100py to 14.9/100py, the pooled estimate was 9.1/100py (95%-CI: 7.7–10.9). Incidence rates for gonorrhoea and chlamydia of any site ranged from 13.3/100py to 43.0/100py and 15.1/100py to 48.5/100py, respectively. Considering only studies that met the criteria for sensitivity analysis yielded pooled estimates of 39.6/100py (95%-CI: 32.9–47.6) and 41.8/100py (95%-CI: 33.9–51.5), respectively. The overall estimate for hepatitis C incidence was 1.3/100py (95%-CI: 1.0–1.8). Conclusions Despite partly heterogeneous results, the data depict high incidence rates of STIs among MSM who engage in higher-risk sexual behaviours such as condomless sex with casual partners. This subpopulation of MSM requires access to STI screening at close intervals. By offering access to structures that provide regular STI monitoring and prompt treatment, PrEP may not only decrease HIV incidence but also have beneficial effects in decreasing the burden of STIs.


Introduction
Data on the epidemiology of sexually transmitted infections (STIs) suggest a high burden of disease worldwide [1].Increases in the incidence of chlamydia, gonorrhoea and syphilis infections have been observed in numerous countries, particularly among men who have sex with men (MSM) [2][3][4][5][6].At the same time, the decreasing susceptibility of bacterial pathogens such as Neisseria gonorrhoeae [7][8][9][10][11] and Mycoplasma genitalium [12,13] to antimicrobials is becoming a major public health problem, especially given evidence that the risk of HIV transmission is higher in the presence of other STIs [14][15][16][17].
Reliable data on the incidence of STIs among sexual minorities are difficult to obtain [18][19][20].In some countries, reporting systems for STIs may include information on the probable route of transmission.However, calculating incidence rates for STIs among MSM from these population-based data can rely only on reported cases for the numerator and estimates of the denominator, which is unknown (i.e., the total number of MSM at risk of acquiring an STI) [21,22].Therefore, from an epidemiological viewpoint, a more reliable method to generate estimates of STI incidence rates in sexual minorities is to use incidence data from longitudinal cohort studies.
In the present systematic review and meta-analysis, we aimed to generate effect estimates for incidence rates of STIs in MSM who engage in sexual behaviours that put them at high risk of acquiring STIs.To do so, we used data on the incidence of STIs diagnosed during the study period in participants from studies on HIV pre-exposure prophylaxis (PrEP).PrEP is a biomedical form of primary HIV prevention that has been shown to be highly effective in clinical trials [23][24][25][26] and cohort studies [27][28][29][30][31][32][33][34][35].Most PrEP trials and cohort studies only enrol MSM who indicate that they engage in behaviour that puts them at high risk of acquiring HIV, which is usually defined in these studies as reporting having had condomless anal intercourse within a specified amount of time with a specified minimum number of partners.Because they also screen for STIs at regular intervals, these studies may serve as a valuable source of STI incidence data in this subpopulation of MSM.

Materials and methods
We conducted a systematic review and meta-analysis of data on the incidence of STIs in published trials and cohort studies of PrEP in MSM at risk of HIV infection.A review protocol was developed and followed but not published (S1 File).

Eligibility criteria
Publications were eligible if they reported results from clinical studies of PrEP in MSM at risk for HIV infection and had a follow-up period of at least three months.To be included, studies had to report the incidence of STIs other than HIV and provide sufficient data on the personyears of follow-up to calculate incidence rates.Any form of testing for STIs had to be performed at regular intervals; studies that relied exclusively on self-reported incidence data were not included.We considered full-text publications and conference abstracts.Language was restricted to English, French and German.

Information sources and search strategy
We searched Ovid MEDLINE, Embase and Cochrane CENTRAL on 28 September 2017 and conducted an update search on 25 July 2018 covering the period from the inception of the data base through the date of the search.Autoalerts for all searches were included through 1 August 2018.The full search strategies can be viewed in the online supplement (S1 File).Resource limitations meant that authors were contacted for missing data only for studies that had been included in the original search.

Study selection and data collection
We checked all identified records for eligibility using a two-step process, screening titles and abstracts first, and assessing full texts of records included thereafter.Data were extracted by two evaluators (RNW and CD).Disagreement was resolved by discussion or by involving a third investigator (AN, MG).Apart from incident STI diagnoses and corresponding personyears of follow up, data were extracted on study type and characteristics, baseline characteristics of the sample, and any reported changes in sexual behaviour or STI incidence.We did not extract data that were only presented graphically.As data from a randomised controlled trial suggests that PrEP could have an effect on the incidence of genital herpes [36], we considered data on genital herpes only if these were reported separately for placebo or non-intervention arms.For data on hepatitis B, we chose the same approach since tenofovir is an established medication for chronic hepatitis B.

Assumptions, simplifications and calculations
If incidence rates with 95% confidence intervals (95%-CIs) and the total person-years of follow-up were not reported, we calculated person-years as follows: If possible, we multiplied the number of participants per visit by the length of time since the previous visit, or by multiplying the number of participants by the mean follow-up time.Where only the overall number of participants at baseline and at the end of follow-up were reported and the number of participants for each time point was missing, we calculated person-years by multiplying the number of participants who completed the study plus half the number of participants who had not completed the closing visit by the length of the follow-up period.For all of these approaches, we used the number of incident diagnoses whenever possible; however, in some cases only the number of participants infected was available.
Incidence rates were calculated as events per 100 person-years of follow-up.Assuming a Poisson distribution, we used log incidence rates and corresponding standard errors to determine incidence rates with 95%-CIs.If no incident cases occurred during follow-up, we added a correction of 0.5 to the number of cases and person-years of follow-up as described in similar studies [37,38].

Synthesis of the results
STI incidence data were grouped according to study type and geography.We calculated pooled estimates of the incidence rates and 95% CIs using a random effects model to account for methodological and clinical differences between the studies.Heterogeneity was quantified by means of the Higgins's I-squared test statistic [39].All calculations were performed using Stata SE 14 (metan command package).

Assessment of the quality of the data
In order to evaluate the quality of the data, three criteria were established: First, we evaluated whether the data on incidence rates or numbers of incident STI diagnoses and person-years of follow-up were explicitly reported or easy to calculate or interpret.Second, we assessed whether studies reported having performed screening procedures for the detection of STIs during the follow-up.To satisfy this quality criterion, studies had to explicitly state having performed screening procedures at a minimum of six monthly intervals and in accordance with the Centers for Disease Control and Prevention (CDC) recommendations for STI screening in sexually active MSM [40].Third, we evaluated the size of each study, considering more than 500 personyears of follow-up to be an indicator of higher quality incidence data.

Further analyses
To examine sources of heterogeneity, we separately evaluated subgroups along study design and geography.We performed a sensitivity analysis for each outcome, only considering data extracted from studies that explicitly reported incidence rates (or provided easy-to-interpret data for numbers of incident diagnoses and person-years of follow-up) and that reported having applied STI screening methods that met our quality criterion #2.We did not undertake any formal tests of publication bias in our meta-analysis because our outcome of interest (i.e., STI incidence rates) was not an efficacy outcome that might have led to the selective publication of study results.

Results
Our literature search yielded 2416 records, of which 1581 remained after duplicates had been removed.Of these, 1456 were excluded during the title and abstract screening, leaving 125 publications to be evaluated as full texts.Finally, 24 publications reporting the results of 20 studies were included.Fig 1 depicts the numbers of records identified, included and excluded [41].

Changes in sexual behaviour and STI incidence during the study period
Two RCTs [24,26] and four cohort studies [28,32,33,50] reported that sexual behaviour had not changed with respect to condom use and/or the number of sexual partners.Two of these studies [28,50] and one further cohort study [49] additionally reported no change in the incidence of STIs.One retrospective cohort study [51] reported an increase in STI incidence compared to the time before PrEP initiation.However, this increase was not significant after adjusting for the number of screening visits.A decrease in the number of sexual partners and instances of condomless anal intercourse was observed in an RCT [23] and its open-label extension [31].A smaller overall STI incidence in the second compared to the first half of the study was reported in one cohort [34].In contrast, decreasing rates of condom use were reported in three cohort studies [30,47,52], and increasing incidence rates for STIs observed in two cohort studies [27,29].Four studies did not report on changes in sexual behaviour or STI incidence during the follow-up [25,[44][45][46].

Gonorrhoea and chlamydia
Data on the incidence of gonorrhoea and chlamydia were reported in three placebo-controlled RCTs [23,24,26] one active-controlled RCT [44] one open-label RCT [25] and in eight cohort studies [27][28][29]33,45,47,50,51].Grant et al. 2010 reported only testing for urethral gonorrhoea and chlamydia infection upon positive urine leukocyte esterase [23] and Lal et al. reported testing for genital gonorrhoea only upon clinical suspicion [27].For two studies, routine sampling did not include the throat [44] or it remained unclear, whether this site was included [45].In one cohort study, screening tests only included anal and pharyngeal sites for gonorrhoea, and anal and urethral sites for chlamydia [27].Three studies reported on persons infected, not on events of infection during follow-up [26,44,50].In the following we have put the different anatomical locations in boldface to improve readability.
One cohort study reported data on the combined outcome of urethral gonorrhoea and/or chlamydia [47].In 179.5 person-years of follow-up, 32 cases of urethral STIs were diagnosed, leading to an incidence rate of 17.8/100py (95%-CI: 12.6-25.2).
One placebo-controlled RCT [26], one head-to-head RCT [44], one open-label RCT with data only derived from the PrEP group [25], and four cohort studies [27,28,45,51] reported on gonorrhoea and chlamydia of any of the body sites tested.The cumulative follow-up was 2093.2 and 2057.2 person-years, respectively.The incidence of gonorrhoea of any site ranged from 12.2/100py (95%-CI: 9.0-16.4) in an active-controlled RCT reporting the number of persons infected during follow-up rather than reporting the number of events of infection [44] to 43.0/100py (95%-CI: 37.5-49.3) in the US PrEP Demonstration Project [28].The pooled estimate from all studies reporting this outcome was 27. ).The heterogeneity of the incidence rates remained high when different geographic regions were pooled separately.Three studies [25,28,51] had incidence data easily accessible and reported applying proper screening procedures for these outcomes (nucleic acid amplification tests performed on anal and throat swabs and urine samples).Considering these studies only, the pooled estimate was 39.6/100py (95%-CI: 32.9-47.6,I 2 = 57.6%).

Mycoplasma and ureaplasma
No study reporting on the incidence of genital or rectal mycoplasma or ureaplasma infection was identified.

Viral hepatitis
One placebo-controlled RCT [26], one open-label RCT [25], and seven cohort studies [30,46,48,[50][51][52][53] reported data on the incidence of hepatitis C (HCV) during a cumulative follow-up of 3730.5 person-years.Of these studies, only five [25,30,46,51,52] reported using serology assays to screen for HCV at regular interviews.In the study by McCormack et al. [25] this applied only to participants who reported injecting or snorting drugs, fisting or the use of sex toys.The lowest HCV incidence was reported in a Canadian retrospective cohort study with zero events during 109 person-years of follow-up [51].The highest HCV incidence rate was reported for the Dutch AmPrEP cohort [48] and totalled 1.9/100py (95%-CI: 1.1-3.4).The overall estimate for HCV incidence was 1.3/100py (95%-CI: 1.0-1.8,I 2 = 0.0).Looking separately at the cohort studies, the pooled estimate was almost identical.Although the incidence rate of HCV in European studies was twice the incidence rate in North American studies, the confidence intervals for these estimates overlapped, see Fig 5. Pooling data only from the sensitivity analysis sample [25,30,46] yielded a pooled estimate for HCV incidence of 1.2/ 100py (95%-CI: 0.8-1.8,I 2 = 0.0).
One RCT [26] reported no incident cases of hepatitis B during 175 person-years of followup in the placebo group.
The incidence of hepatitis A was reported in one French cohort study [52].Six cases were diagnosed during a follow-up of 486 person-years, yielding an incidence rate of 1.2/100py (95%-CI: 0.6-2.8).

Participants with at least one STI and other STIs reported
Six studies reported on the outcome 'participants diagnosed with at least one STI', and one study each reported on anogenital warts, herpes simplex seroincidence, molluscum contagiosum and trichomonas.These data are reported in the online supplement (S2 File).

Discussion
The present meta-analysis aimed to generate estimates for incidence rates for various STIs among MSM who engage in high-risk sexual behaviours, such as condomless anal intercourse with casual partners.The results indicate that the incidence rates for STIs in this subgroup of MSM are high.The incidence rate for a diagnosis with any STI was about 84/100py in the four cohort studies that met our criteria for higher-quality data.The bulk of diagnoses with any STI were due to chlamydia and gonorrhoea infections at different anatomical sites.Here, rectal gonorrhoea, chlamydia or both played a major role, with a pooled incidence rate of more than 50/100py.This is worrying given the rising levels of antimicrobial resistance seen especially in Neisseria gonorrhoea in recent years [7][8][9][10][11].Although Mycoplasma genitalium infections have recently received increased attention as a difficult-to-treat STI [12] and epidemiological research partly supports screening for this infection in populations at higher risk, such as MSM [12,[54][55][56], no corresponding incidence data were reported in the studies included in our meta-analysis.
Our overall estimate for the incidence rate of hepatitis C infections was 1.3/100py.This is substantially higher than the incidence of 0.4 per 1000 persons per year seen in HIV-uninfected MSM and even the incidence of 7.8 per 1000 persons per year seen in HIV-positive MSM reported in a recent meta-analysis that included both prospective and retrospective studies [57].Recent research suggests that only a proportion of the hepatitis C infections in this susceptible population are sexually transmitted and that practices involving intravenous or nasal drug use play an important role [58,59].Due to the potentially severe sequelae and the high cost of treatment, our findings are worrying and suggest that policy makers may wish to consider redoubling efforts at primary prevention in those at highest risk (e.g.routine screening, needle exchange programmes).In the one study reporting incidence data for hepatitis A, a rate of 1.2/100py was observed.While this is a high rate by any measure, it must be kept in mind that this French cohort study was conducted during an outbreak of hepatitis A among MSM in several European countries [60][61][62][63][64][65], including France [66].Nevertheless, these data underline the importance of increasing hepatitis A vaccination coverage in MSM, specifically in those who engage in sexual behaviour that puts them at high risk of acquiring HIV and other STIs.
Our study does not attempt to compare STI incidence rates among MSM taking PrEP versus MSM not taking PrEP.One such comparison has been performed previously [67].However, the validity of such studies is questionable [68].To be eligible to participate in clinical trials of PrEP, participants had to be at high risk of infection with HIV (and hence other STIs) already.Comparing this group of MSM to other MSM who are not at high risk, or to the overall population of MSM, will lead to predictable and meaningless results.Moreover, the frequent testing for STIs in the PrEP studies will lead to an increase in the number of detected STI diagnoses.Indeed, one of the PrEP cohort studies included reported that screening every six months instead of quarterly would have delayed diagnosis of STIs in 24% of the participants [47].A retrospective cohort study that compared the incidence of STIs in the year before PrEP initiation to the year after, found a significant increase only when the data were not adjusted for testing frequency [51].Two longitudinal studies comparing the incidence of STIs before PrEP initiation to that during PrEP intake have reported mixed results: in the first, which assessed the incidence of rectal STIs, no increase was seen [69], whereas in the second, a modest but significant increase in 'any STI' and Chlamydia infections was reported [70].Generally, the majority of studies included in our meta-analysis reported either a stable or decreased riskbehaviour and/or STI incidence during follow-up.This finding is supported by a recent casecrossover study [71].
Since the participants included in the studies benefit directly from their participation (free PrEP medication and STI testing), the follow-up may be more complete than in studies designed to assess STI incidence rates in susceptible sexual minorities.However, using data from PrEP trials and cohort studies to estimate STI incidence rates has a number of important limitations.First, the majority of studies included in our meta-analysis were not designed specifically to assess STI incidence rates.We have attempted to address this limitation with our sensitivity analysis.The results of this analysis generally support the validity of our approach.A second limitation is our inclusion of data from heterogeneous study designs, as well as different regions and populations of MSM.While participants included in double-blind placebocontrolled trials will not have known their group assignment, participants in the other types of studies, particularly in PrEP cohort studies, will all have been aware that they were using PrEP.This may have affected participants' sexual risk taking behaviour and therefore their risk of acquiring STIs.We have accounted for this by building subgroups according to each of the different study designs and presenting the data separately for these where appropriate.Nevertheless, when interpreting the overall effect estimates, readers will need to bear this limitation in mind.The methodological and clinical heterogeneity is reflected by I-squared as a statistical tests for heterogeneity.This was often high, with values exceeding 75%.In a few cases, heterogeneity was substantially reduced when we analysed subgroups separately according to study design or region.While being at risk of acquiring HIV and therefore being eligible for PrEP was the common inclusion criterion for most of the studies, the actual risk behaviour of the participants may have varied between studies (and indeed over time) and may partly explain the heterogeneity of our results.Because the risk behaviour of participants in the included studies was not reported in ways that were consistent across the studies, comparing these risk behaviours is difficult and stratifying them in meaningful ways not feasible.Further factors, such as age and ethnicity of participants, may also partly explain high values of I-squared obtained for some of our pooled estimates.
If the limitations of our study are borne in mind, our estimates of STI incidence rates among MSM who are at high risk of STI infection can be helpful to policy makers and health professionals working in the fields of primary prevention and clinical care.For example, our estimates could be used to make recommendations on the frequency of STI screening and access to it in this susceptible population.Indeed, offering PrEP to this group of MSM may not only be a promising strategy to reduce the incidence of HIV.It could also help embed the target population into a comprehensive programme of frequent STI screening and treatment, potentially reducing the incidence and prevalence of STIs in the medium term, as predicted by a recent modelling study [72].

Conclusions
Despite the heterogeneous designs of the studies included in our meta-analysis and the heterogeneity of some of our results, our overall findings suggest that the incidence rates of various STIs among MSM who engage in high-risk sexual behaviour is high.The high rates of gonorrhoea, chlamydia and hepatitis C infections in the included cohort studies raise particular concerns.This subgroup of MSM can benefit from access to STI testing and treatment at close intervals.The use of PrEP is a highly effective means of preventing HIV infection and should be embedded in a comprehensive programme targeting primary and secondary prevention and treatment of other STIs.