Variations in sexual network connectivity may explain dramatic variations in sexually transmitted infection prevalence between populations and over time: a selected four-country analysis

Background: The incidence of sexually transmitted infections (STIs) has been noted to vary dramatically between population groups and over time. Here, the hypothesis that changes in network connectivity underpin these changes is explored. Methods: The incidence/prevalence estimates of HIV, herpes simplex virus-2, syphilis, chlamydia, and gonorrhoea, as well as two markers of sexual network connectivity (partner concurrency and multiple partnering) by ethnic group and sexual orientation in Kenya, South Africa, the United Kingdom (UK) and the United States (USA) were extracted from published studies. Pearson’s correlation was used to test the association between the markers of network connectivity and the incidence/prevalence of these five STIs. A literature review was performed to evaluate the possible causes of the increases and decreases in syphilis incidence over the past 60 years. Results: In each country, the five STIs were found to cluster in particular ethnic groups and sexual orientations and to be positively associated with the two markers of network connectivity. Syphilis incidence in the UK and USA was found to increase dramatically in the 1960s/1970s, decline in the 1980s and again increase in the late 1990s. These changes took place predominantly in men who have sex with men, and were preceded by corresponding changes in network connectivity. The large decline in antenatal syphilis prevalence in Kenya and South Africa in the 1990s were likewise preceded by declines in network connectivity. Conclusions: Although other explanatory variables are not controlled for, the present analysis is compatible with the hypothesis that differential network connectivity is a parsimonious explanation for variations in STI incidence over time and between populations.


Introduction
The reasons underpinning the dramatic variations in sexually transmitted infection (STI) incidence between different populations are incompletely understood 1,2 . Not infrequently the incidence of specific STIs varies by an order of magnitude or more between ethnic groups and between groups defined according to sexual orientation 3-6 . Furthermore, over the past century, there have also been numerous examples of dramatic changes in the incidence of specific STIs. The incidence of syphilis in men who have sex with men (MSM) in the United States (USA), for example, was high in the 1970s, but declined dramatically in the early 1980s only to return to 1970s levels more recently (Figure 1) 7 . In this paper we test the hypothesis that differences in the connectivity of local sexual networks play an important role explaining these cross sectional and longitudinal variations in STI incidence 8,9 . Although the connectivity of local sexual networks is determined by a number of parameters, we focus on two of the most important -the rate of partner change and the proportion of partnerships that run concurrently 8,9 . Previous analyses have typically found that one, or a combination of markers of network connectivity, are raised in populations with a higher prevalence of a particular STI 3,4,9,10 . These studies have however typically been limited to single countries and frequently only investigated variations in a single STI and either considered variations between ethnic groups or sexual orientations but not both [11][12][13][14][15][16][17][18][19][20] . Global country-level analyses have produced a mix bag of results with a small number of studies 21,22 finding positive associations between markers of network connectivity and STI prevalence, and a larger number finding no association [23][24][25][26] . A possible explanation for this discrepancy is a type of misclassification bias that could be termed a 'fruit salad bias' 22,27 . It is increasingly appreciated that each country is made up of a number of sexual networks (defined by sexual orientation, ethnicity, class, religion etc.) that interconnect to varying degrees 2,4,9,28 . If the connectivity of these sexual networks is a major determinant of STI prevalence, then it would be most appropriate to compare the association between network connectivity and STI prevalence between sub-populations of sexual networks within a country rather than combining them all into one national population (fruit salad) as a number of studies have done 10,24,25 .
These insights provided the motivation for the current paper where we aimed to conduct a more systematic analysis of the association between network connectivity and STI prevalence in national subpopulations. The current study expands the previously conducted analyses of this association in four ways. Firstly, this association was assessed in four countries: two high income countries (the United Kingdom [UK] and [USA]) where STIs are predominantly concentrated in MSM and certain ethnic groups and two low/middle income countries that have been affected by generalized HIV epidemics (Kenya and South Africa) but where certain ethnic groups have largely escaped this epidemic. These countries were chosen due to

Amendments from Version 1
All of the criticisms made by the two reviewers have been addressed in the new version. These include a change to the title of the article, changes to the introduction which detail how the four countries used in the study were selected, more details in the methods (including more clearly defined terms), a number of additional limitations added to the discussion and the inclusion of importance of development of STI vaccines to control STI spread.
the familiarity of the author with these countries and the availability of nationally representative datasets with both sexual behaviour and STI prevalence data. Secondly, in each country we assess this association at the levels of both ethnic groups and MSM versus heterosexuals. Thirdly, we assess this association in five STIs (chlamydia, gonorrhoea, HIV, herpes simplex virus-2 [HSV-2] and syphilis). Fourthly, we conduct longitudinal analyses over the past 60 years to assess if increases and decreases in syphilis incidence and prevalence are preceded by corresponding changes in network connectivity.
Cross sectional STI data. The incidence and prevalence estimates of chlamydia, gonorrhoea, HIV, HSV-2, syphilis, concurrency and multiple partnering in Kenya, South Africa, the UK, and USA were extracted from published studies, as detailed in Table 1 and Table 2. When two or more studies were found, we favoured data from studies based on nationally representative samples. Where this data was not available other sources were used such as detailed in Table 2. In addition, for the cross-sectional analyses, we limited studies to those based on data from after the year 1990 and favoured the most recent data.
Symptomatic STI: In the case of Kenya and South Africa, we could not find high quality data on STI prevalence, or behaviour reported by sexual orientation. In addition, we could not find reasonable quality data reporting the incidence/prevalence of chlamydia or gonorrhoea by ethnic group in these two countries. As a result, we use the proportion of men reporting urethral discharge as a composite proxy for combined chlamydia plus gonorrhoea incidence per ethnic group 29,30 .
Longitudinal syphilis data. We chose syphilis for the evaluation of longitudinal changes in STI prevalence for a number of reasons, including the fact that syphilis surveillance programmes from the four countries have provided reasonable quality The prevalence of concurrency and multiple partners were taken from the 1992 National Health and Social Life Survey (NHSLS). This was a cross-sectional study that used a nationally representative stratified random sample of 3 432 women and men between the ages of 18 and 59.
Abbreviations: STI -sexually transmitted disease, DHS -Demographic and Health Survey, NATSAL -National Survey of Sexual Attitudes and Lifestyles, NHANES -National Health and Nutrition Examination Surveys.
estimates of syphilis incidence for the last 60 years and longer -a far longer period than that for which reasonable incidence estimates are available for other STIs 45 . To a large extent, this is related to the fact that relatively accurate serological tests have been available for the diagnosis of syphilis for longer than other STIs and these have been extensively used to diagnose antenatal syphilis 45-47 . Antenatal syphilis surveys have been shown to produce reasonable estimates of the prevalence of syphilis in the general population 45-47 . For Kenya and South Africa, we used large antenatal syphilis prevalence surveysnational in the case of South Africa and limited to the capital in the case of Kenya (Table 1). For the UK and USA, we used incidence estimates based on national case reporting data (Table 1). In the UK, the sexual orientation of individuals with syphilis was not available for much of this period and thus we follow the example of others who compare the ratio of the incidence in men to that in women as a proxy for the incidence in MSM 1,48,49 .
Multiple partners and concurrency. Studies use a wide variety of definitions of multiple partnering and concurrency.
Concurrency: In all the comparisons of ethnic groups, excluding the UK, the prevalence of concurrency was defined as the percentage of men (15-49 years old) who reported having two or more sex partners at the time of the survey. In the UK, concurrency referred to the percentage of men who reported any overlapping sexual relationships in the past year 39 . In the comparisons of sexual orientations, for the USA, concurrency was defined as the respondent reporting two or more sexual partnerships with overlapping dates in the prior year 50 . As far as the UK was concerned, this referred to the proportion reporting overlapping sexual partnerships at any point in the past 5 years 51 .
Multiple partners: For the ethnic group comparisons, the prevalence of multiple partners was defined as the percentage of men (15-49 years old) who reported having two or more The incidence figures for these three STIs were taken from Public Health England's report 'Sexually transmitted infections and screening for chlamydia in England, 2018' 38 . This report provides incidence estimates for chlamydia, gonorrhoea and syphilis for heterosexual men, heterosexual women, HIV-infected MSM and 'HIV-uninfected or HIV-undiagnosed MSM' in 2018. We report the results for the first and last two of these categories. There was little difference in the incidence estimates for heterosexual men and women 38 .

Concurrency, multiple partners 2012 51
The third British National Survey of Sexual Attitudes and Lifestyles (NATSAL 3) was a nationally representative sample of 15,162 men and women aged 16-74 years conducted in 2012 51 . We extracted the relevant variables from a study report which provided prevalence estimates of sexual behaviour for MSM and heterosexual men. The multiple partnering variable referred to the mean lifetime number of sexual partners reported by respondents. The proportion reporting overlapping sexual partnerships at any point in the past 5 years was used as the measure of concurrency.
sex partners in the last year. In the UK, this variable was not reported by ethnic group. As a result, multiple partnering here refers to the percentage of men who reported one or more new heterosexual relationship in the prior year. For the MSM vs. heterosexual comparisons, this variable referred to the mean lifetime number of sexual partners reported by respondents in the UK (aged 16-74) and the USA (aged 35-39).
These variations in definitions between studies mean it would be inappropriate to compare these variables between studies.
Since all the studies we used were limited to specific countries, we only compare subpopulations within each study/country.

Regions vs. ethnic group in Kenya.
Many of the national STI surveillance surveys in Kenya do not collect data on ethnic group 31 . There is however a high correlation between region of residence and ethnic group, with large differences in STI incidence and sexual behaviour between regions 5,32 . We therefore follow the approach used in previous studies and use region of residence as a proxy for ethnic group 32,54 .

Stratification of data by ethnic group:
Each study reported the variables under consideration stratified by ethnic group. We used this data for our analyses.

Statistical analysis
We report the incidence/prevalence of the five STIs, multiple partnering and concurrency by ethnic group (all four countries) and MSM vs. heterosexuals in each country with available data (UK and USA). Where this data is provided, we report 95% confidence intervals. We used the Chi-squared test to assess if there was a difference in incidence/prevalence between the subgroups with the highest and lowest incidence/ prevalence values in each country. For continuous variables (lifetime number of partners) the Man-Whitney U test was used.
The correlation between the prevalence of the different STIs per subgroup in each country, and between the prevalence of each STI and the two behavioural variables was calculated using Pearson's correlation. Because of the small sample sizes (between 2 and 8), tests of statistical significance were not performed for these calculations. In the case of UK and USA, the STI data is presented as an estimated incidence per 100 000 per year without reporting the number of cases and the total size of the study populations. As a result, we did not test if the association between the prevalence of each STI and each of the markers of network connectivity is statistically significant. The key differences were however large, which allows substantive inferences to be drawn. Where data was only provided in graphical form, we digitized the data using GetData Graph Digitizer version 2.26. STATA v15.2 was used for all analyses (StataCorp LLC, College Station, TX, USA).
Longitudinal association between network connectivity and syphilis prevalence/incidence A literature review was performed to evaluate the possible causes of increases and decreases in syphilis incidence in each subpopulation. We used a narrative approach here to evaluate if there was evidence from the literature that increases and decreases in syphilis prevalence were preceded by corresponding changes in markers of network connectivity. The approach adopted was not to perform a systematic review or a quantitative analysis, but rather to assess if there is evidence in the literature that corroborates or refutes the network connectivity hypothesis.

Terminology
Sex, gender, ethnicity and race: Where this was specified, each data source used a slightly different definition of sex, gender, ethnicity and race. Wherever possible, we used the same terms used in these original sources.
The region with the lowest HIV prevalence (the North East) was not included in the behavioural survey that provided our data. Other data sources have however found the rate of multiple partnering in this region to be very low 5 . In our data, the region with the second lowest HIV prevalence (Eastern) had the lowest reported prevalence of multiple partnering (12.2%; 95% CI 7.4-19.5%) and concurrency (6.4%; 95% CI 4.2-9.2%; all P<0.001). Nyanza reported the highest prevalence of multiple partnering (30.6%; 95% CI 23.8-38.3%) and the second highest prevalence of concurrency (18.2%; 95% CI 14.9-21.9%).
With the exception of syphilis, which was negatively associated with concurrency prevalence (r--0.34), the prevalence of each STI was positively correlated with both the multiple partnering (r-0.38 to r-0.84) and concurrency (r-0.42 to r-0.72) variables.

Ethnic group
HIV prevalence varied 8-fold between 0.2% (95% CI 0.2-0.2%) in the non-Hispanic White group and 1.7% (95% CI 1.6-1.8%) in the non-Hispanic Black group ( Figure 2, Table 4). The incidence of syphilis varied 4-fold between 5 and 24 per 100 000 per year in these same two groups. Likewise, the incidence of chlamydia varied 5-fold and that of gonorrhoea 8-fold between the same groups (all P<0.001). The prevalence/incidence of chlamydia, gonorrhoea, HIV, HSV-2 and syphilis by ethnic group were strongly positively correlated with each other (r-0.97 to r-1.00; Table 6). The same is true for syphilis in Kenya and South Africa. In the USA and the UK, chlamydia, gonorrhoea and syphilis are reported as cases per 100 000 per year (Point estimates with 95% Confidence Intervals; see Table 1 for sources of data).

United Kingdom
Ethnic group HIV prevalence varied between 2.8% in the Black Caribbean and Black African groups and 0% in the White and Asian groups (Table 4, Figure 2). The incidence of chlamydia varied 11-fold between the Black Caribbean (1540/100 000/year) and Asian (136/100 000/year) groups. Gonorrhoea incidence varied 8-fold between these same groups (840 and 109/100 000/year, respectively). Likewise, the incidence of syphilis varied 4-fold between 71 and 18 per 100 000 per year in these same two groups. The prevalence/incidence of chlamydia, gonorrhoea, HIV and syphilis by ethnic group were positively correlated with each other (r-0.71 to r-0.98; Table 6).
The prevalence of concurrency and multiple partnering were approximately 1.5 to 2.5-fold higher in the Black African and Black Caribbean groups than the Asian and White groups. The prevalence of each STI was positively correlated with both the multiple partnering (r-0.63 to r-0.96) and concurrency (r-0.37 to r-0.91) variables.
Longitudinal analyses UK and USA. In both the UK and the USA, the incidence of syphilis increased in the 1960s and 1970s but this increase was largely limited to MSM (Figure 1) 1,7,49 . The incidence remained fairly stable in heterosexuals. Although quantitative data is very limited from this period, a number of analyses attributed this increased incidence in MSM to behavioural factors, such as increases in rates of partner change 7,49,55,56 .
In the early 1980s, incidence plummeted in MSM in both countries 1,7 . The AIDS epidemic played a large role in this decline. Individuals most centrally placed in sexual networks were more likely to die from AIDs, which resulted in a fragmentation of sexual networks 7,57 . So too reductions in multiple partnering and increases in condom use both served to reduce effective network connectivity 49,58 . From around 2000, the incidence of syphilis has been increasing in both countries and this increase has been largely limited to MSM 1,7 . A very similar trend has been evident for MSM in the UK for chlamydia and gonorrhoea where reasonable surveillance data is available and is likely the case in the USA where surveillance data is less detailed as regards sexual orientation 1,44 . In both countries, numerous lines of evidence suggest that the introduction of effective antiretroviral therapy from 1996 onwards together with other factors resulted in increases in sexual network connectivity, which was in turn responsible for these increases in STI incidence 1,48 . Reduced deaths from AIDS, increases in rates of partner change and reductions in condom usage all played a role in this reconstitution of network connectivity 7,48,58,59-61 .
Kenya and South Africa. The prevalence of antenatal syphilis in both Kenya and South Africa was between 5 and 12% in the pre-HIV period (Figure 4; Table 3 and Table 4). A number of analyses attributed this high prevalence of syphilis to a number of factors that included behaviours that translate into dense sexual networks 45,62,63 . Syphilis prevalence declined rapidly following the AIDS epidemic to below 1% in both countries 62 . Similar, contemporaneous declines in diagnoses of cases of syphilis took place in these countries 45,64 . Using a range of data sources and study methodologies, a number of analyses have concluded that a combination of connectivity-reducing factors was responsible: reductions in multiple partnering (number of partners and concurrency), delayed sexual debut, increased condom usage and the effect of AIDS mortality removing more centrally placed transmission nodes 45,62,65-67 .

Discussion
Our analysis confirmed previous analyses of dramatic variations in STI prevalence between populations 3-6 . A strong tendency  Table 1 for sources of data).
towards clustering of STIs within subpopulations was evident. Furthermore, the cross-sectional variations in STI prevalence were almost all positively associated with two markers of network connectivity. In a similar vein, dramatic increases and declines in syphilis incidence could also be explained by corresponding changes in network connectivity. Taken together these findings suggest that network connectivity could represent a parsimonious explanation for both variations in STI incidence between populations and over time.
There are however a number of caveats to bear in mind. Our study was not based on a systematic review of all possible studies. In fact, the study was limited to an opportunity sample of only 4 out of over 200 countries, and the data it used was limited in a number of ways. Not all the data was from nationally representative samples. The data also came from different and limited time periods. The way the behavioural variables were defined varied between studies and in the case of Kenya, region was used as a proxy for ethnic group. For the sake of simplicity, we focused on male sexual behaviours. On the one hand, our analysis is purely ecological and therefore susceptible to the ecological inference fallacy. On the other hand, the prevalence of STIs is to a large extent a function of network level properties. Networks are inherently ecological level entities and thus ecological analyses are appropriate and necessary forms of analysis. We assume that there is a degree of segregation of sexual network by ethnic group or sexual orientation. A high degree of sexual partner homophily by ethnic group has indeed been shown for Kenya, South Africa and the USA 4,5,28,69,70 . We could find no analyses investigating this   USA 1963 to 2013 7 The primary and secondary syphilis incidence estimates for 1963 to 2013 (reported as cases per 100 000 population) are taken from a paper that estimated incidence in MSM and heterosexual men based on cases of primary and secondary syphilis reported to the CDC in this time period 7 .  The definitions used for STI, symptoms, concurrency and multiple partners varied between studies -see Table 1 for details  type of homophily in the UK. If there was little or none of this type of homophily, this would however result in a homogenization of STI incidence between ethnic groups and would therefore be expected to reduce the strength of the association between STIs and behaviours by ethnic group.
Not all the associations between multiple partnering/ concurrency and STI incidence/prevalence were strongly positive. In fact, one association (between syphilis and concurrency by ethnic group in Kenya) was weakly negative (Table 6). A further crucial weakness is that we did not control for potential confounders. The prevalence of circumcision and condom use are two examples of factors that could confound our analyses. Other papers have, however, evaluated this question and concluded that these are likely to be only one of many determinants of differential STI prevalence 3,4,32,65 . For example, in South Africa, the highest STI prevalence ethnic group has been noted to have considerably higher circumcision rates and use of condoms than the two lower STI prevalence populations 4,65 . Likewise, rates of condom usage are typically higher in MSM than heterosexuals 50 and as far as we can ascertain, the prevalence of circumcision does not differ significantly between these two populations. A further limitation is that we have not controlled for other differences between MSM and heterosexual sexual networks such as the impact of roleversatility in MSM 71 . These limitations mean that we cannot conclude how generalizable our findings are or even if the associations we found are causal. What we can conclude is that our results are compatible with the network connectivity explanation. We cannot exclude the possibility that other unmeasured variables are the predominant cause.
Our findings are however logical if we recall that the number of sexual contacts per unit time is a key determinant of the rate of spread of an STI 2 . So too increases in concurrency have been shown to lead to non-linear increases in sexual network connectivity 9,72 . A number of mathematical modelling studies from the USA and South Africa have established that the actual self-reported differences in number of partners and concurrency between ethnic groups in the USA and South Africa were able to explain a large part of the differential spread of HIV in these populations 3,65 .
A strength of our analysis is its finding that the association between network connectivity and STI prevalence held in four different countries for both comparisons by ethnic group and MSM versus heterosexuals. In the case of Kenya, all the ethnic groups/regions, including those with very low STI rates, were 'black' Africans. The association was found for five different STIs and in the case of syphilis, was able to explain longitudinal changes in incidence. These findings do not fit with racist theories that biological differences between races explain differential STI spread 73 . They fit better with sociological explanations that differences in sexual behaviours result in differences in network connectivity and consequent STI incidence 9,74 . Furthermore, they suggest that behaviour change is possible and can result in rapid reductions in STI incidence. In fact, if network connectivity is a key determinant of STI incidence then it follows that reducing network connectivity would be an optimal way to effect radical STI prevention. Whilst high STI prevalence populations would first need to decide if they wanted to change behaviours that enhance connectivity, there are a number of positive precedents for this. Uganda's 'Zero grazing' campaign was a national campaign to reduce multiple partnering in order to reduce the risk of HIV. This campaign was credited with large scale reductions in multiple partnering and a subsequent decline in HIV prevalence 75 . Similar campaigns amongst MSM in the 1980s had a similar effect 7,57 .  These insights are relevant in the current setting of increasing incidence of numerous STIs in high income countries 1 . A number of analyses have found that a large proportion of these increases are occurring in MSM with high rates of partner change and particularly in contemporary HIV pre exposure prophylaxis (PrEP) recipients 1,38,76 . It is not unusual for PrEP recipients to report a median of 12 partners (IQR 6-25) per 3 months 77 . Unsurprisingly these translate into very high STI incidence -the combined incidence of gonorrhoea, chlamydia and syphilis can be up to 169 per 100 person years 78 . Tackling this high incidence via intensive screening and treating without addressing the underlying cause may result in antimicrobial resistance 79 .
Further studies are required to further elucidate the links between norms, structural factors, behaviour, sexual networks and STI incidence. In the interim, the type of evidence reviewed here suggests that reducing STI incidence/prevalence in high prevalence populations to those of low prevalence population may be difficult without reducing network connectivity and/or without vaccination where vaccines are available.

Data availability
Underlying data All data underlying the results are available as part of the article and no additional source data are required (see Table 1- Table 3 and Table 7).
Page 15 of 28 28. Laumann EO: The social organization of sexuality: Sexual practices in the I confirm that I have read this submission and believe that I have an appropriate level of expertise to confirm that it is of an acceptable scientific standard.

Version 1
Reviewer Itzchak Levy 1 Tel-Aviv University, Tel-Aviv, Israel 2 Sheba Medical Centre, Tel-Aviv, Israel The study done by Chris Kenyon is very interesting and important. It tries to show the association between the connectivity within a group (which in this work is expressed by the number of partners in a certain time and the concurrency of partners) and the incidence/prevalence of certain sexually transmitted infections (STIs). This is important because understanding sexual networks and how they work and what influences them is very important in the prevention of STIs. It adds to the knowledge we get from the study of social and sexual networks as well to phylogenetic studies of the spread of STIs. However, since this paper is constructed from many different works that studied different questions (incidence of some STIs, prevalence of STIs, sexual behavior data, etc.) as a clinician I don't understand how the conclusions of the writer can be made and I feel that the input of a statistician would be much more helpful. It sounds more reasonable to take one or two STIs which are compatible with a dynamic study (e.g. gonorrhea and chlamydia) and to concentrate on a more defined population. However, this study still gives the "wide picture". HIV in my opinion is not suitable for this study because its prevalence is influenced not only by the incidence but also by some other factors such as the introduction of anti-retroviral therapy (ART) in 1996 and the introduction of better treatments like integrase inhibitors in the 2000s. Incidence may be also influenced by introduction of earlier treatment (treatment as prevention) and introduction of pre exposure prophylaxis (PREP) so it is also less suitable for this study. The methods section should be somewhat explained better. It is not understandable if the work is a systematic review of the literature, then it would be helpful if you can explain the method by which the specific studies were chosen and why others were not (a flow chat may be helpful).
There is a mixture of studies, some show incidence and some show prevalence -this mixture is confusing.
With such a big population I don't quite understand what is the meaning of the correlation.
There are a lot of prevalence studies which utilize serology like studies 42 and 43 (HSV2 and syphilis). There is no way to know the date of infection in those studies so it is a very big limitation to use them for the purpose of this study which tries to show dynamics.
In Kenya (page 6) ethnicity was a region of residence served as a proxy for ethnicity. I am not sure that this approximation can be done safely for the purpose of this study because it is well established that MSM tend to reside in big cities without regard to their ethnicity so I would leave it as "regional". In Figure 2 (page 7) The rate of MSM in each ethnic group is not given and this may influence the data and the conclusions. In Figure 4 legend (page 9) the word the united states should be omitted.

Is the work clearly and accurately presented and does it cite the current literature? Partly
Is the study design appropriate and is the work technically sound? Partly

Are sufficient details of methods and analysis provided to allow replication by others? No
If applicable, is the statistical analysis and its interpretation appropriate? I cannot comment. A qualified statistician is required.
However, since this paper is constructed from many different works that studied different questions (incidence of some STIs, prevalence of STIs, sexual behavior data, etc.) as a clinician I don't understand how the conclusions of the writer can be made and I feel that the input of a statistician would be much more helpful. It sounds more reasonable to take one or two STIs which are compatible with a dynamic study (e.g. gonorrhea and chlamydia) and to concentrate on a more defined population. However, this study still gives the "wide picture". HIV in my opinion is not suitable for this study because its prevalence is influenced not only by the incidence but also by some other factors such as the introduction of anti-retroviral therapy (ART) in 1996 and the introduction of better treatments like integrase inhibitors in the 2000s. Incidence may be also influenced by introduction of earlier treatment (treatment as prevention) and introduction of pre exposure prophylaxis (PREP) so it is also less suitable for this study.

Reply:
Thank you for these useful comments. The limitations section has been expanded to include the concerns raised by the reviewer (P 23, second paragraph): Differential uptake of antiretroviral therapy is an important potential determinant of differential HIV prevalence between sub-populations. Once again, we did not control for this.
It should be noted that in the case of South Africa the differential HIV prevalence between ethnic groups was of a similar magnitude in surveys done before ART became available to those following the widespread use of ART.
The methods section should be somewhat explained better. It is not understandable if the work is a systematic review of the literature, then it would be helpful if you can explain the method by which the specific studies were chosen and why others were not (a flow chat may be helpful).

Reply:
As noted above the countries were not selected via a systematic review methodology. The rationale for selecting the studies has been added to the introduction. This limitation has been added to the relevant section of the limitations in the discussion (P 23, 3 rd Paragraph).
In addition, the following sentence has been added to the introduction to make it clear how the four countries were chosen (P3, third last sentence).
These countries were chosen due to the familiarity of the author with these countries and the availability of nationally representative datasets with both sexual behaviour and STI prevalence data.
Finally, we have changed the title of the paper from "a four-country analysis" to "a selected four-country analysis." There is a mixture of studies, some show incidence and some show prevalence -this mixture is confusing. With such a big population I don't quite understand what is the meaning of the correlation.
There are a lot of prevalence studies which utilize serology like studies 42 and 43 (HSV2 and syphilis). There is no way to know the date of infection in those studies so it is a very big limitation to use them for the purpose of this study which tries to show dynamics.

Reply:
It is true that one cannot know the date of infection in studies that define STI incidence/prevalence with HIV, HSV-2 serology or a treponemal test only in the case of syphilis. These variables are then best considered as a measure of cumulative incidence. A number of studies have found that these variables offer a valuable marker of how connected sexual networks are and what their risk is for other STIs (1, 2). As an example, the prevalence of HSV-2 in 20-24 year olds can predict subsequent HIV prevalence (2). The same might be true for emergent STIs such as monkeypox. For most of the studies of syphilis incidence/prevalence used in this paper, syphilis was defined by both a treponemal and non-treponemal test.
In Kenya (page 6) ethnicity was a region of residence served as a proxy for ethnicity. I am not sure that this approximation can be done safely for the purpose of this study because it is well established that MSM tend to reside in big cities without regard to their ethnicity so I would leave it as "regional". In Figure 2 (page 7) The rate of MSM in each ethnic group is not given and this may influence the data and the conclusions.

Reply:
We provide a justification for the use of regions as a proxy for ethnicity in Kenya in the Methods section (P 11, third paragraph): In addition, we have added this as a limitation in the discussion section (P 23, second paragraph): The way the behavioural variables were defined varied between studies and in the case of Kenya, region was used as a proxy for ethnic group

Regions vs. ethnic group in Kenya.
Many of the national STI surveillance surveys in Kenya do not collect data on ethnic group 29 . There is however a high correlation between region of residence and ethnic group, with large differences in STI incidence and sexual behaviour between regions 5,31 . We therefore follow the approach used in previous studies and use region of residence as a proxy for ethnic group 31, 54 .
In Figure 4 legend (page 9) the word the united states should be omitted.

Reply:
Thanks for pointing out this error, which has been corrected.
The author uses the terms "heterosexual men" and "men who have sex with women" ( Table  2). I think it would be helpful to choose one throughout. There are also places where the author states "heterosexuals" which I assume to be men only but it is a little unclear 5.
There appears to be a typo in the Figure 4 title (the phrase "the United States" should not be there) 6.

Major comments:
The paper is not described as a systematic review, but was it intended to be? Given that there are summary quantitative estimates provided, it feels like a systematic review. If so, I think it would have been helpful in the results to see how many papers were identified using the search criteria and how many were eliminated, and for what reasons. When I ran this search (using "or" between the terms but also making sure the countries were included) I get >225,000 results. I assume the author used a different order of terms. Could the full order of search terms be included in the paper? 1.
Somewhat related to the above, I am having a bit of trouble figuring out how studies were selected. For example, it appears that there were only 3 studies from Kenya published since 1990 that make comparisons on the prevalence of STIs/HIV, multiple partners, or concurrency. Is that the case? Or were these the only population-based studies? I think adding in inclusion/exclusion criteria would be helpful.

2.
To calculate chi-square values for incidence/prevalence, did the author pool raw numbers from different studies or use some other method? Would there be potential for violation of an assumption of independence of observations?
3. The discussion focuses on the relationship between STIs and concurrency. Unless I misunderstand the data, it seems that these data are mostly from Table 6, which shows the correlation of STIs and concurrency and multiple partners. To me, it does not look like the associations are particularly strong in Kenya, and it does not really hold for syphilis in the UK or gonorrhea in the UK. Are there other data from which the author is drawing to make these conclusions?
In the conclusion, I think it may be worth adding a comment about investment in STI vaccines being a plausible solution to curbing STI rates. Probably a more viable solution than behavior change. 8.

Is the work clearly and accurately presented and does it cite the current literature? Yes
Is the study design appropriate and is the work technically sound? Partly

Are sufficient details of methods and analysis provided to allow replication by others? No
If applicable, is the statistical analysis and its interpretation appropriate? I cannot comment. A qualified statistician is required.
Are all the source data underlying the results available to ensure full reproducibility? Yes

Are the conclusions drawn adequately supported by the results? Partly
Competing Interests: No competing interests were disclosed.
Reviewer Expertise: STI/HIV epidemiology I confirm that I have read this submission and believe that I have an appropriate level of expertise to confirm that it is of an acceptable scientific standard, however I have significant reservations, as outlined above.
Author Response 24 Aug 2022 Chris Kenyon, Institute of Tropical Medicine, Antwerp, Antwerp, Belgium Reply to reviewer 1: 1 In this review, Dr. Kenyon reviews data on STI/HIV incidence/prevalence and partner concurrency from (mostly) population-based studies in four counties. The results show that STIs cluster within certain groups defined by race/ethnicity and gender of sex partners, and that STIs were correlated with number of sex partners and concurrency. It is helpful to have these data summarized for these countries. The author appropriately addresses the limitations. My main comment has to do with the transparency in how these papers were selected. I think adding in the full search criteria, the full exclusion/inclusion criteria, the number of studies identified and eliminated, etc would be appropriate here. I also had trouble identifying which data supported which conclusions. These and other comments are detailed below.
Reply: Thanks for pointing this error out. It has been corrected.

Major comments:
The paper is not described as a systematic review, but was it intended to be? Given that there are summary quantitative estimates provided, it feels like a systematic review. If so, I think it would have been helpful in the results to see how many papers were identified using the search criteria and how many were eliminated, and for what reasons. When I ran this search (using "or" between the terms but also making sure the countries were included) I get >225,000 results. I assume the author used a different order of terms. Could the full order of search terms be included in the paper? ○ Reply: As noted above the countries were not selected via a systematic review methodology. The rationale for selecting the studies has been added to the introduction. This limitation has been added to the relevant section of the discussion (P 23, 3 rd Paragraph). Somewhat related to the above, I am having a bit of trouble figuring out how studies were selected. For example, it appears that there were only 3 studies from Kenya published since 1990 that make comparisons on the prevalence of STIs/HIV, multiple partners, or concurrency. Is that the case? Or were these the only population-based studies? I think adding in inclusion/exclusion criteria would be helpful.

Reply:
Where possible surveys using nationally representative samples were used. Where this data was not available other sources were used such as detailed in Table 2. This information has been added to the methods section.
To calculate chi-square values for incidence/prevalence, did the author pool raw numbers from different studies or use some other method? Would there be potential for violation of an assumption of independence of observations? ○ Reply: Chi-squared tests were performed in the raw numbers per group. It is acknowledged that this approach is of limited utility and if the reviewer prefers, these statistical tests could be removed. The discussion focuses on the relationship between STIs and concurrency. Unless I misunderstand the data, it seems that these data are mostly from Table 6, which shows the correlation of STIs and concurrency and multiple partners. To me, it does not look like the associations are particularly strong in Kenya, and it does not really