Impact of daily, oral pre-exposure prophylaxis on the risk of bacterial sexually transmitted infections among cisgender women: a systematic review and narrative synthesis

Background: There are concerns that the use of pre-exposure prophylaxis (PrEP) may result in an increased incidence of sexually transmitted infections (STIs). Evidence for this is mixed and has mostly been based on reviews focussed on gay and bisexual men and transgender women, while none have summarised evidence in cisgender women. Methods: We conducted a systematic review to explore whether daily, oral PrEP use is associated with changes in bacterial STI occurrence (diagnoses or self-reported) and/or risk among HIV seronegative cisgender women (ciswomen). The quality of evidence was assessed using the Grading of Recommendations, Assessment, Development and Evaluation (GRADE) tool. Results: We included 11 full text articles in a narrative synthesis, with the studies published between 2012 and 2021. The studies were mostly based in Africa (n=7, 63.6%) and reported on 3168 ciswomen using PrEP aged 16–56 years. Studies had marked differences in variables, including measurements and definitions (e.g., STI type) and limited data available looking specifically at ciswomen, principally in studies with both male and female participants. The limited evidence suggests that PrEP use is not associated with increased STI rates in ciswomen generally; however, adolescent girls and young women in Sub Saharan Africa have a higher prevalence of bacterial STIs prior to PrEP initiation, compared to adult ciswomen and female sex workers. Conclusions: We suggest future PrEP research make efforts to include ciswomen as study participants and report stratified results by gender identity to provide adequate data to inform guidelines for PrEP implementation. PROSPERO registration: CRD42019130438


Introduction
The use of daily antiretroviral pre-exposure prophylaxis (PrEP) has demonstrated efficacy for the prevention of HIV transmission amongst men who have sex with men (MSM), transgender women and heterosexual couples [1][2][3][4][5][6] . PrEP can be administered as a daily, oral tablet or long-acting injection, with global estimates indicating that approximately 925,000 people were enrolled on PrEP in December 2020, with just under a quarter (22%) based in the United States.
A theoretical 'syndemic' relationship has been described between HIV prevention and other sexually transmitted infections (STIs), although the causal connection between the two remains unclear 7 . There is a perceived risk between PrEP uptake, increased condomless sex and an increased risk of STIs being reported [8][9][10] . Among PrEP users, there have been reports of actual or intended changes in risks after uptake, including increased condomless sex acts and multiple partners 11,12 , and increased STI incidence 13 . For instance, a systematic review and meta-analysis of 17 studies 14 describe a 24% increased pooled risk of any STI diagnosis following PrEP use by HIV-negative MSM and transgender women [OR, 1.24; 95% CI 0.99, 1.54; p = 0.059]. Ong et al. 15 undertook a random effects metaanalysis of any bacterial STI and reported an overall pooled incidence of 72.2 per 100 person-years (95% CI 60.5, 86.2) following PrEP initiation.
The findings from reviews are predominantly based on open-label clinical trials, limiting the application of results outside a controlled setting [16][17][18] . It is unclear what may be driving differences in risk seen among PrEP users however some theories have been postulated. Firstly, that participants of PrEP trials may be more likely to engage in 'risky' sexual practices and, subsequently, be generally more likely to acquire STIs 10,14,19 . For instance, an intended benefit of PrEP use is the freedom for people to have condomless sex, if they wish, without the risk of acquiring HIV 20 . Secondly, temporal changes suggest an increased trend in the risk of STIs amongst the general population; some argue that increased STI rates were observed before the introduction of PrEP 10 . Alternatively, some studies have reported no significant difference in either STI incidence nor risk [1][2][3]16,[21][22][23][24][25] . Overall, the evidence for any causal relationship between PrEP use and increased risk remains inconclusive 10,15,26,27 .
Reported changes in STI incidence following PrEP initiation could also be attributed to the sexual risk context. Factors such as partnership practices (e.g. multiple sexual partners), behaviours (e.g. condom use) 9 , the socio-structural context (e.g. transactional, 'survival' sex or mobility) and gender identities or relationship dynamics 9,28-30 may, in turn, directly or indirectly influence the likelihood of an individual acquiring a bacterial STI. In addition, the socio-political context and nature of the healthcare system could influence the frequency of screening. Providers' perceptions of risk profiles of patients or educational initiatives could influence STI testing decisions, introducing a detection bias of predominantly asymptomatic STIs within a particular population 31 . Thus, the association between PrEP and risk is context-specific, whether by country, setting or study design 10 ; population group 28-30 ; or the availability of other prevention methods 16,17 .
Systematic reviews and meta-analyses exploring the impact of PrEP on STI acquisition have primarily focussed on MSM and often subsume transgender women who have sex with men within analyses, report small sample sizes of ciswomen or group ciswomen with other 'non-MSM' populations 15,26,32 . This limits the generalisability and interpretation of findings due to differing psychosocial factors, including behavioural vulnerabilities 33 . The values, preferences and acceptability of PrEP among women have been previously explored, however few clinical trials have reported results specifically on ciswomen, despite accounting for almost half of new HIV infections in adults globally 19,34 . For instance, there is a well-established underrepresentation of pregnant, breastfeeding and women of childbearing potential in PrEP trials who are either explicitly

Amendments from Version 1
The minor changes in this revised manuscript are as follows: Introduction: 1. Clarifications were made in relation to describing a theoretical 'syndemic' relationship between STIs and HIV prevention and implications for any causal inferences, including a new reference added.
3. Additional information on the underrepresentation of pregnant, breastfeeding and women of childbearing potential in PrEP trials, including a new reference. 12. Reflection on limitation of our review not searching specific trial registers for those registered in Africa.
Any further responses from the reviewers can be found at the end of the article REVISED excluded in study protocols or unenrolled if a pregnancy occurs during the study, ultimately limiting the acquisition of safety data 35 . Ancillary studies of clinical trials with female participants have begun to explore the population sexual risk context and its influence on PrEP adherence 30,36-38 . One study 39 reports no evidence of any change in sexual behaviour among cisgender female participants of the Partners Demonstration Project, however this study does not report STI diagnoses.
We aim to synthesise evidence from the published literature on the association between oral pre-exposure prophylaxis and the risk of bacterial sexually transmitted infections among cisgender women (gender identity is aligned with sex assigned at birth), including the impact of PrEP on sexual behaviour.

Methods
The review is registered on PROSPERO (registration: CRD42019130438, 12 April 2019) and we use Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines to report the findings 40,41 . A protocol was developed and followed but not published 41 .

Eligibility criteria
We included open-label RCTs, demonstration and implementation projects and observational studies. We assessed each article according to our inclusion criteria as stated in our PICO statement 41 .
Ciswomen were defined to be over the age of 15 to align with UNAIDS statistics. We excluded studies if they solely related to perceptions, acceptability or willingness to take PrEP, rather than actual PrEP use or if they exclusively measured HIV acquisition during PrEP use, rather than STI, as previously described 14 . To be eligible for inclusion, studies had to include at least three-months follow-up of STI diagnoses, which could be reported as a composite measure (i.e., including non-bacterial STIs).
We included publications that were full-text, peer-reviewed journal articles, conference abstracts or grey literature published in English, with no restrictions placed on publication date, status, and geographic location.

Study selection
Two authors (VP/EC) independently used a screening tool for the first 30 articles to cross-check for consistency in the process and then independently screened titles and abstracts for eligibility. We imported all references into Covidence (RRID:SCR_016484) and conducted screening of titles, abstracts and full texts. Any reasons for excluding studies were noted and one author (VP) contacted study authors where full-text articles were unavailable. Duplicate records were also excluded during screening.
If both conference abstracts and journal articles were reported for the same study, the most recent publication was included. If the selected study was a clinical trial, the most recent publication of results linked to the register was screened. Any conflicts were discussed and resolved by consensus.

Data extraction
We used a data extraction template 41 in Microsoft Excel (Version 2112) (RRID:SCR_016137) and contacted study authors to provide a breakdown of data by gender, if the results were reported as a combined dataset (e.g., heterosexual couples). The following data were recorded: (1) study design and characteristics; (2) participant demographics and baseline characteristics; (3) key findings and outcome measures; (4) follow-up time; and (5) assessment of study quality.

Risk of bias
The quality of evidence was assessed using the Grading of Recommendations, Assessment, Development and Evaluation (GRADE) tool 41,42 taking into account study limitations (e.g., risk of bias), imprecision of the effect estimates, inconsistency of results (or heterogeneity), indirectness of evidence and risk of reporting (or publication) bias 43 . For observational studies (or non-randomised studies of interventions), grading is completed by rating upwards (starting from 'low quality') whereas randomised controlled trials have a 'high-certainty' rating to begin with 42 . Assessments range across four levels of quality: high, moderate, low and very low and indicate whether further research would "change our confidence in the estimate of effect" or an association of interest 42,43 .

Synthesis of results
We planned to report STI incidence rates with 95% confidence interval for each study presented on a forest plot. However, due to methodological heterogeneity between the studies identified, findings are presented as a narrative synthesis.

Results
Following the removal of duplicates, we screened a total of 2625 studies. Of these, 122 full-text studies were assessed for eligibility and 11 studies were included in the review 44-54 . Detail of our search is outlined in Figure 1.

Study characteristics
The 11 included studies were published between 2012 and 2021 (Table 1). Six studies were observational in design 44-47,51,54 , three were demonstration projects 48,49,53 and two were RCTs 50,52 . Sub-groups of the studies included adolescent girls and young  40 We present both search methods together as we did not record these separately when conducting our review in 2019 as it was not a requirement of previous PRISMA guidelines. Text in red indicates updated search (November 2020).
women and female sex workers. Seven of the studies were conducted in Sub Saharan Africa 44,46,47,49-52 , three in the United States 45,48,54 , and one in Taiwan 53 . In total, the studies followed 3168 ciswomen (range 11-1062 participants) enrolled to or using PrEP and aged between 16 to 56 years old, with follow-up for an average of 1.5 years. Loss to follow-up ranged from 14% to 91%. Table 2 provides a summary of baseline prevalence of STIs and reported risk after PrEP initiation. The definition of STIs across studies varied; some encompassed broader terminology of 'STI infections' and included viral types such as genital warts 44 , while others used site-specific types such as genital gonorrhoea 45 .

Synthesis of results: sexually transmitted infections
Baseline prevalence. The baseline prevalence of STI infections in five studies, using a broader definition, ranged between 6% to 47% 44,46-49 . Baseline prevalence of chlamydia was reported in six studies and ranged from 6% to 29% 45,46,49-52 . Two studies with a population of adolescent girls and young women reported a baseline prevalence of 25% and 29% respectively 49,50 . Gonorrhoea diagnoses at baseline were reported across six studies and ranged from 6% to 11% 45,46,49-52 . Syphilis diagnoses, reported in four studies, were low at baseline (prevalence ranging from 0% to 2%) 45,50,52,53 .

Follow-up prevalence.
Overall, the prevalence of diagnosed STIs reduced or remained stable during follow-up. However, there were no significant trends nor consistent methods for the measurement and reporting of STIs. Giguère et al. 46 found that STI prevalence (defined as trichomoniasis, chlamydia and gonorrhoea) declined from 15.7% (95% CI 11.8, 21.1) at baseline to 2.1% (95% CI 0.4, 10.2) at 24-months, but found no difference between baseline and 12 months, nor 12 to 24 months. Eakle et al. 44 reported bacterial STI diagnoses (including pelvic inflammatory disease, which can be used as an indicator of untreated gonorrhoea and chlamydia), at three-month intervals (up to 21 months) and also found no significant difference in diagnoses at baseline and follow-up.
Gonorrhoea diagnoses decreased during follow-up in most studies; for instance, Maljaars et al. 47 reported a 24.5% reduction in STI infections (including gonorrhoea) 12-weeks following PrEP initiation. Van Damme et al. 52 found gonorrhoea infections declined from 6.0% at baseline to 4.9% at follow-up among PrEP users (5.5% to 3.2% in placebo arm, P=0.25) and chlamydial infections from 15.1% to 13.3% (12.9% to 12.0% in placebo arm, P=0.65). However, 148 women (13.9%) were lost to follow-up 52 . One study 45 reported a small increase of genital gonorrhoea at one-year (0% to 3.3%) but this was based on a small sample (n=44), i.e., only 1.5% of the whole study were ciswomen).

Incident infections during follow-up.
Three studies calculated the incidence rate of chlamydia diagnoses at between 30 and 42 cases per 100 person-years 49-51 . In two studies of adolescent girls and young women, around two-thirds of chlamydia infections were incident, i.e., not present when PrEP started 49,50 . Furthermore, three studies that calculated the overall incidence rate of gonorrhoea infections reported these at 12-14 cases per 100 person-years 49-51 . In two studies of adolescent girls and young women, over three-quarters of gonorrhoea infections were incident 49,50 . Two studies reported subsequent syphilis diagnoses at follow-up; no syphilis was detected during follow-up by Tabidze et al. 45 and Wu et al. 53 reported one new case of syphilis among 11 ciswomen using PrEP (including on-demand or mixed dosage).

Synthesis of results: sexual behaviour
Analysis of reported behaviour showed inconsistent results, with no clear signal of increase in risk following PrEP initiation. Table 3 illustrates risks reported across the studies. Sexual partners were variably defined across the studies; some studies specified by the 'type' of sexual partner (e.g., casual, regular, main), with sex workers definitions of 'client' also specified (e.g., occasional, regular). Condom use also varied in definition across the studies in terms of 'consistency' or 'condomless sex acts.' All studies were based on self-reports; however, one study also explored the accuracy (underreporting) of reporting condomless sex by comparing self-reports against a biological assessment (vaginal swabs) which detected traces of prostate-specific antigen (PSA) and Y-chromosomal DNA Generally, consistent condom use was unchanged or improved. One study 46 reported a significant reduction in unprotected sex after 12 months, compared to baseline (27.2% to 18.1%, P=0.04). Two studies 45,47 found no change and Eakle et al. 44 had insufficient data due to loss to follow-up.

Quality assessment
Overall, the quality of evidence was very low 41 . Nine studies (81.8%) scored very low and two studies (18.2%) low quality when using the GRADE assessment. Firstly, the sample size of ciswomen using daily PrEP in four of the studies 45,47,53,54 was small (range: 3-98) which will have resulted in imprecise measurements. Secondly, high loss to follow-up (ranging from 13.9% to 90.8%) of participants was reported in most studies (n=6), introducing substantial risk of bias. Thirdly, the presentation of data limited the ability to determine any significant changes to risk. For instance, Maljaars et al. 47 did not stratify behaviour outcomes by gender, therefore it is not possible to determine whether any changes in condom use are influenced by gender.

Discussion
We found no consistent evidence that PrEP use increased the risk of ciswomen acquiring bacterial STIs, with some studies indicating that it may be associated with reduced risk. This aligns with similar findings in MSM and transgender women, described in observational studies 16,24 , demonstration projects 21,22,25 , pilot studies 23 and RCTs 1-3 . Additionally, our review found no clear evidence that PrEP use results in an increased likelihood of engaging in an action that can make ciswomen more vulnerable to acquiring an STI. We found that adolescent girls and young women in sub Saharan Africa have a high prevalence and incidence of bacterial STIs (particularly chlamydia and gonorrhoea) which is linked to higher vulnerabilities based on age-disparate sex, transactional sex, gender norms and lifetime gender-based violence 55-58 .
Increasingly diverse methods of PrEP (including delivery and dosage) provide an increasingly personalised approach to HIV prevention 7 . Our review focussed on oral PrEP however, other methods for HIV prevention include the vaginal ring (either dapivirine or tenofovir), which has been found to be an acceptable method among women, with high continuation rates seen in both observational studies and randomised controlled trials (RCTs) 59-62 as well as injectable PrEP 63 . The HPTN 084 study found 8-weekly injections of long-acting PrEP (cabotegravir) to be 89% more effective (HR 0.11; 95% CI 0.01, 0.31) among ciswomen in sub-Saharan Africa than the daily, oral PrEP (tenofovir/emtricitabine) 63 .

Strengths and limitations
To our knowledge, this is the first exploration of this research question in this population group. While the included studies had widely varying methods, they all followed participants for at least three months. We used a systematic approach to critically summarise and review the limited literature that currently exists on this topic. There are several limitations; firstly, despite an extensive literature search, we found few studies examining STI risk following PrEP initiation which contrasts with the evidence on PrEP efficacy. Our review did not include studies using other modalities of PrEP (e.g., injectable, vaginal ring); therefore, we are unable to draw conclusions on whether different modalities effect STI risk or changes in or changes in sexual behaviour and would need separate studies to review these relationships. Secondly, several of the included studies were very small and all were assessed as 'low' or 'very low' quality. We also did not search specific clinical trials registers for trials conducted in Africa (e.g., the South African Clinical Trials Register (SANCTR) or Pan African Clinical Trials Registry (PACTR)) which may have introduced a publication bias relating to population setting. Thirdly, there is marked variability in the measurement of STI outcomes (specifically timing, types, and composite endpoints) ≥4 casual sexual partners (2/11, 18.2%) * Some included studies did not provide follow-up data but are presented here to provide context when interpreting findings † Numerator backcalculated by VP ‡ Study authors contacted for further data as this was provided in a conference abstract / poster. Outcome is changes between year before and after PrEP start. § Study authors contacted for further data as this was provided in a conference abstract / poster. Data breakdown not provided by type of PrEP (n=3, daily; n=8, on-demand and mixed).

Consistent condom use at baseline
and sexual behaviour risk, which are essential to interpret patterns and causal pathways of STI acquisition to guide public health interventions. For instance, STI diagnoses varied across studies and are likely to have been influenced by attrition bias; notably in the FEM-PrEP study 52 , where less than half of participants had a pelvic examination which would then confirm the self-reported data on risk. Finally, there were high numbers of loss to follow-up (range: 14%-91%) across studies; for instance, Wu et al. 53 explained that loss to follow-up was influenced by women who could not afford or were unwilling to pay for PrEP as participants were only offered a maximum of 105 pills for a year.
Due to the heterogeneity in study designs and outcome measures, it was not appropriate to conduct a meta-analysis, restricting our synthesis to a narrative review. Just over one-quarter (n=3/11, 27.3%) of the included studies reported sexual behaviour at follow-up and baseline, therefore we were unable to measure the full extent of this potential impact.

Study implications
Our findings suggest that, like previous reviews including MSM and transgender women, there is no evidence that ciswomen using PrEP have a changed risk of bacterial STIs; however, adolescent girls and young women in sub-Saharan Africa had very high prevalence of bacterial STIs at PrEP initiation. Overall, we found very limited and low-quality evidence, making it difficult to draw any solid conclusions. This highlights an important issue of gender data biases in clinical research design, conduct and reporting, and specifically HIV prevention trials [64][65][66] . Despite this, the provision of PrEP presents an opportunity to engage women in programmes to prevent and treat STIs, providing opportunities for sexual health promotion and advice 15,67,68 . This is particularly important for those populations of adolescent girls and young women in sub-Saharan Africa with a high burden of bacterial STIs 69,70 .

Conclusions
Based on this review, there is insufficient evidence to show whether PrEP use is associated with increased STI diagnoses for ciswomen. Specifically, the quality of evidence from included studies were limited and emphasises a need for larger scale studies of cisgender women in different settings which also measure sexual behaviour including condom use and number of sexual partners at both baseline and during follow-up. We emphasise the need for larger PrEP studies with ciswomen using standard periods of follow-up that align with testing guidelines. We also suggest consistent definitions of STIs, stratification of data by gender identity and validated and standardised methods for measuring risk are used in these studies to provide more robust data to help inform PrEP implementation guidelines.

Response/revisions
We agree that a need remains to highlight this potential bias, so we have included an additional sentence in the Discussion section.

Response/revisions
We have clarified this sentence in the 'strengths and limitations' section of the Discussion to highlight that this refers to changes in sexual behaviour at baseline and follow-up.

Methods
The methods are transparent and comprehensible. It would be useful to expand on (in the 'Risk of Bias' section where GRADE is referred to) the four levels of evidence that are produced by a GRADE assessment and how these relate to the symbols in Table 1.

Results
The results are clearly presented. Some clarifications are needed. It would be useful in the 'Study characteristics' section for the studies to be referenced as they have been throughout the rest of the results section.
1. Table 1 -it would be useful to know how the reported outcomes were collectedlab/clinician confirmed/diagnoses? self-reported?
2. Table 1 -GRADE symbols need to have been described in the Methods section so that this column is interpretable to anyone not familiar with GRADE.
3. Table 2 -column % change -please add (in the footnote or text) how this calculation was made.

4.
Synthesis of results: sexual behaviour -are all these behaviours self-reported (assume so!)?this should be made clear in this section.

5.
Quality assessment. This section needs to be contextualised by outlining the scoring for GRADE as per the comment in the Methods section. A summary of how many studies scored ++, and how many scored +, would contextualise the first sentence 'Overall, the quality of evidence was low' -and (as 9/11 studies were assessed as '+' is it not more accurate to say that 'Overall, the quality of evidence was very low'? 6.

Discussion
The discussion is succinct and describes the results clearly in light of the limited evidence. A few minor details to consider: Strengths and limitations: 'Secondly, several of the included studies were very small and some of poor quality' - Table 1 would indicate that they were all of 'low' or 'very low' quality.

1.
Paragraph 4: 'Just over one-quarter of the included studies reported risk at follow-up and baseline' -it is not clear what 'risk' refers to here.

2.
Otherwise no further comments or recommendations.

Are the rationale for, and objectives of, the Systematic Review clearly stated? Yes
Are sufficient details of the methods and analysis provided to allow replication by others?
land. Paragraph 7 -acronyms for PrEP and STIs have already been defined in paragraph 1, I am not sure about the added value for defining again here. However, as in the first paragraph of the introduction, there has been a description of the long-acting injectable, cabotegravir, it would be worth clarifying that this review focuses on oral PrEP rather than injectable PrEP.

Response/revisions
To highlight that our study focusses on oral PrEP only, we have moved detail on other modalities into the Discussion and have clarified that we are unable to draw conclusions on these methods (as a limitation). We have included more information on vaginal ring trials and other modalities of PrEP in the Discussion, including 4 additional references:

Methods
The methods are transparent and comprehensible. It would be useful to expand on (in the 'Risk of Bias' section where GRADE is referred to) the four levels of evidence that are produced by a GRADE assessment and how these relate to the symbols in Table 1.

Response/revisions
We agree that additional information on how the GRADE tool is used would be helpful for readers so have included additional information about how to interpret ratings/assessment, including a new reference:

Comments: Results
The results are clearly presented. Some clarifications are needed. It would be useful in the 'Study characteristics' section for the studies to be referenced as they have been throughout the rest of the results section.
1. Table 1 -it would be useful to know how the reported outcomes were collectedlab/clinician confirmed/diagnoses? self-reported?
2. Table 1 -GRADE symbols need to have been described in the Methods section so that this column is interpretable to anyone not familiar with GRADE.
3. Table 2 -column % change -please add (in the footnote or text) how this calculation was made.

5.
We have added a sentence in the opening paragraph of the "synthesis of results: sexual behaviour" section to clarify how this was measured across studies. We have also added detail in Table 1 under outcome column.
6. We have added a sentence outlining the number of studies which scored very low and low quality with the GRADE assessment in this sub-section.

Comments: Discussion
The discussion is succinct and describes the results clearly in light of the limited evidence. A few minor details to consider: Strengths and limitations: 'Secondly, several of the included studies were very small and some of poor quality' - Table 1 would indicate that they were all of 'low' or 'very low' quality. Paragraph 4: 'Just over one-quarter of the included studies reported risk at follow-up and baseline' -it is not clear what 'risk' refers to here.