Circumcision and risk of sexually transmissible infections in a community-based cohort of HIV-negative homosexual men in Sydney, Australia

Abstract

Background

Circumcision status was examined as an independent risk factor for STIs in the HIM cohort of homosexual men in Sydney.

Methods

From 2001–2004, 1,427 initially HIV-negative men were enrolled and followed to mid-2007. All participants were offered annual STI testing. Past history of STIs was collected at baseline and information on sexual risk behaviors, every 6 months. At annual face-to-face visits, participants reported STI diagnoses made in the previous year.

Results

Circumcision was not associated with prevalent or incident HSV-1, HSV-2, or self-reported genital warts. There was also no independent association of circumcision with incident urethral gonorrhea or chlamydia. Being circumcised was associated with a significantly reduced risk of incident (HR 0.35, 95% CI 0.15–0.84), but not prevalent (OR 0.71, 95% CI 0.35–1.44) syphilis. The association was somewhat stronger among men who reported predominantly insertive unprotected anal intercourse (HR 0.10, 95% CI 0.01–0.82).

Conclusions

These are the first prospective data in homosexual men to assess circumcision status as a risk factor for STIs. Circumcised men were at reduced risk of incident syphilis but no other prevalent or incident STIs. Circumcision is unlikely to have a substantial public health impact in reducing acquisition of most STIs in homosexual men.

INTRODUCTION

Men who have sex with men (MSM) bear a disproportionate burden of sexually transmissible infections (STIs) in resource-rich countries. A substantial decline in rates of many STIs occurred in MSM soon after HIV/AIDS was recognized, but many countries have witnessed a resurgence of STIs among this population [1]. Traditional methods of STI control are clearly failing to control the STI epidemic in MSM and novel strategies to reduce STI transmission require investigation.

Observational evidence among heterosexual men suggests that circumcision may reduce the risk of syphilis, chancroid and possibly HSV-2 [2]. Recently published data from African randomized trials have reported that circumcised heterosexual men are at reduced risk of HSV-2 acquisition [3, 4], but not syphilis infection [3]. Compared with uncircumcised men, circumcision also resulted in a lower prevalence of high-risk penile human papilloma virus (HPV) [3, 5] but had no impact on incidence [6] or prevalence [7] of urethral gonorrhea or chlamydia. As the range of sexual behaviors of homosexual men is substantially different from those of heterosexual men, it is likely that these results cannot be generalized to homosexual men.

The small number of studies performed among MSM, none of which was prospective, have described little impact of circumcision on selected STIs [8–10] or STIs overall [11]. The association of circumcision status and STIs was assessed in the community-based Health in Men (HIM) cohort of HIV-negative homosexual men in Sydney, in which the incidence and other risk factors for a wide range of anogenital STIs had already been well-characterized [12–15].

METHODS

Participants

Study participants were men who were recruited from a range of community-based sources between June 2001 and December 2004, and followed until June 2007. Details of recruitment sources and methods of the HIM study have been described elsewhere [15]. Briefly, men eligible for participation in the HIM study met the following criteria: (1) they reported having sex with men in the 5 years prior to enrolment, (2) they lived in, and/or participated in the gay community, of Sydney, and (3) they tested negative for HIV at baseline. Signed, informed consent was obtained from all participants. Ethics approval was granted by the University of New South Wales.

Data Collection

All participants underwent annual face-to-face interviews, with six-monthly telephone interviews between these visits. Detailed behavioral data were collected every six months including number and position (insertive/receptive) of unprotected anal intercourse (UAI) acts. In addition, each year participants reported their preference for anal intercourse position (preference for insertive role, receptive role or no preference). Participants were asked to report if they had been diagnosed with anogenital gonorrhea, chlamydia or warts in the last twelve months at the annual face-to-face interviews. They were also asked if they had been diagnosed with anogenital warts in their lifetime at the baseline interview. As part of the study design, participants were offered annual testing for STIs at the time of the face-to-face interview. Circumcision status was reported by participants at baseline. Self-report was validated by clinical examination in a subgroup of 237 participants, and correlated with examination findings in 98% of cases [16].

Laboratory Studies

Detailed information on specimen collection and laboratory diagnostic methods have previously been published [12–15, 17] and are briefly summarized below.

Neisseria gonorrhoeae and Chlamydia trachomatis

Nucleic acid amplification tests (NAAT) for N. gonorrhoeae and C. trachomatis were introduced to the HIM study in January 2003. Consenting participants collected first-void urine and self-obtained anal swab samples annually for NAAT testing by the BD ProbeTec assay (BD Diagnostics, Sparks, MD). For anal N. gonorrhoeae diagnosis, all ProbeTec positive results underwent supplementary porA testing. The full methods of this supplementary testing have been previously described in detail [17]. Only porA positive samples were included as true anal N. gonorrhoeae infections in the analyses.

Syphilis

Consenting participants were screened annually for syphilis by enzyme immunoassay (EIA; ICE Syphilis, Murex Biotech Ltd, Dartford, UK). Positive EIAs were confirmed with the Treponema pallidum particle agglutination assay (TPPA) and fluorescent treponemal antibody absorption test (FTA-Abs). The rapid plasma reagin test (RPR) was used to assist clinical staging and to detect reinfection.

Herpes simplex virus

Baseline sera and sequential specimens were tested for antibody to HSV using an algorithm involving the Enzygnost® anti-HSV IgG enzyme linked immunosorbentassay (ELISA, Behring), and type-specific ELISAs (HerpeSelect® 1 & 2 ELISA IgG, Focus Technologies) [15]. Equivocal results were resolved by Western Blot. Anogenital warts. No HPV testing or routine anogenital examinations for warts were performed in the HIM study. Participants self-reported a lifetime history of genital or anal warts at baseline, and reported annually whether they had been diagnosed with anogenital warts since the last face-to-face interview [12].

Incidence Definition

For each STI, incidence was expressed as the number of incident cases per 100 person-years (PY).

Neisseria gonorrhoeae and Chlamydia trachomatis

To be able to more comprehensively estimate the incidence of urethral and anal N. gonorrhoeae, participants who reported a diagnosis of urethral or anal N. gonorrhoeae or C. trachomatis, or tested positive to this infection at their HIM study visit were treated as incident cases. Since the NAAT testing was added to the study in January 2003, this calculation was only applicable from 2003 forward. [14].

Syphilis

Incident syphilis was defined as syphilis seroconversion or re-infection during the study period. Re-infection was defined as a four-fold increase in RPR titre on sequential serum specimens collected at annual visits.

Herpes Simplex Virus

Incident HSV-1 and HSV-2 infections were defined as HSV-1 and/or -2 seroconversion during the study.

Anogenital warts

For genital and anal warts, prevalent infection was defined as a self-reported lifetime diagnosis of warts at the baseline interview. Incident warts were defined as self-reported diagnoses of anogenital warts in the previous 12 month period.

Statistical Analysis

Statistical analyses were performed using STATA 10.0 (STATA Corporation, College Station, TX). Participants’ preference for anal intercourse position was correlated with reported UAI behavior (Table 1). As the data for number of UAI acts were highly skewed, median was used as the measure of central tendency. The exact binomial method was used to calculate 95% confidence intervals (CIs) for prevalence and incidence values. Crude and adjusted analyses were performed to identify the association of circumcision status with prevalent and incident urethral N. gonorrhoeae, urethral C. trachomatis, syphilis, HSV-1, HSV-2 and anogenital warts. Odds ratios (for prevalent infection) and Hazard ratios (for incident infection) and their corresponding 95% CIs were calculated for these associations. For analyses of the entire cohort, multivariate logistic regression (for prevalent infections; Table 2) and Cox regression (for incident infections; Table 3) models included sexual behavioral variables found to be associated with prevalent and incident infections in previous HIM analyses of anogenital N. gonorrhoeae and C. trachomatis [14], syphilis [13], herpes simplex virus [15] and anogenital warts [12]. Covariates were entered in the multivariate model if they were found to be independent predictors of each STI in these previous analyses and are outlined in the footnotes of Table 2 and Table 3. For analyses of participants predominantly practicing the insertive role in anal intercourse, the multivariate Cox regression model (Table 4) included the total number of insertive UAI acts in last 6 months. Circumcision status and age were a priori covariates in each multivariate model. To avoid the possibility of residual confounding, age was analyzed as a continuous variable due to the strong independent association of age with circumcision status in the HIM study [18] and the association of age with urethral N gonorrhoeae and C. trachomatis infections [14], genital warts [12] and HSV-1 infection [15]. Of other demographic variables associated with circumcision status in HIM [18], neither country of birth or ethnicity were associated with any prevalent or incident STI so were not included as covariates in the multivariate models.

Table 1.

Correlation of reported preference of anal intercourse position with anal intercourse behavior among participants reporting any unprotected anal intercourse (UAI) in the HIM study

	Prefer insertive role in anal intercourse (1498.3 PY a)		P	No preference of role in anal intercourse (2120.1 PY a)		P	Prefer receptive role in anal intercourse (774.2 PY a)		P
	Insertive UAIa	Receptive UAIa		Insertive UAIa	Receptive UAIa		Insertive UAIa	Receptive UAIa
Median number of acts in last 6	5.29	0		5.44	5.0		1.0	7.0
months (IQR) a	(0.35–21.58)	(0–1.48)	<0.001 b	(0.80–23.0)	(0.32–21.63)	0.118 b	(0–6.67)	(0.33–27.73)	<0.001 b
Median percentage acts last 6	98.4	1.6		51.3	48.7		14.6	85.4
months (IQR) a	(84.1–100)	(0–15.9)	<0.001 b	(39.5 –68.0)	(32.0–60.5)	0.025 b	(0.2–35.0)	(65.0–99.8)	<0.001 b

^aPY= person-years of follow-up; UAI= unprotected anal intercourse; IQR= interquartile range

^bWilcoxon signed-rank test

Table 2.

Circumcision status and prevalent (baseline) sexually transmissible infections in the HIM cohort

Sexually transmissible infection	Baseline prevalence		Univariate			Multivariate a
	Circumcised n (%)	Uncircumcised n (%)	OR b	95% CI b	P value b	OR b	95% CI b	P value b
Genital warts c (n=1332) d	75 (8.5)	44 (9.9)	0.84	0.57–1.24	0.378	0.76	0.50–1.13	0.176
Anal warts c (n=1334) d	186 (21.0)	76 (17.0)	1.29	0.96–1.73	0.091	1.14	0.84–1.55	0.397
HSV-1 (n=1370) d	683 (75.4)	347 (74.8)	1.03	0.80–1.34	0.807	0.87	0.66–1.15	0.340
HSV-2 (n=1370)d	225 (24.8)	94 (20.3)	1.30	0.99–1.71	0.058	0.95	0.69–1.31	0.760
Syphilis e (n=1396)d	30 (3.25)	13 (2.74)	1.19	0.62–2.31	0.603	0.71	0.35–1.44	0.338

^amultivariate models adjusted a priori for age and for factors previously found to be independently associated with each STI [12–15]. These were: prevalent genital warts – lifetime number of male sexual partners, lifetime number of female sexual partners; prevalent anal warts – lifetime number of male sexual partners; prevalent HSV-1 – lifetime number of male sexual partners, lifetime number of female sexual partners, prior hepatitis B virus infection, educational level attained; prevalent HSV-2 - lifetime number of male sexual partners, lifetime number of female sexual partners, prior hepatitis B virus infection, educational level attained, preference of anal intercourse position; prevalent syphilis - lifetime number of male sexual partners

^bOR = Odds ratio; CI = Confidence interval; p-values = p for homogeneity

^dn varies & does not total 1426 due to missing data, including non-consent to STI testing and participants unsure of their history of genital or anal warts

^cself-reported lifetime history of warts

^eincludes past treated infection and infections newly identified on baseline testing

Table 3.

Circumcision status and incident sexually transmissible infections in the HIM cohort

Sexually transmissible infection	Incident cases, no.	Incidence, per 100 PY a		Univariate			Multivariate b
		Circumcised	Uncircumcised	HR a	95% CI a	P value a	HR a	95% CI a	P valuea
Genital warts c	39	0.89	1.20	0.74	0.39–1.41	0.362	0.93	0.49–1.77	0.827
Anal warts c	63	1.87	1.78	1.05	0.62–1.78	0.860	1.16	0.68–1.98	0.577
HSV-1	70	7.17	8.21	0.87	0.53–1.42	0.571	1.07	0.62–1.83	0.812
HSV-2	75	2.37	2.32	1.00	0.62–1.63	0.993	0.94	0.58–1.53	0.817
Syphilis	22	0.33	0.82	0.40	0.17–0.92	0.031	0.35	0.15–0.84	0.019
Urethral gonorrhea d	127	2.96	2.85	1.04	0.70–1.58	0.818	1.09	0.73–1.63	0.667
Anal gonorrhea d	113	2.58	2.64	0.98	0.62–1.54	0.933	1.13	0.74–1.73	0.560
Urethral chlamydia d	275	6.31	6.42	0.99	0.73–1.34	0.942	1.10	0.70–1.73	0.669
Anal chlamydia d	192	4.49	4.28	1.05	0.73–1.53	0.782	1.04	0.71–1.52	0.836

^aPY = person years; HR = Hazards ratio; CI = Confidence interval; p-values = p for homogeneity

^bmultivariate models adjusted a priori for age and for behavioral factors in previous 6 months independently associated with incidence of each STI [12–15]. These were: incident genital warts – frequency of insertive anal fingering with casual partners; incident anal warts – frequency of insertive anal fingering with casual partners, frequency of insertive fisting with casual partners; incident HSV-1 – number of regular male partners, frequency of insertive oro-anal sex with regular male partners, frequency of insertive penile-oral sex to ejaculation with casual partners ; incident HSV-2 - frequency of insertive anal fingering with casual partners, frequency of receiving a dildo with casual partners; incident syphilis - number of male sexual partners, number of receptive UAI acts with HIV positive or status unknown partners; incident urethral gonorrhea - sexual contact with gonorrhea, number of casual male partners, unprotected anal intercourse according to position; incident anal gonorrhea - sexual contact with gonorrhea, unprotected anal intercourse according to position, frequency of receptive anal fingering with casual partners; incident urethral chlamydia - sexual contact with chlamydia, number of casual male partners, unprotected anal intercourse according to position, insertive oral sex to ejaculation with casual partners; incident anal chlamydia - sexual contact with chlamydia, number of casual male partners, unprotected anal intercourse according to position, frequency of receptive oro-anal sex (“rimming”) with casual partners

^cself-reported warts diagnosed within the last year

^dcombined study visit and self-reported diagnoses in the last 12 months

Table 4.

Circumcision status and incident sexually transmissible infections in the HIM cohort among participants who predominantly practiced the insertive role in unprotected anal intercourse

Sexually transmissible infection	Incident cases, no.	Incidence, per 100 PY a		Univariate			Multivariate b
		Circumcised	Uncircumcised	HR a	95% CI a	P value a	HR a	95% CI a	P value a
Genital warts c	13	0.77	1.53	0.50	0.17–1.46	0.203	0.55	0.18–1.65	0.285
Anal warts c	21	1.69	1.89	0.88	0.35–2.21	0.784	0.94	0.37–2.42	0.905
HSV-1	25	8.69	8.89	0.97	0.40–2.34	0.951	0.93	0.40–2.17	0.868
HSV-2	20	1.70	2.26	0.73	0.30–1.79	0.497	0.72	0.30–1.76	0.474
Syphilis	6	0.001	0.01	0.10	0.01–0.81	0.032	0.10	0.01–0.82	0.033
Urethral gonorrhea d	56	4.13	3.50	1.20	0.63–2.29	0.572	1.19	0.63–2.24	0.589
Anal gonorrhea d	17	1.13	1.31	0.85	0.26–2.75	0.782	0.93	0.29–3.02	0.902
Urethral chlamydia d	106	7.23	7.83	0.93	0.57–1.54	0.787	0.93	0.57–1.53	0.778
Anal chlamydia d	33	1.96	3.05	0.67	0.30–1.49	0.324	0.68	0.30–1.53	0.347

^aPY = person years; HR = Hazards ratio; CI = Confidence interval; p-values = p for homogeneity

^bmultivariate models adjusted a priori for age and for total number of insertive unprotected anal intercourse acts in last 6 months

^cself-reported warts diagnosed within the last year

^dcombined study visit and self-reported diagnoses in the last 12 month

Multiple failures in the subsequent interview(s) were only allowed for incidence calculations involving gonorrhea, chlamydia and syphilis. For incidence risk factor analyses of these STIs, mixed models were used that allowed for repeated measures in the same individuals. These models, which allow for within and between subject variability, provide robust variance estimates, and therefore appropriate p-values and confidence intervals [19].

Participants’ own circumcision status would have had no plausible effect on their STI acquisition via receptive anal sex. We examined circumcision as a risk factor for anal gonorrhea, chlamydia and warts, only to contrast these findings with the same infections at genital sites.

RESULTS

From June 2001 to December 2004, a total of 1,427 participants were enrolled. The median age at enrolment was 35 years (range 18–75 years), 68% were Australian-born and 74% were of Anglo ethnicity. 95% of participants self-identified as gay or homosexual. One participant who reported a surgically reconstructed foreskin was excluded, therefore analyses of circumcision status and risk of STIs were based on 1,426 participants.

By the end of the HIM study in June 2007, the total follow-up time was 5160.7 person-years, and median follow-up time for participants was 3.9 years. The follow-up time from commencement of gonorrhea and chlamydia NAAT testing in January 2003 to the end of the HIM study in June 2007 was 3781.5 person-years. The uptake of N. gonorrhoeae and C. trachomatis anal and urine testing was high: a total of 4090 NAAT (90.2% of eligible visits) were performed during this period. All participants consented to HSV and syphilis serological testing. The proportion of circumcised participants among those lost to follow-up was very similar to the proportion of circumcised participants in the entire cohort at baseline (65.4% vs. 65.8%, p=0.879).

Circumcision status and prevalent STIs

Results of analyses of the association of circumcision status and prevalent (baseline) STIs are shown in Table 2. There was no univariate or multivariate association of circumcision status with syphilis, HSV-1, or HSV-2 seropositivity at baseline. Nor was there an association of circumcision status with a lifetime history of genital or anal warts on either univariate or multivariate analyses.

Circumcision status and incident STIs

The association of circumcision status with incident STIs in the HIM cohort is shown in Table 3. Circumcised participants were at reduced risk of acquiring syphilis during the study period after controlling for age, number of male sexual partners and number of receptive UAI acts with HIV positive or status unknown partners (HR 0.35, 95% CI 0.15–0.84, p=0.019). Circumcision status was not associated with any other incident STI on univariate or multivariate analyses. Comparison of analyses of genital and anal STIs revealed little difference in hazard ratios for incident warts, gonorrhea and chlamydia.

Only 10.1% of total person-years of follow-up were among men who exclusively practiced insertive, but no receptive, UAI. However, 33.1% of total person-years of follow-up were among participants who reported a preference for the insertive role in anal intercourse, and whose sexual behavior closely reflected this preference (Table 1). Among these 404 participants, the protective effect of circumcision on incident syphilis infection remained despite only 6 incident infections in this subgroup (HR 0.10, 95% CI 0.01–0.82, p=0.033) (Table 4). There was no association of circumcision status with any other STI in men who predominantly practiced the insertive UAI role.

DISCUSSION

Circumcised participants in the Health in Men cohort were at significantly reduced risk of incident syphilis infection, although there was no association of circumcision with baseline syphilis seropositivity. Circumcision was not associated with prevalent or incident urethral gonorrhea or chlamydia, HSV or genital warts.

Only three previous published studies, all cross sectional, have examined the relationship between circumcision and specific STIs in homosexual men [8–10]. The first was a study of almost 900 men who have sex with men (MSM) attending a US STD clinic [9], which found no univariate association of being circumcised with prevalent urethral gonorrhea or chlamydia. Second was a small study of mostly uncircumcised Peruvian MSM, quoted in a subsequent meta-analysis [2, 10] in which there was no independent association of circumcision with prevalent syphilis. Most recently, among over 20,000 San Francisco STD clinic attendees, there was no univariate association of being circumcised with syphilis infection among either HIV-positive or HIV-negative MSM [8]. A recent meta-analysis which included some unpublished data also found no overall association of circumcision status with STIs among MSM [11]. While other published studies of circumcision and STI risk have included MSM, results were not presented separately for this subgroup [20–24].

Of the STIs considered in the HIM study the published evidence for a protective association with circumcision is strongest for syphilis. However, the majority of previous observational studies have been of cross sectional design among heterosexual males in developing countries. Most have found circumcised men to have lower odds of syphilis infection [2]. Recent data from the Rakai randomized trial in heterosexual men found no effect of circumcision of incident syphilis [3]. However, this finding may have resulted from lack of power due to the small number of syphilis infections identified on follow-up testing [25]. Although the adjusted odds ratio (OR) for prevalent syphilis in HIM was 0.71, and was lower than that of the other prevalent STIs examined (Table 1), it was not statistically significant. Syphilis was very common among Australian homosexual men in the 1970s, but was virtually eliminated during the late 1980s and 1990s [26]. One possible explanation for an association with incident but not prevalent syphilis is that participants who initiated MSM sexual activity during the late 80s and 90s would have been at very low risk of acquiring syphilis irrespective of their sexual behavior or circumcision status. Only recently, since 2001, has syphilis re-emerged in Australian MSM [13].

Recent publications have described a reduction in HSV-2 risk among circumcised heterosexual men in two of the African circumcision trials [3, 4]. We observed no such association of circumcision with incident HSV-2 infection in HIM and ours in the first study to assess the association among homosexual men.

In the HIM study, as in most longitudinal studies, HSV and syphilis were diagnosed serologically and the anatomical site of acquisition was not known [27, 28]. The initial site of infection may have been oral or anal for which a man’s own circumcision status could have no plausible protective effect. Should circumcised homosexual men be at reduced risk of genital acquisition of both these infection, the lack of data on site-specific diagnosis could, in part, explain the lack of association with circumcision in homosexual men.

No reduction in incidence of urethral gonorrhea or chlamydia was observed among circumcised participants in the HIM study. Circumcision had no influence on prevalent or incident urethral gonorrhea or chlamydia in two of the African heterosexual circumcision trials [6, 7]. Likewise, the only published study in MSM [9] found no association of circumcision status with urethral gonorrhea or chlamydia among STD clinic attendees. These findings suggest circumcision has little impact on acquisition of these urethral bacterial infections.

Data on the association of circumcision status and clinically apparent genital warts among heterosexual men is conflicting [9, 21, 23, 29–33]. Although a lower prevalence of PCR-based diagnoses of high-risk human papilloma virus (HPV) was recently reported among circumcised heterosexual men in two circumcision trials [3, 5], observational data are inconsistent regarding the effect of circumcision on penile human papilloma virus (HPV) infection [34]. There was no overall association of circumcision with self-reported genital warts in the HIM study, and this is the first time such a relationship has been examined in MSM.

Several strengths of this study should be noted. First, the prospective nature of the HIM study avoided prevalence-incidence bias which may have been an issue in most previous studies assessing circumcision status and STI risk which employed a cross-sectional design. This may be an important bias as there is some evidence that the presence of a foreskin could lead to an altered clinical expression of infection with sexually transmitted pathogens [3, 29–31, 33]. Second, potential demographic confounding factors were assessed [18] and, where appropriate, included in the multivariate model. Third, the community-based nature of the HIM study avoided selection bias which may be problematic in clinic-based studies if men of a particular circumcision status more commonly have, or notice, abnormal genital lesions. Fourth, self-reported circumcision status was validated by clinical examination in a subgroup of HIM participant and self-report was found to be a valid measure of circumcision status in this group [16]. Fifth, non-differential misclassification was minimized by the use of highly sensitive and specific diagnostic testing algorithms for gonorrhea, chlamydia, syphilis and HSV. In addition, the overall uptake of STI testing offered to HIM participants was high; over 90% of men agreed to STI testing during the study. Importantly, we sought to differentiate genital STIs for which the presence of a foreskin could plausibly affect the risk of acquisition during insertive sexual practices by two means: (1) analyses of site-specific anal gonorrhea and chlamydia infections and anal warts allowed direct comparison between anal infections for which circumcision should afford no protection and the same infections at genital sites; and (2) analyses of the subgroup of men who predominantly practiced the insertive role in UAI.

A number of limitations of the study also need to be addressed. First, it is possible that some STIs may be associated with circumcision status, but that the power of the study was insufficient to detect a significant effect during the study period. Nonetheless, a significant reduced risk of syphilis was observed in HIM despite the low incidence of infection (Table 3 & Table 4). Second, for prevalent infections such as HSV and syphilis, especially those in whom life-long seropositivity persists [27, 28], imprecise or inadequate control for past sexual behavioral confounding factors may have influenced the outcome of this analysis. However, no prevalent infection examined in HIM was associated with circumcision status on either univariate or adjusted analyses (Table 2). Third, as mentioned above, the results of HSV and syphilis serological tests do not reveal the site of infection, and the lack of association with HSV-2 observed in HIM could be a result of infections being acquired at non-penile sites for which circumcision would have no plausible benefit. Finally, anogenital warts were self-reported in HIM. This may have resulted in some misclassification as has been previously described for anogenital warts among participants in the US Multicenter AIDS cohort study [35].

Circumcision was associated with a reduced risk of incident syphilis in this cohort. There was little, if any, effect of circumcision on most other prevalent and incident STIs. To confirm our findings, further prospective studies are warranted among MSM with higher STI incidence. Alternative biologic or behavioral public health interventions to combat rising STI rates in homosexual men are needed.