Randomized clinical trials in South Africa, Uganda, and Kenya showed 50–61% efficacy in reducing HIV incidence among men undergoing circumcision compared with those who remained uncircumcised [1–3]. On the basis of these data, the World Health Organization now recommends male circumcision as an important element of HIV prevention programs . There is concern, however, that, for some men, male circumcision might increase HIV infection risk during the immediate postoperative period, if sexual intercourse is resumed before full wound healing. HIV infection risk may be increased through local inflammation during healing, compromised dermal integrity, or other mechanisms. To better inform recommendations for men and their partners, we assessed the risk of HIV seroconversion during the immediate postoperative period among men participating in three African circumcision trials.
Study populations and criteria for inclusion in this analysis
The Kisumu and Orange Farm trials enrolled men aged 18–24 years and the Rakai trial enrolled men aged 15–49 years. This analysis is limited to HIV-negative men who were randomized to and underwent circumcision. The analysis does not include control arm men who became circumcised through nontrial services. Kisumu participants had to be sexually active for trial eligibility, whereas the Orange Farm and Rakai trials included men with no prior sexually activity. This analysis included men who were not sexually active at enrollment as they could have initiated sex during follow-up. Trial recruitment, enrollment, reasons for refusing enrollment, and follow-up have been described previously [1–3]. The three trial protocols were approved by institutional review boards in the countries where the studies were conducted and in the donor countries. All trials were overseen by data and safety monitoring boards [1–3].
Circumcision procedures and surgical follow-up
The Orange Farm and Kisumu trials both used the ‘forceps guided’ procedure [1,3], whereas the Rakai trial employed a sleeve resection method . In the Orange Farm trial, participants were circumcised by general practitioners in their surgical offices. In the Kisumu and Rakai trials, circumcisions were performed by trained and certified medical doctors or clinical officers in the studies' operating theatres.
Postoperative follow-up visits for wound assessment were scheduled at 24–48 h, 5–9 days, and 4–6 weeks for Rakai; at 3, 8, and 30 days for Kisumu; and at 3 months for Orange Farm. In the Orange Farm trial, a nurse conducted a genital examination and recorded adverse events and wound healing at each follow-up visit. In the Rakai trial, adverse events and wound healing were assessed by clinical officers (similar to ‘physician's assistants’ ) and in the Kisumu trial by a medical doctor or clinical officer .
HIV testing occurred prior to randomization in each trial, and detailed methods have been reported previously [1–3]. In Kisumu, testing was repeated 1, 3, 6, 12, 18, and 24 months after randomization . In Rakai, testing was repeated 6, 12, and 24 months after randomization . In Orange Farm, testing was repeated 3, 9, and 21 months after randomization. Each trial has reported their HIV testing methods [1–3]. This analysis includes HIV status at 3 months and at 6 months for Kisumu, at 3 months for Orange Farm, and at 6 months for Rakai.
Behavioral counseling on wound care and postoperative abstinence
Participants in the Orange Farm trial were counseled by certified counselors and the general practitioners performing the surgery to abstain from sex for 6 weeks . In the Rakai trial, men were counseled by clinical officers, given written information on wound care, and advised to abstain from sex until wound healing was certified as complete . In the Kisumu trial, men were counseled at the 3-day, 8-day, and 30-day follow-up visits by professional counselors, were given written instructions including directions on wound care, and were advised to abstain from sex for at least 1 month after the procedure .
Behavioral data collection
During scheduled postoperative visits, participants were asked if they had had intercourse since circumcision. In the Rakai trial men were asked, ‘Have you had intercourse since the operation?’ In addition, the Rakai trial asked how many days after circumcision participants had first had sex. In the Kisumu trial at 3 and 8 days after circumcision, men were asked, ‘Have you had sexual intercourse since circumcision?’ At their 30-day visit, they were asked, ‘How many times in the past month have you had sexual intercourse?’ In the Orange Farm trial at the 3-month visit, men were asked, ‘Did you have sex between your circumcision and today?’, and if yes, they were also asked how many days after circumcision they first had sex.
For this analysis, the outcome was HIV seroconversion at 3 months (Orange Farm and Kisumu trials) and at 6 months (Rakai and Kisumu trials). The primary exposure variable was early resumption of sex after circumcision. For uniform comparison, we defined resumption of sex after circumcision as early intercourse if sex was reported less than 42 days after circumcision in Rakai and Orange Farms trials. For the Kisumu trial, men were not asked how many days after circumcision they first had sex; if the participant reported intercourse since circumcision on the 30-day visit and the visit date occurred within 42 days of circumcision, we inferred that intercourse occurred less than 42 days after circumcision. The secondary exposure variable was complete wound healing at 1 month after circumcision. The Rakai and Kisumu trials had clinician-assessed measures of wound healing. For the Kisumu and Rakai trials, we also compared wound healing at the 30-day or 4-week visit and HIV seroconversion at 3 months (Kisumu only) and at 6 months. Comparable data were not available from Orange Farm.
Chi-square tests and Fisher's exact tests were used to assess associations within trial site. Data for Kisumu and Rakai participants aged 18–24 years and for Kisumu and all Rakai participants were pooled to compare HIV seroconversion at 6 months for men reporting early sex to those who did not report early sex. Exact methods were used (StatXact, version 8.0; Cytel Software, Inc., Cambridge, Massachusetts, USA) to calculate site-specific and common odds ratios (ORs) with exact confidence intervals (CIs).
Socio-demographics and sexual behaviors
The age distribution among 18–24-year-old men was similar across sites (Table 1). A greater proportion of men in Rakai were currently married or living as married as compared with other participants. Among Orange Farm participants, 108 (8.4%) reported no sexual activity at enrollment, although 24 of these men reported sexual activity at their 3-month follow-up visit. Among 1095 Rakai participants aged 18–24 years, 140 (12.8%) reported no sexual activity at enrollment and none of these men reported sexual activity at their 4-week follow-up. Kisumu participants were younger at sexual debut, had longer duration of sexual activity, and reported more partners during the past 12 months and in their lifetimes than the other trial sites.
The 30-day and 4-week follow-up visits were attended by 1266 (95.1%) of Kisumu participants and 2282 (98.1%) of Rakai participants [1068 (97.5%) aged 18–24] (Table 2). For Orange Farm, 1293 of men attended the 3-month follow-up visit. Most 30-day and 4-week visits occurred 28–34 days after circumcision for Kisumu (89.7%) and Rakai (72.8%) participants. Among 18–24-year-olds, the proportion of men reporting early sex was similar for Kisumu (3.9%) and Rakai (5.4%) participants but was much higher among Orange Farm trial participants (22.5%).
In all three trials, early sex was reported more often among men who were married or living as married and among men with more sex partners in the past 12 months and in their lifetimes (Table 3). Early sex was not associated with earlier age at the first intercourse, but was more likely among men with longer duration of sexual activity, although this was not statistically significant in the Rakai trial.
HIV seroconversions at 3 months by early resumption of sexual activity: Kisumu and Orange Farm trials
At the 3-month follow-up visit, there were three HIV seroconversions among 1266 Kisumu participants (Table 2). Among 49 participants who reported early sex, there were no seroconversions, whereas three of 1207 (0.2%) men who did not report early sex seroconverted (P > 0.999). There were seven Kisumu participants who reported having intercourse since circumcision at their scheduled 30-day visit, but the visit date occurred more than 42 days after surgery, so we were unable to determine whether or not this was early sex. None of these men HIV-seroconverted at the 3-month visit. There was one HIV seroconversion at 3 months among 1293 Orange Farm participants. Among 289 men who reported early intercourse, there was no seroconversion, whereas one of 995 participants who did not report early intercourse seroconverted (0.1%, P > 0.999). Combining data from the Orange Farm and Kisumu trials, the HIV seroprevalence at 3-month follow-up for both trials was 0.0% among men who reported early sex compared with 0.2% among men who did not report early sex after circumcision (P > 0.999).
HIV seroconversions at 6 months by early resumption of sexual activity: Kisumu and Rakai trials
Among 1206 Kisumu participants who did not report early sex, five (0.4%) seroconverted by the 6-month follow-up visit, compared to one seroconversion among 49 (2.0%) men who reported early sex (Table 4). Among 1008 Rakai participants aged 18–24 years, there were eight (0.8%) seroconversions among 951 men who did not report early sex compared with one (1.8%) seroconversion among 56 men who reported early sex (Table 4). Combined, the seroprevalence at 6 months for 18–24-year-olds in the Kisumu and Rakai trials was 1.9% among men who reported early sex and 0.6% among men who did not report early sex. The combined OR for seroconversion at 6 months for men reporting early sex was 2.99 (95% CI 0.32–13.6; P = 0.168; Table 4). If the seven Kisumu men in whom early sex could not be determined due to visit timing were classified as having early sex, then the combined OR for seroconversion at 6 months for 18–24-year-old Kisumu and Rakai men reporting early sex would have been 2.84 (P = 0.181).
Among 2170 Rakai participants aged 15–49 years, there were 12 (0.6%) HIV seroconversions among 1193 men who did not report early sex compared with two (1.1%) seroconversions among 177 men who reported early sex (P = 0.637). Including all Kisumu and Rakai participants, the combined OR for seroconversion at 6 months for men reporting early sex was 2.42 (95% CI 0.45–8.57, P = 0.154). Including as early sex, the seven Kisumu men whose follow-up visit occurred more than 42 days after circumcision resulted in a combined OR for seroconversion at 6 months of 2.35 (P = 0.163).
Circumcision wound healing and HIV seroconversion at 3 and 6 months
Among Kisumu participants, 16 (1.3%) men did not have complete wound healing at the 30-day visit. One (6.3%) of these 16 men seroconverted at 3 months, whereas there were two (0.2%) seroconversions among 1247 men with complete wound healing at the 30-day visit (P = 0.075). One of the 16 men with incomplete healing reported having early sex, but this was not the participant who seroconverted. The 30-day follow-up visit and wound assessment for the participant who seroconverted occurred at 24 days after circumcision, and he reported having no sex partners since circumcision. At the 6-month visit, there was still only one seroconversion among the 16 men who did not have complete wound healing at the 30-day visit, whereas there were five (0.4%) seroconversions among 1246 men who had clinically documented complete wound healing (P = 0.145).
Among 18–24-year-old Rakai trial participants, 122 (11.4%) men did not have complete wound healing at the 4-week visit. Of these men, 112 had HIV test results available at 6 months and none seroconverted. This did not differ from seroconversions among men with complete wound healing [9/888 (1.0%); P = 0.684]. Two (1.6%) of the 122 men with incomplete wound healing reported early sex. Among 15–49-year-old Rakai participants, 318 (13.9%) men did not have complete wound healing at the 4-week visit. Of these men, 302 had HIV test results available at 6 months and two (0.7%) seroconverted. This did not differ from seroconversions among men with complete wound healing [12/1869 (0.6%); P > 0.999]. Fifteen (5.0%) of the 302 men with incomplete wound healing reported early sex; none of these men seroconverted.
Only a small proportion of men in the Kisumu and Rakai circumcision trials reported sexual intercourse within 42 days of circumcision. We found no seroconversions at 3-months postcircumcision in men who reported early sex and no association between early resumption of sex and HIV seroconversion by 3 or 6 months compared with men who did not report early sex. Nevertheless, the point estimates for the site-specific and combined ORs of HIV infection for those men who resumed sex early against those who did not were greater than one among Kisumu and Rakai trial participants, suggesting the possibility of an increased risk. In the immediate postoperative period, it is conceivable that HIV infection risk may be increased through local inflammation during healing or compromised dermal integrity. There is little evidence of an association between incomplete wound healing and HIV acquisition. There was just one seroconversion at 3 months among men who reported incomplete wound healing in Kisumu and no association between incomplete wound healing and HIV seroconversion at 6 months in Kisumu and Rakai.
Each trial anticipated the potential for increased HIV infection during the postoperative period and included robust counseling of participants on wound care and abstinence. The Kisumu trial recommended abstinence for 4 weeks, whereas the Rakai trial recommended abstinence until certified wound healing. Over 94% of men in the Kisumu and Rakai trials reported abstinence for at least 42 days suggesting that this health education was effective. In the Orange Farm trial, 22% of men reported early sex after circumcision, although the longer time until assessment (3 months) may have limited men's ability to recall when they resumed sex. Alternatively, more men in the Orange Farm trial may have had early sex. Differences in reported adherence to advice to abstain from sex may reflect differences in the counseling programs. The counseling regimen in the Orange Farm trial was designed to approximate programmatic implementation of circumcision. Nevertheless, the differing counseling approaches resulted in 78–95% client compliance with abstinence recommendations. In contrast to the clinical trials, implementation of circumcision programs in community settings will likely not provide multiple, immediate follow-up visits and counseling after the surgery. Thus, identifying effective and reproducible counseling strategies in diverse settings is critical for programmatic male circumcision implementation to limit the potential for increased HIV infection risk during the early postoperative period.
More men in Rakai than in Kisumu did not have certified wound healing at the 4-week follow-up. This may be explained in part by the different timing of visits, with a greater proportion of Rakai than Kisumu participants attending a follow-up visit at 21–27 days (Table 2). Healing was assessed by physical examination in both trials. In Rakai, complete wound healing was defined as: ‘healthy scar formation; no scab or open wound.’ In Kisumu, complete wound healing was defined as, ‘no scab, open wound, swelling or redness’. Such subtle differences in definitions or in ascertainment may affect comparability of healing. In addition, it is possible that differences in time until complete wound healing reflected the different surgical methods (i.e. sleeve versus forceps guided procedures, cautery versus ligation for hemostasis). Despite the greater proportion of Rakai men who were classified as having incomplete wound healing at the 1-month visit, we found no association between incomplete wound healing and HIV seroconversion at 6 months. The precise date of complete wound healing is unknown, as wound status was only determined when the participant presented at the follow-up visit. Only 17 (5.3%) of 318 men with incomplete wound healing in the two trials reported early intercourse. Although intercourse prior to complete wound healing may have been underreported, men with incomplete wound healing may be less likely to have sex as a result of comprehensive counseling or physical discomfort. In the Rakai trial, resumption of intercourse prior to complete wound healing was associated with increased surgical complication rates .
As adult male circumcision is implemented in programmatic settings, HIV-positive men undergoing circumcision may require different recommendations for wound healing and sexual abstinence, as wound healing is delayed in HIV-positive men compared with HIV-negative men [5–7].
This analysis was limited by the low number of seroconversions in the 6 months after circumcision and had insufficient power to detect meaningful associations. For example, with a sample of 2200 individuals (approximately the number of participants included across two trials), 10% of men reporting early sex, and 0.6% HIV seroconversions, the probability of detecting an OR of 3.0 as statistically significant at the P < 0.05 level is 48% (two-sided Wald test). Sexual intercourse less than 1 month after surgery will occur less frequently than the 10% estimate of early sex using a 6-week definition. As to what the exact recommendation for delay of sexual intercourse after circumcision should be (until complete wound healing, 30 days, or 6 weeks) requires closer study of variability and progression of wound healing. This analysis was limited to men who were HIV negative at enrollment, as the goal was to ascertain whether there was an increased risk for HIV infection associated with early resumption of sex after surgery. The trials differed in the timing of follow-up visits, surgical and counseling approaches, assessment of wound healing, and methods to determine resumption of sex after surgery, and these differences may constrain this combined analysis.
In the Kisumu and Rakai trials, less than 6% of men reported having early sex after circumcision, and very few reported sex prior to wound healing. The proportion reporting early sex was higher in Orange Farm. It will be important to evaluate adherence to instructions to delay resumption of intercourse once service programs roll out. Although the difference in HIV incidence was not statistically significant among men reporting early resumption of sex, the data do not preclude the possibility that their risk of acquiring HIV may be increased. The protective effect of male circumcision on HIV seroconversion observed in the three trials incorporates any potential increased risk associated with early resumption of sex. Thus, whereas there may be increased risk of HIV seroconversion among men undergoing circumcision who resumed sex early compared with men who did not resume sex early, the protective effect of circumcision over 18–24 months was strong and highly statistically significant. However, the trial results were obtained in the context of vigorous counseling against resumption of sex before full wound healing. When circumcision is made widely available, there may be less opportunity for such intense counseling. Therefore, monitoring of early sex and assessments of HIV infections attributable to early sex should be part of programs providing widespread MC services. Circumcision programs should ensure counseling and instruction for wound care, personal wound assessment, and sexual abstinence to limit the possible risk of HIV infection during the early postoperative period. Enhanced counseling should include active involvement of female partners where possible and intense counseling for married men who are more likely to resume sex early.
Foremost we thank the men who volunteered to participate in these studies. We thank the project staff for their long hours of work and dedication to sound research. The authors would like to thank John Krieger for his useful comments on an earlier draft of this manuscript; Stephen Watya for providing training and oversight for surgery; and Carolyn Williams, Melanie Bacon, and J.O. Ndinya-Achola for their support. The Orange Farm Trial was funded by ANRS, Paris, France; the National Institute for Communicable Diseases, Johannesburg, South Africa; and the Institut National de la Santé et de la Recherche Médicale, Paris, France. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript. The Rakai Trial was supported by a grant (UO1 AI11171-01-02) from the National Institutes of Allergy and Infectious Disease (NIAID), Division of AIDS, United States National Institutes of Health (NIH) and in part by the Division of Intramural Research, NIAID, NIH. The Kisumu Trial was supported by grant number AI50440 from the NIAID, Division of AIDS, NIH; and by grant number HCT 44180 from the Canadian Institutes of Health Research. S.M. was supported by a CIHR Investigator Award. We thank the members of the NIH data safety monitoring board who monitored the Kisumu and Rakai trials, as well as the institutional review boards that provided oversight. We are also grateful for the advice provided by the Rakai community advisory board. The Orange Farm, Rakai, and Kisumu trials have been registered in http://www.clinicaltrials.gov under the numbers NCT00122525, NCT00425984, and NCT00059371, respectively.
Principal contributions of authors are as follows:
B.A., R.C.B., R.H.G., G.K., S.M., A.P., D.S., D.T., and M.J.W. contributed to the study concept and design. Acquisition of data was conducted by K.A., B.A., R.C.B., R.H.G., G.K., S.M., A.P., D.S., D.T., and M.J.W. Drafting of the manuscript was conducted by S.D.M. Critical revision of the manuscript for important intellectual content was conducted by K.A., B.A., R.C.B., R.H.G., G.K., S.D.M., S.M., C.B.P., A.P., D.S., D.T., and M.J.W. Statistical analysis was performed by S.D.M. and C.B.P. Analysis and interpretation of data were performed by B.A., R.C.B., R.H.G., S.D.M., S.M., C.B.P., M.J.W.B.A., R.C.B., R.H.G., S.M., and M.J.W. obtained funding. Administrative, technical, or material support was provided by K.A., B.A., R.C.B., R.H.G., G.K., S.M., C.B.P., A.P., D.S., D.T., and M.J.W.