THE PRACTICE OF ANAL INTERCOURSE (AI), both in homosexual and heterosexual couples, has been associated with a high level of risk for human immunodeficiency virus (HIV) infection in numerous studies.1–8 In particular, receptive anal intercourse has been associated with the highest risk of HIV infection.1,2,6,8
Emphasis on increased condom use is a cornerstone of HIV prevention programs. However, condoms manufactured in the United States generally are labeled for vaginal use only. This labeling reflects the concern that condoms designed for use during vaginal intercourse (VI) may fail at an unacceptably high rate when used during AI, resulting in an increased risk of exposure to HIV and other sexually transmitted diseases (STDs). Thus, although epidemiologic studies indicate that correct, consistent condom use reduces the overall risk of sexual transmission of HIV,9–12 there is uncertainty as to the level of protection that condoms designed for use during VI can provide during AI.
Surveys have produced widely varying estimates of the prevalence of AI.13–25 Despite differences in methods, such surveys have shown that AI is practiced at least occasionally by a substantial proportion of the sexually active population. For instance, Peterson et al19 reported that 73% of adult homosexual and bisexual black men surveyed in 1990 had engaged in AI with primary or secondary partners during the preceding 6 months. Furthermore, 52% had engaged in “unprotected” AI during that time, compared to 15% to 20% of white homosexual and bisexual men engaging in this behavior noted in two 1988 surveys. This practice is not confined to the homosexual population. Among sexually active adolescents, surveys have estimated the proportion having practiced AI to be as high as 27% of men and 35% of women.17,22 A survey of sexually active college students in Canada estimated that 14% of men and 19% of women had practiced AI at some time in their lives.16 The National Health and Social Life Survey,25 conducted in 1992, estimated that 26% of men and 20% of women ages 18 to 59 had engaged in AI during their lifetimes. Of greatest concern is that in several groups studied, the majority of those who practiced AI were unprotected against HIV and other STDs, in that they used condoms inconsistently or not at all.14,15,17,20,21
Observational studies of the effectiveness of condoms may be subject to a number of limitations, depending on the study design, the population under study, and the methods used to ascertain condom use history, risk behaviors, and outcomes. Figure 1 depicts the complex interactions occurring between categories of sexual behavior and condom use frequency and shows the difficulties that may arise in correctly ascertaining these characteristics for observational studies. For instance, consider a hypothetical survey that asks sexually active participants about condom use frequency without also asking about sexual activity. A respondent who states that he is a nonuser of condoms might be engaging in unprotected VI (represented by the letter “a” in Figure 1), unprotected AI (letter “b” in Figure 1), or both unprotected AI and VI (letter “c”). Frequency of condom use also is difficult to ascertain. Survey participants given the option of responding that they “regularly” use condoms may not be using them 100% of the time and may use condoms with different frequencies for AI and VI or for primary and secondary partners. Clearly, each of these factors involves a level of risk that might not have been determined adequately by the survey. Furthermore, studies that use a cross‐sectional design, in which condom use history is correlated with current HIV status, fail to account for the possibility that subjects seroconverted before adopting their current condom use patterns or that other risk behaviors may have been responsible for HIV seroconversion.
The same epidemiologic studies that showed overall effectiveness of condoms in preventing HIV transmission did not definitively show effectiveness during AI, in part because some studies did not distinguish between “protected” and unprotected AI1,2 or, in the case of studies of heterosexual couples, failed to consider condom use during AI and VI separately.3,4,7 Another weakness of existing studies in this area is that confounding by other risk factors for HIV infection, such as contact with multiple partners or other high‐risk sexual practices, was not addressed adequately.
Despite these methodologic problems, it is possible to infer from some epidemiologic studies a risk of unprotected AI that is greater than that of protected AI (Table 1).
The Multicenter AIDS Cohort Study8 differentiated between subjects who used condoms with all, some, or none of their partners and found that among homosexual or bisexual men who practiced anogenital intercourse, those who used condoms with all partners had a significantly lower rate of HIV seroconversion than those who did so with only some partners or with none. Paradoxically, the rate of seroconversion was higher for those using condoms with some partners than for those using them with none. This finding is most likely because of the confounding frequency of condom use by number of partners. Thus, “never” users of condoms were more likely to be monogamous and therefore at lower risk for HIV exposure, resulting from contact with multiple partners, than were “sometime” users. The seroconversion rate for subjects reporting condom use with all partners was calculated at 7 per 1,000 per 6‐month interval, suggesting either that condoms provide incomplete protection or that they were not used consistently even by those using them with all partners.8
The apparently significant, although incomplete, protective effect of condom use during AI detected by the Multicenter AIDS Cohort Study is in contrast to the findings of a case‐control study of HIV seroconversion among homosexual men in three San Francisco cohorts.26 In an attempt to control for changes in sexual behaviors over the period of follow‐up (1984–1989), seroconverters were matched to three seronegative control subjects by date of first seropositive visit and the number of months elapsed from the previous visit. This analysis detected a positive association between condom use during receptive anal intercourse and HIV seroconversion, with odds ratios of 5.4 (95% CI 2.1–20.0) and 1.4 (95% CI 0.6–4.1) for condom use by some or all partners, respectively, relative to no condom use. Although the apparent association of condom use with an increased risk of HIV infection appears to contradict the findings of the Multicenter AIDS Cohort Study, it may be explained by a methodologic concern common to both studies, which is that high‐risk sexual practices (such as contact with multiple high‐risk partners) may be associated with increased condom use and may act as confounders.26 Another weakness of this study is that condom use data were available for only 48% of cases and 25% of control subjects.
Studies of HIV seroconversion in heterosexuals engaging in AI also have been difficult to interpret. The European Study Group on Heterosexual Transmission of HIV27 collected condom use information on 245 initially HIV‐negative, heterosexual partners of HIV‐positive men and women. Among 124 couples who used condoms for each vaginal or anal contact, there were no seroconversions to HIV (95% CI 0.0–1.5 per 100‐person years) as compared to 12 seroconversions among the 121 couples using condoms inconsistently (4.8 per 100‐person years, 95% CI 2.5–8.4). In an attempt to assess the effect of AI on seroconversion, the authors reported that 2 of 8 women who continued to have unprotected AI with their HIV‐infected male partners seroconverted compared to 6 of 65 women who were inconsistent condom users, but did not engage in unprotected AI. The authors concluded that the difference in cumulative rate of seroconversion between these two groups after 24 months of follow‐up (27.8 versus 11.7) was not statistically significant, possibly because of the small number of women reporting unprotected AI. This analysis is flawed in that women who reported that they did not practice unprotected AI could either have been engaging in protected AI or avoiding AI completely. The more appropriate comparison, that of seroconversion rates between study subjects engaging in unprotected AI and those engaging in protected AI, was not reported and cannot be calculated from the data provided. In an Italian study of a cohort of female partners of HIV‐infected men, condom use was not assessed separately for AI and VI. However, a small and statistically nonsignificant increased risk of HIV infection (relative ratio 1.4, 95% CI 0.4–4.8) was found in women practicing AI relative to those not practicing AI only in the subgroup of women reporting that they did not “always” use condoms.28
Estimates of Device Failure Rates
Given the practical and ethical difficulties inherent in attempting to study the effect of condom use on HIV and STD transmission, other measures of condom effectiveness are needed. Condom failure rates, as defined by slippage and breakage combined, are accepted indirect indicators of their ability to provide protection against STD, including HIV.29 Retrospective surveys, in which participants are asked to recall the number of condoms used during a particular period as well as the number that had failed during AI, have provided estimates of condom breakage ranging from 0.5% to 6%23,24,30,31 and slippage ranging from 3.8% to 5% (Table 2).24,30 One survey estimated total failure at 2.7% for receptive anal intercourse and 3.3% for insertive anal intercourse.32 Retrospective surveys may be plagued by problems with recall for the total number of devices used, the number that have failed, and the exact circumstances of device failure. Such problems may explain the range in failure estimates generated by these studies. Furthermore, the populations surveyed in these studies differed from one another in a number of characteristics, which may have had an impact on the comparability of study results, including age, gender, and sexual orientation, socioeconomic status, and geographic location.
Of particular importance is the finding that rate of condom failure varies widely with the subjects' experience in using them. For instance, a 1987 survey of male homosexuals in New York City found that whereas perepisode rate of failure was calculated as 2.7% for receptive AI and 3.3% for insertive AI overall, the range varied from a high of 15% in those who had used a condom only once in the year before the survey to a low of 1% in those having used a condom more than 10 times during the preceding year.32 The negative association of condom failure with experience is consistent with the findings of studies of condom use during VI.33,34 In addition, the types of condoms available to respondents appear to have affected estimates of failure. For instance, a survey of male homosexuals in the Netherlands reported an overall rate of per‐condom failure of 8% but found that users of condoms designated as “anal‐qualified” had a failure rate of 3%, significantly lower than for other varieties.24
Prospective data on condom performance, although likely to be somewhat more accurate, also are limited. A prospective study of condom performance in 4 female and 30 male prostitutes found 0.5% breakage during AI, compared to 0.8% during VI. These low rates of failure were attributed to a number of factors specific to this study, including the use of new, good quality condoms, experienced users, and the relatively short duration of intercourse practiced by commercial sex workers, which would tend to place less stress on the condom.35 In contrast, a prospective study of family planning clinic patients detected a much higher failure rate; of 19 contacts involving AI, 6 breaks (32%) and 4 slips (21%) were reported, significantly different from the breakage and slippage rates of 5.3% and 6.3%, respectively, calculated for this group during VI.34
New Condom Designs
There is limited clinical trial evidence to support the widespread notion that thicker or stronger condom varieties will be less likely to fail during AI. A small clinical study in the Netherlands compared seven combinations of condom characteristics and lubricant volume in a group of male homosexuals who were not regular condom users. Among the condoms tested, breakage and slippage varied from 0% to 22% and 0% to 33%, respectively. The stiffest condom tested, with an inflation volume of 30.6 1, did not break or slip. However, the stiffest condoms also were considered the most uncomfortable by the study subjects. The results of this study prompted the development and marketing of an anal‐qualified condom in the Netherlands.36 In one survey, male homosexual users of condoms designated as anal‐qualified had a failure rate of 3%, significantly lower than for other varieties.24 A larger clinical study of a condom reported to be “30% thicker than standard condoms” found a breakage rate of approximately 2% and a slippage rate of approximately 7%. The authors state that because of ethical concerns, no control condom was used in this second study; therefore, it is not known whether the group of users studied would have had equal failure rates with a standard condom.37
As our understanding of patterns of HIV transmission and the effectiveness of available preventive measures has evolved, periodic changes have been made to health education messages. The concept that condom use equals universally “safe sex” has given way to the promotion of options that result in “safer sex” for those who are sexually active. These options include limiting sexual contacts to those known to be uninfected, avoidance of those activities, such as AI, that are associated with the highest risk of infection, and correct, consistent condom use. Although there is strong evidence that condom use generally reduces sexual transmission of HIV, solid data showing the effectiveness of currently available condoms during AI, a particularly high‐risk sexual practice, still are lacking. Because unprotected AI is known to carry with it a higher risk of HIV transmission than does unprotected VI, the consequences of condom failure during AI are potentially far more devastating than during VI. In addition, the possibility always exists that for some users, the false sense of security provided by an improperly used or inadequate protective measure could increase, rather than decrease, the risk of infection.
Increasing the effectiveness of condoms, particularly for anal use, will require development of new devices that resist slippage and breakage and present less of a challenge for the inexperienced user. Although there is some evidence to suggest that it is possible to engineer a condom to better withstand the stresses of AI, such a product may be rejected by many users as a result of reduced comfort and sensation. User acceptance of condoms that are designed to be resistant to slippage and breakage has varied in the two small trials for which reports are available.36,37 An additional consideration for new condom development is that the available data on condom effectiveness during AI, whether from epidemiologic studies, clinical trials, or surveys, are based on use of latex condoms and cannot be extrapolated to condoms made of other materials, such as polyurethane.
Testing of condoms for slippage and breakage during AI also presents certain ethical issues, because a controlled clinical trial of a new anal‐qualified condom against a standard condom would require that researchers prescribe the use of two devices of unknown effectiveness to a group of subjects engaging in a behavior with known risks. This problem may be mitigated somewhat by limiting participation in such trials to couples who have been accustomed to engaging in unprotected AI before the start of the study, who are documented to be free from clinically detectable sexually transmitted infections at the the start of the study, and who agree to be mutually monogamous for the duration of the study. Such selection criteria should protect investigators from the perception that they are placing study subjects at increased risk by encouraging a high‐risk behavior using a protective device of unproven effectiveness. An added benefit of these selection criteria would be to allow the assessment of acceptability of new devices among nonusers of condoms. Testing of condoms in this group would provide the best “real world” assessment of slippage, breakage, and user preferences. Clinical trials would need to evaluate the effect on failure rates of such factors as user experience, duration of intercourse, partner position, and use of additional lubricant and should ideally allow for measurement of device failure rates in both heterosexual and homosexual couples and during AI as compared to VI.
Condoms provide less‐than‐perfect protection against sexual transmission of HIV and other STDs during AI, not only because they are unacceptable to many users and therefore are used inconsistently, but because users, particularly those who are inexperienced, face slippage and breakage rates that markedly could increase their risk of infection. Although under conditions of perfect use, condom failure rates during AI and VI may be comparable, the wide range of failure rates found by both prospective and retrospective studies shows the practical limitations of condoms as a protective measure during AI, particularly for the inexperienced user. Furthermore, epidemiologic studies of HIV seroconverters have provided conflicting results regarding the effectiveness of condoms in preventing HIV transmission during AI. The message of the currently available data is that there is no way to eliminate entirely the risk of AI with a partner who is or who has a high likelihood of being infected. However, condoms used correctly and with every act of AI provide better protection than those used incorrectly, sporadically, or not at all, although the reduction in risk cannot be quantified. For individuals who choose to continue engaging in AI, despite an understanding of the risks that this behavior entails, consistent, correct condom use should be recommended as a risk‐reduction strategy.
Efforts to develop new condoms are predicated on the assumption that products that are more resistant to slip‐page and breakage during AI use also will confer increased protection against HIV. Although this assumption is plausible, it would, of course, be ethically indefensible to conduct a controlled clinical trial of the ability of any product to prevent HIV transmission in discordant couples. Therefore, long‐term follow‐up studies of HIV‐seronegative individuals who choose to remain sexually active with HIV‐infected partners will still be needed. Such studies should build on previously published efforts by separating condom use during VI and AI and by including variables measuring both frequency of condom use and frequency of failure in analysis of seroconversion rates.
1. Winklestein W, Jr., Lyman DM, Padian N, et al. Sexual practices and risk of infection by the human immunodeficiency virus: The San Francisco Men's Health Study. JAMA 1987; 257:321–325.
2. Darrow WW, Echenberg DF, Jaffe HW, et al. Risk factors for human immunodeficiency virus infections in homosexual men. Am J Public Health 1987; 77:479–483.
3. Seidlin M, Vogler N, Lee E, Lee YS, Dubin N. Heterosexual transmission of HIV in a cohort of couples in New York City. AIDS 1993; 7:1247–1254.
4. Lazzarin A, Saracco A, Musicco M, Nicolosi A. Man-to-woman sexual transmission of the human immunodeficiency virus: risk factors related to sexual behavior, man's infectiousness, and women's susceptibility. Arch Intern Med 1991; 151:2411–2416.
5. European Study Group on Heterosexual Transmission of HIV. Comparison of female to male and male to female transmission of HIV in 563 stable couples. BMJ 1992; 304:809–813.
6. Kingsley LA, Rinaldo CR Jr., Lyter DW, et al. Sexual transmission efficiency of hepatitis B virus and human immunodeficiency virus among homosexual men. JAMA 1990; 264:230–234.
7. Padian NS, Marquis L, Francis DP, et al. Male-to-female transmission of human immunodeficiency virus. JAMA 1987; 258:788–790.
8. Detels R, Visscher BR, Jacobson LP, et al. Sexual activity, condom use, and HIV-1 seroconversion. In: Voeller BR, Reinisch JM, Gottlieb M, eds. AIDS and Sex: an Integrated Biomedical and Biobehavioral Approach. New York: Oxford University Press, 1990:13–19.
9. Roper WL, Peterson HB, Curran JW. Commentary: condoms and HIV/STD prevention-clarifying the message. Am J Public Health 1993; 83:501–503.
10. Cates W, Stone KM. Family planning, sexually transmitted diseases and contraceptive choice: a literature update-Part I. Fam Plann Perspect 1992; 24:75–84.
11. Nicolosi A, Leite MLC, Musicco M, Arici C, Gavazzeni G, Lazzarin A, for the Italian Study Group on HIV Heterosexual Transmission. The efficiency of male-to-female and female-to-male sexual transmission of the Human Immunodeficiency Virus: a study of 730 stable couples. Epidemiology 1994; 5:570–575.
12. Centers for Disease Control and Prevention. Update: barrier protection against HIV infection and other sexually transmitted diseases. MMWR 1993; 42:589–591, 597.
13. Bolling DR. Prevalence, goals and complications of heterosexual anal intercourse in a gynecologic population. J Reprod Med 1977; 19:120–124.
14. Kelly JA, Murphy DA, Roffman RA, et al. Acquired immunodeficiency syndrome/human immunodeficiency virus risk behavior among gay men in small cities. Findings of a 16-city national sample. Arch Intern Med 1992; 152:2293–2297.
15. Lemp GF, Hirozawa AM, Givertz D, et al. Seroprevalence of HIV and risk behaviors among young homosexual and bisexual men. JAMA 1994; 272:449–454.
16. McDonald NE, Wells GA, Fisher WA, et al. High-risk STD/HIV behavior among college students. JAMA 1990; 263:3155–3159.
17. Moscicki AB, Millstein SG, Broering J, Irwin CE Jr. Risks of human immunodeficiency virus infection among adolescents attending three diverse clinics. J Pediatr 1993; 122:813–820.
18. Osmond DH, Page K, Wiley J, et al. HIV infection in homosexual and bisexual men 18 to 29 years of age: the San Francisco Young Men's Health Study. Am J Public Health 1994; 84:1933–1937.
19. Peterson JL, Coates TJ, Catania JA, Middleton L, Hilliard B, Hearst N. High-risk sexual behavior and condom use among gay and bisexual African-American men. Am J Public Health 1992; 82:1490–1494.
20. Remafedi G. Predictors of unprotected intercourse among gay and bisexual youth: knowledge, beliefs and behavior. Pediatrics 1994; 94:163–168.
21. Silvestre AJ, Kingsley LA, Wehman P, Dappen R, Ho M, Rinaldo CR. Changes in HIV rates and sexual behavior among homosexual men, 1984 to 1988–92. Am J Public Health 1993; 83:578–580.
22. Stanton B, Li X, Black M, et al. Sexual practices and intentions among pre-adolescent and early adolescent low-income urban African-Americans. Pediatrics 1994; 93:966–973.
23. Tindall B, Swanson C, Donovan B, Cooper DA. Sexual practices and condom usage in a cohort of homosexual men in relation to human immunodeficiency virus status. Med J Aust 1989; 151:318–322.
24. van Griensven GJP, de Vroome EMM, Tielman RAP, Coutinho RA. Failure rate of condoms during anogenital intercourse in homosexual men. Genitourin Med 1988; 64:344–346.
25. Michael RT, Gagnon JH, Laumann EO, Kolata G. Sex in America: A Definitive Survey. New York: Little, Brown and Company, 1994:140.
26. Samuel MC, Shiboski S, Engel RR, et al. Factors associated with Human Immunodeficiency Virus seroconversion in homosexual men in three San Francisco cohort studies, 1984–1989. J AIDS 1993; 6:303–312.
27. deVincenzi I, for the European Study Group on Heterosexual Transmission of HIV. A longitudinal study of human immunodeficiency virus transmission by heterosexual partners. N Engl J Med 1994; 331:341–346.
28. Saracco A, Musicco M, Nicolosi A, et al. Man-to-woman sexual transmission of HIV: longitudinal study of 343 steady partners of infected men. J AIDS 1993; 6:497–502.
29. Steiner M, Trussell J, Glover L, Joanis C, Spruyt A, Dorflinger L. Standardized protocols for condom breakage and slippage trials: a proposal. Am J Public Health 1994; 84:1897–1900.
30. Golombok S, Sketchley J, Rust J. Condom failure among homosexual men. J AIDS 1989; 2:404–409.
31. Weinstock HS, Lindan C, Bolan G, Kegeles S, Hearst N. Factors associated with condom use in a high-risk heterosexual population. Sex Transm Dis 1993; 20:14–20.
32. Thompson JL, Yager TJ, Martin JL. Estimated condom failure and frequency of condom use among gay men. Am J Public Health 1993; 83:1409–1413.
33. Steiner M, Piedrahita C, Glover L, Joanis C, Spruyt A, Foldesy R. The impact of lubricants on latex condoms during vaginal intercourse. Int J STD AIDS 1994; 5:29–36.
34. Sparrow MJ, Lavill K. Breakage and slippage of condoms in family planning clients. Contraception 1994; 50:117–129.
35. Richters J, Donovan B, Gerofi J, Watson L. Low condom breakage rate in commercial sex [Letter]. Lancet 1988; 2:1487–1488.
36. Wigersma L, Oud R. Safety and acceptability of condoms for use by homosexual men as a prophylactic against transmission of HIV during anogenital sexual intercourse. Br Med J Clin Res 1987; 295:94.
37. Golombok S, Sheldon J. Evaluation of a thicker condom for use as a prophylactic against HIV transmission. AIDS Educ Prev 1994; 6:454–458.