HUMAN PAPILLOMAVIRUS (HPV) is a common sexually transmitted infection, and it is estimated that over 50% of sexually active men and women aged 15 to 49 years have been infected with one or more genital HPV types at some point in their lives. ¹ Of the over 38 different types of this small DNA virus known to infect the genital tract, ² only two types (HPV 6 and 11) are clearly linked with genital warts, whereas virtually all types have been linked with squamous intraepithelial lesions (SILs) of the uterine cervix. Furthermore, genital HPV types play an etiologic role in the development of virtually all cases of invasive cervical cancer (ICC). ³ Invasive cancers of the vagina, vulva, penis, and anus have also been linked with genital HPV infection, but compared with cervical cancer, these cancers are rare. ⁴ Papanicolaou (Pap) smear screening has allowed for early detection and treatment of HPV-associated cervical lesions, dramatically decreasing cervical cancer incidence and mortality among women.

As with other infections that cause significant morbidity and mortality, preventing the spread of the infectious agent throughout a susceptible population is generally more cost-effective than approaches involving early detection and treatment. In the absence of a vaccine, abstinence, mutual monogamy, and condoms are options for preventing genital HPV infection. In response to the recent congressional mandate to provide the public with more accurate information on the efficacy of condoms in preventing HPV infection, ⁵ a comprehensive review of the literature was undertaken and a meta-analysis performed.

The strongest evidence of the role of male condoms in preventing disease transmission is for HIV. A meta-analysis of 25 studies of HIV-serodiscordant heterosexual couples provided a summary efficacy estimate of 87% (95% CI, 60–95%). ⁶ The evidence is less clear, however, for other STDs. In vitro testing has shown that condoms provide 100% protection as a physical barrier to chlamydia, herpes simplex virus-2 (HSV-2), and HIV, ⁷ and some reports indicate protection against gonorrhea, chlamydia, and HSV-2. ^8–10 However, a study of Baltimore STD clinic attendees revealed that condom use was not associated with any reduction in the incidence of gonorrhea, chlamydia, trichomoniasis, or syphilis. ¹¹ Although no studies have been conducted explicitly to assess the effectiveness of condoms in preventing HPV infections, numerous publications present data on the relationship between condom use and HPV-related conditions. The present analysis examines the existing evidence on the effectiveness of condoms against HPV infections and HPV-related conditions (e.g., genital warts, subclinical HPV infection, cervical intraepithelial neoplasia [CIN], and ICC).

Methods

We examined the peer-reviewed literature on condoms and HPV-related conditions published in the English language since 1980 to identify studies for inclusion in this analysis. The search was limited to the past 20 years of published literature because before the mid-1980s HPV DNA testing was not available, and before the late 1970s classification schemes of HPV-related cervical pathology were not standardized. After initial discussions with experts in the field to identify appropriate studies, we engaged in a MEDLINE search and reviewed reference lists from published articles to identify other publications not already selected. The following search terms were used: human papillomavirus (HPV) and condoms, cervical cancer and condoms, SIL or cervical intraepithelial neoplasia (CIN) and condoms, and warts and condoms, as well as each of these search terms independently. Only studies that examined use of the male condom were included. To date, there have been no studies of the female condom and HPV.

Inclusion Criteria

Studies were included in this analysis if they met the following criteria: (1) a clear definition of the endpoint (HPV infection; HPV type; warts, SIL, CIN; cancer), (2) an exclusive definition of condom use, and (3) an appropriate comparison group. Nine studies that examined the effect of barrier contraceptive methods lumped together (condom, diaphragm, cervical cap) were excluded, as it was not possible to disentangle the effect of condoms from the other methods. Two studies conducted in populations with high rates of HIV infection were also excluded because the presented data did not allow for assessment of condom use and prevention of HPV-related conditions in the absence of immunosuppression. Compared with individuals who are not infected with HIV, those who are infected tend to have more evidence of HPV-related disease. ⁴ In case–control studies, we required controls to be selected from the same populations that gave rise to the cases ¹²; two studies did not meet this criterion. Four other studies were excluded because of inadequate assessment of either condom use or outcome.

Because of the instability of estimates derived from studies of small sample size, we also excluded all studies that had fewer than 20 cases and/or controls or less than 20 condom users (2 studies). When more than one publication reported on the same study population, we included only the one with the clearest presentation of data on the relationship between condom use and the outcome of interest. When several definitions of condom use were evaluated or more than one comparison was reported, we elected to present only one comparison, giving higher priority to the definition of “always use versus never use” and less priority to definitions suggesting occasional use. For many studies, however, condom use was measured as only “yes versus no” or “ever versus never.”

For biologic homogeneity, separate evaluations were performed for the five outcomes of interest: (1) external genital warts, (2) HPV DNA detection in genital samples, (3) cervical warts and SIL, or CIN (grade I or unspecified), (4) CIN II or III, and (5) ICC (squamous cell carcinoma and adenocarcinoma not differentiated). Cervical SIL (Bethesda classification system) and CIN (Richart's classification scheme) define the same set of lesions, although the SIL category includes cellular features of HPV infection (e.g., koilocytosis) alone, as well as classic cellular features of CIN with or without koilocytosis. ⁴

Whenever the odds ratio (OR) or relative risk (RR) presented in the publication was calculated from a clear baseline category, we utilized that estimate and corresponding 95% confidence intervals in our analysis. Because most of these studies were not designed to assess the relationship between HPV-related conditions and condom use, in some cases the OR presented in the publication did not reflect always-versus-never use of condoms. For those studies, we recalculated the OR and CI, from data presented in the publication ¹³ or data obtained from the investigators themselves. ^14,15 We also calculated ORs and 95% CIs for studies that presented percentages. ^16,17 In all instances, Epi-Info (version 6.04b; Centers for Disease Control and Prevention, Atlanta, GA) was used in the calculation of the 95% CIs. ¹⁸ When provided, adjusted ORs (OR_adj) or adjusted RRs (RR_adj) were presented.

We conducted a chi-square test of homogeneity ¹⁹ for each outcome and concluded that substantial heterogeneity was present in each group. Therefore, we elected not to calculate pooled odds ratios and have presented only the individual estimates.

Results

A total of 20 studies met all the inclusion criteria (Tables 1 and 2). Five of these studies assessed more than one outcome, providing a total of 27 different estimates on the relationship between HPV-related conditions and condom use. Two studies included data on the relationship between condom use and external genital warts ^13,17; in 6 studies the primary outcome was detection of HPV DNA in cervical samples ^14,18–24; 4 studies examined cervical SIL, cervical warts, or CIN (grade I or unspecified) ^16,25–27; 6 studies used CIN grade II or III as the outcome ^25,27–31; and 5 studies focused on ICC. ^{15,25,32–34} The majority of studies measuring HPV DNA as the primary outcome were cross-sectional, with the exception of one cohort study. In contrast, all publications on cervical cancer were reports of case–control studies. Studies assessing (1) external genital warts, (2) cervical SIL, cervical warts, or CIN (grade I or unspecified), and (3) CIN II or III represented a mixture of case–control and cross-sectional designs.

The overwhelming majority of these studies were conducted among women, with one study of men only and another of both men and women. Three of the studies on women, however, gathered condom use data directly from their husbands. ^15,28,33 Although the age range of all subjects was large, in general, studies of genital warts, HPV DNA detection, or SIL were conducted among men and women in their twenties and early thirties, while studies of ICC were conducted among slightly older women.

Several different condom measures were used in these studies, but all relied on self-report, which can sometimes result in misclassification. Time intervals were rarely stated, and only four studies specified a defined measurement period for condom use. Two studies compared users for 5 or fewer years with those who had used condoms for greater than 5 years, 1 assessed condom use in the past 4 months, and 1 compared users for 6 or more months with those who had used them for fewer than 6 months. The type of condom used (latex versus nonlatex) was not specified in any of the studies.

External Genital Warts

Only two studies assessed the relationship between condom use and external genital warts, and both found protective effects (Figure 1). One analysis examined young male military recruits, while the other evaluated men and women attending an STD clinic. Among military recruits, those who reported always using condoms were 70% less likely to have genital warts or subclinical HPV infection of the penis than male recruits who occasionally or never used them (OR, 0.3; 95% CI, 0.2–0.5). ¹⁷ Investigators tested potential confounding factors, but none influenced the association. Among both men and women attending an STD clinic, reported condom use (always versus never) reduced the likelihood of genital warts by approximately 60% (OR_adj, 0.4). Although 95% CIs were not presented for these estimates, the crude ORs with 95% CIs were similar to the adjusted estimates (men: crude OR of 0.3 and 95% CI of 0.2–0.4; women: crude OR of 0.6 and 95% CI of 0.4–0.9). ¹³

Cervical HPV DNA

All six studies measuring HPV DNA detection as the outcome were conducted among women, and only one showed a statistically significant protective effect for condom use. Among Danish sex workers who reported always using condoms with their commercial partners, those who also always used them with their private partners during the past 4 months were 80% less likely to have HPV DNA detected than those who never did (OR_adj, 0.2; 95% CI, 0.1–0.6). ²³

The odds ratios for the four other cross-sectional studies assessing condom use and HPV DNA detection ranged from 0.8 to 1.6, with the exception of the study reported by Kjaer et al. Among a random sample of women from the general population of Denmark, those who reported current use of condoms were almost four times more likely to have a nononcogenic HPV type detected than women who reported never using condoms (OR_adj, 3.8; 95% CI, 1.2–11.6). ²² A similar effect, but of lower magnitude, was reported for oncogenic HPV types (OR_adj, 1.6; 95% CI, 0.8–3.3). Of the four remaining studies, only one showed a statistically significant elevated risk for HPV DNA detection among women who reported condom use. Young et al ²⁴ found that women who reported using condoms “most of the time” or “always” were 50% more likely to have HPV DNA detected than those who reported never using condoms (OR_adj, 1.5; 95% CI, 1.1–2.0). Among female adolescents, Jamison et al ²¹ found a slight, nonsignificant increase in risk of HPV DNA detection for women who used condoms >75% of the time in comparison with those who used them <25% of the time (OR, 1.2; 95% CI, 0.7–2.1). However, Davidson et al ²⁰ reported virtually no effect among native Alaskan women (RR_adj, 0.9; 95% CI 0.5–1.5).

In the sole cohort study, Ho et al ¹⁴ found that, compared with women who reported never using condoms, those who reported always using them were slightly less likely to become HPV DNA–positive, but the effect was not statistically significant (RR_adj, 0.8; 95% CI, 0.4–1.4).

Cervical Warts and SIL or CIN (Grade I or Unspecified)

Four studies examined the effect of condom use among women in whom cervical SIL, cervical warts, CIN I, or CIN of an unspecified grade had been detected. In two of these studies there was no effect of condom use on CIN I (grade I or unspecified) after adjustment for confounding factors. ^25,27 Zondervan et al ²⁵ enrolled women attending family planning clinics and found no relationship between ever having used condoms (as a contraceptive method) and dysplasia (OR_adj, 1.0; 95% CI, 0.7–1.4). Similarly, Adam et al ²⁷ studied women referred to colposcopy clinics with abnormal Pap smears and found no association between ever using condoms and CIN I (OR_adj, 1.1; 95% CI, 0.7–1.8).

In contrast, two Finnish studies reported increased risk of cervical SIL with condom use as a contraceptive method. Syrjanen et al ¹⁶ found a slight, nonsignificant increase in risk for women attending an outpatient OB/GYN unit who reported using condoms compared with those who did not (OR, 1.4; 95% CI, 0.7–2.9). Kataja et al ²⁶ showed an 80% increased risk of SIL among women who reported using condoms (OR, 1.8; 95% CI, 1.4–2.4), but the final estimate was not adjusted for potential confounding factors, and cases were significantly younger than controls.

CIN II or III

Six studies assessed the relationship between condom use and CIN II or III. All of these studies adjusted for confounding factors, but no two studies adjusted for the same factors. Zondervan et al, ²⁵ Munoz et al, ²⁸ Becker et al, ²⁹ Wang et al, ³¹ and Adam et al ²⁷ compared women with either CIN II or III with women who had normal cervical cytology. Two of these investigative teams found a significant protective effect, yet three others reported no significant effect of condom use. Wang et al ³¹ interviewed women and their husbands and determined that women attending cervical cancer screening sites who reported any use of condoms were 70% less likely to have CIN II or III than women who reported no use of condoms (OR_adj, 0.3; 95% CI, 0.1–0.8). Adam et al ²⁷ estimated that among women referred to colposcopy clinics for abnormal Pap smears, those who reported ever having used condoms were 40% less likely to have CIN II or III than those who had never used condoms (OR_adj, 0.6; 95% CI, 0.4–0.9).

Becker et al ²⁹ surveyed women attending university-based women's and maternal/infant care clinics and found that, compared with women who had never used condoms, those who reported ever using condoms were slightly less likely to have high-grade lesions (OR_adj, 0.8; 95% CI, 0.5–1.3). Munoz et al ²⁸ conducted parallel studies in Spain and Colombia and interviewed women with and without CIN II or III and their husbands. In Spain a slight protective effect for condom use (ever versus never) was observed, although this was not statistically significant (OR_adj, 0.8; 95% CI, 0.6–1.7). In Colombia, no relationship between condom use and CIN II or III was found (OR_adj, 1.0; 95% CI, 0.5–2.7). A slightly elevated, nonsignificant risk of CIN II or III among women who had ever used condoms compared with those who had never used them (OR_adj, 1.2; 95% CI, 0.8–1.8) was reported by Zondervan et al. ²⁵

In contrast to the above studies, which compared women with CIN II or III with women with no cervical pathology, Ho et al ³⁰ enrolled only women with histologically confirmed CIN and compared those with CIN II or III with those with CIN I. Compared with women who reported never using condoms, those who reported using condoms most or all of the time were approximately 30% less likely to have CIN II (OR_adj, 0.7; 95% CI, 0.3–1.8) and nearly 70% less likely to have CIN III (OR_adj, 0.3; 95% CI, 0.1–0.8). However, the comparison for women with CIN II was not statistically significant.

Invasive Cervical Cancer

A further five studies evaluated condom use as a risk factor for ICC. Four of the five studies showed a range of protective effects for condom use, while the fifth showed no effect. None of the studies showed any increased risk.

Two studies revealed a significant protective effect for condom use. Kjaer et al ³³ selected women with cervical cancer from the Danish Cancer Registry and compared them with women from the general population. The analysis was restricted to “monogamous women” (women with one or two lifetime sex partners), and data on condom use were collected from the husband. Women whose husbands had ever used condoms were 80% less likely to have cervical cancer than those whose husbands had never used condoms (OR_adj, 0.2; 95% CI, 0.1–0.6). Thomas et al ¹⁵ also gathered data on the husband's use of condoms and found a significantly reduced risk for cervical cancer. Monogamous Thai women whose husbands always or frequently used condoms in their visits to prostitutes were 50% less likely to have cervical cancer than women whose husbands rarely or never used condoms with prostitutes (OR_adj, 0.5; 95% CI, 0.2–1.0).

Slattery et al ³² studied women from the Utah Cancer registry and compared them with women selected by random-digit dialing who had not been diagnosed with cancer. Women who reported using condoms for 6 or more months were 30% less likely to have cervical cancer than women who had used them for less than 6 months (OR_adj, 0.7; 95% CI, 0.5–1.2). Among women attending family planning clinics, Zondervan et al ²⁵ found that ever using condoms was associated with a slight decreased risk (20%) of cervical cancer (OR_adj, 0.8; 95% CI, 0.3–2.0). Hildesheim et al, ³⁴ however, found no effect on risk of cervical cancer when comparing women who had used condoms as a contraceptive method for at least 5 years with those who had never used them (OR_adj, 1.0; 95% CI, 0.6–1.5).

Discussion

Condoms are known to be effective in preventing HIV among both men and women, ^6,35–37 but data on their protective effect against other STDs are less consistent. ¹¹ The studies included here represent the best available data describing the relationship between condoms and HPV-related conditions. Although the absence of nonpublished studies raises the question of publication bias, exclusive reference to the peer-reviewed literature ensures the highest quality of data. Nevertheless, it was difficult to compare results from these studies because of the variety of different measures of condom use and the numerous different outcomes. All of the studies we evaluated had at least one of two problems: (1) they were not designed to evaluate condom use and therefore did not include precise measures of consistent or correct usage and/or (2) the temporal sequence was not established.

Generally, condom use was asked about in the context of contraception and only of female subjects (rather than of their male partners, who are the actual users). Previous studies have shown that important modifiers of condom efficacy include measures of user experience such as frequency and consistency of use and correct coverage. Individuals who often use condoms report fewer problems with breakage or slippage than infrequent users, ^38,39 yet even experienced users report episodes where problems (putting it on inside out, application after penetrative intercourse has begun, breakage, and slippage) resulted in potential transmission risk. ^40,41

Other investigators have determined that the level of reported condom use varies, depending on the type of survey questions used. “Always use” versus “never use” is a popular epidemiologic measure, because the exclusive categories clearly define opportunity for exposure to HPV. However, a comparative study determined that “always use” rarely covered use for more than a year, ⁴² which is inadequate for infections such as HPV that can be present for longer intervals before progressing to CIN II or III or ICC. The type of sex partner also determines the frequency of condom use, with condoms being more often used with new and casual partners than with regular partners. ⁴³ However, questions regarding partner type were rarely posed in these studies: only one study included questions about condom use with prostitutes, and another asked commercial sex workers about use with clients and regular partners.

Temporal sequence is an important criterion in making causal inference, yet only one study ¹⁴ was designed to look at new acquisition of HPV. For all other studies, it was impossible to determine whether condoms were used before or after acquisition of HPV. Without determining that individuals began using condoms while they were still infection-free, it is impossible to accurately assess the role that condoms play in preventing new infections. However, it is possible (but unknown) that condom use with the same partner after acquisition of a new HPV infection might reduce the total amount of virus transmitted and the number of genital sites infected.

Despite design problems inherent in the studies available for review, some observations may be made. Although only two studies were conducted among men, both suggested good protection against development of external genital warts. The sole study of external genital warts among women also showed protection; however, the level of protection was somewhat lower among these women than among men included in the same study. The limited data on heterosexual men suggest that, compared with young men who do not always use condoms with vaginal intercourse, those who do may be less likely to develop external genital warts. The evidence among women for protection from external genital warts and other HPV-related conditions is less clear because it is based on only one study.

There was no consistent evidence of a protective effect of condom use on HPV DNA detection, and in some studies, condom use was associated with a slightly increased risk for these lesions. Only among Danish commercial sex workers was there a strong, statistically significant reduction in risk with report of condom use with private partners. The fact that these sex workers also reported always using condoms with their clients suggests that they were experienced users, and as has been reported following other studies of sex workers, ⁴⁴ they may have been able to ensure their partners used condoms correctly and consistently.

The studies grouped together under the heading “cervical warts and SIL or CIN (grade I or unspecified)” represent a mixture of conditions that are often indicative of transient infections. Therefore, this group may be composed of women who have a more recently acquired HPV infection. In the study by Kjaer et al, condom use was associated with a substantially increased risk of detecting nononcogenic HPV types. Common nononcogenic types are HPV 6 and 11, which are the same types associated with external genital warts. Kjaer et al also measured condom use as current versus never use. It may be that women opted to use condoms after noticing genital warts on their male partner. Because the virus was probably present before the warts appeared, condom use then may have begun after transmission had already occurred, thus giving the impression that condom use increased risk for the nononcogenic HPV types associated with genital warts. Alternatively, condoms may do little to protect against initial infection with HPV and subsequent transient conditions, a supposition that may be supported by the slightly increased risk seen for HPV DNA detection.

The individual odds ratios, and thus evidence of a protective role for condoms, varied for studies of CIN II or III. Two studies demonstrated a protective effect of 70% for condom use, four showed only slight protection (20–40%), and two showed either no effect or a slightly increased risk. Ho et al ³⁰ showed a strong protective effect of condoms for CIN III (compared with CIN I). The effect was attenuated for CIN II, perhaps indicating the indistinct nature of CIN II as an intermediate category of cervical intraepithelial neoplasia. Whereas CIN III is a cervical cancer precursor lesion with a high probability of progression to invasion, CIN I appears to be the acute manifestation of cervical HPV infection and has a high likelihood of spontaneous resolution. ⁴⁵

Among the available data on condom use and prevention of cervical cancer is a hint that a subset of women was at least partially protected. Notably, the two studies that demonstrated a statistically significant protective effect of condoms against ICC were conducted among monogamous women and collected condom use data from the male sex partners of these women. This rigorous study design accounts for the sexual behavior of both partners and comes closer to approximating a true measure of condom use. Although the three other studies of ICC did not demonstrate statistically significant protective effects of condom use, they used less rigorous condom use definitions, and all but one demonstrated a small reduction in risk associated with condom use.

Some level of protection from HPV-related conditions was observed among 17 of the 27 populations studied; however, in most studies, protection was not substantial. It is unlikely that condoms offer the same level of protection against genital HPV infection as they do for HIV. These data suggest that condoms may prevent progression to lesions (warts, high-grade intraepithelial neoplasia), but perhaps not actual infection by HPV. We hypothesize this may be due to a reduction in the amount of virus transmitted with condom use, which could decrease the probability of developing a clinical lesion.

The data also indicate that condoms may provide more protection for a susceptible man than for a susceptible woman. All studies that were conducted among men or incorporated data collected from men on condom use showed significant and substantial protective effects. Thus, it may be that susceptible men who use condoms are more likely to be protected from infectious women than the reverse. We hypothesize that, in the process of putting on a condom, an infected man may touch the shaft of the penis, transferring virus onto his fingertips and subsequently depositing it onto the exterior of the condom as he rolls it down the penis. He would then infect his susceptible female partner during intercourse through the same vehicle designed to protect her. Sonnex et al ⁴⁶ showed that HPV DNA could be detected in fingertip samples from 64% of men with external genital warts. A susceptible man using condoms correctly and consistently might be better protected from an infectious woman because she would be less likely to contaminate the condom.

Complete protection from genital HPV infection may be impossible because infections may occur at epithelial sites not covered by the condom. Also, when condoms are used primarily for contraceptive purposes, the condom may not be put on until after external genital contact has occurred. The least protective estimates for cervical SIL, warts, CIN I, and ICC were from studies that evaluated condom use as a contraceptive method only. ^{16,25,26,31,33}

Due to the inclusion of inadequate measures of condom use and the lack of information on temporal sequence, it is not possible to use available data to draw definitive conclusions on the efficacy of condoms in the prevention of HPV-related conditions. With the potential for HPV vaccines on the horizon, ^47–49 one may ask whether there is a need to collect more conclusive data. Based on what we currently know, the answer is yes. Even if an HPV vaccine becomes available within the next 10 years, the benefits of such a vaccine would not accrue for at least another 10 to 20 years. In addition, as HPV DNA testing becomes more widely used in the management of women with Pap smears showing borderline abnormalities, an increasing percentage of women will be told they have an HPV infection. Both men and women need accurate information on how they can protect themselves and their partners from transmitting or acquiring genital HPV infection.

As suggested in a report from the American College of Obstetricians and Gynecologists, ⁵ there is a need for studies expressly designed and powered to determine the degree to which condoms prevent acquisition of HPV infection or development of HPV-related sequelae. These studies must incorporate accurate measures of condom use (including correct and consistent use), and they must be designed to show a temporal sequence. Unfortunately, the ethical concern associated with randomizing individuals to not using condoms precludes the use of the most rigorous study design: a randomized controlled trial. Given the nature of sexually transmitted diseases, more partner studies similar to those conducted by Kjaer et al ³³ and Thomas et al ¹⁵ are necessary to accurately measure condom use and sexual behavior among both members of the sexual partnership.