Half of all uncircumcised, HIV-negative men in this study from Kenya were infected with at least 1 HPV type at baseline. Nearly 30% of surveyed men had multiple penile HPV type infections, with HPV-16 and 18 accounting for >25% of all infections. If prevalent carcinogenic HPV-16 and 18 infections were prevented by mass vaccination, there is a theoretical possibility that this could result in other HR-HPV types filling the ecological niche of HPV-16 and/or 18 [44
]. Because there are reported differences in genotype distributions across sexes and geographical regions [5
], the potential for HPV type replacement could differ across study populations. The present study represents, to our knowledge, the first detailed investigation of multiple HPV infections and type associations within a cohort of young men from Africa. Using a cross-sectional study design, we did not find any negative associations between vaccine-preventable types HPV 16, 18, 6, and 11, and 41 nonvaccine-preventable HPV types.
All HPV genotypes were more likely to be detected as multiple rather than single-type infections. Further, all associations between HPV types were positive, suggesting that men who become infected are more likely to have >1 HPV infection, possibly due to host behavioral or immunological factors [7
]. When analyses were restricted to HPV-infected men, associations between specific HPV types generally shifted closer toward 1.0, which suggests that specific HPV types are not associated with one another. We also observed a higher number of men than expected with ≥4 HPV type infections, suggesting that individual HPV infections were not independent from each other. This is likely due to the fact that men with ≥4 HPV infections reported less condom use in the last 6 months and more lifetime partners than did men with fewer HPV infections. These observations in men are consistent with results from studies among women [42
]. Numerous prospective studies have found that women with HPV infection at baseline are more likely to acquire additional HPV types [7
] and that acquisition of multiple HPV types occurs more often than expected [7
]. Therefore, it is likely that the observed pattern of number and types of HPV infections reflect differences in HPV persistence or acquisition in men with multiple infections.
Positive associations between HPV types were observed yet appeared not to be type-specific. Our results among Kenyan men are consistent with a prospective study of American female college students that found no 2 HPV types are more likely to be acquired together than any other HPV types [26
]. Most studies that have examined HPV type associations in women have reported positive or no associations between HPV types, regardless of the pairwise combinations, analytical methods, HPV genotyping methods, or study population [26
]. However, among female colposcopy clinic attendees in Italy, where the genotype distribution likely differs compared to the general population, coinfection with species A7 and A10 and with HPV-31 and 52 occurred less often than expected [48
]. A large cross-sectional study of Danish women referred for testing based on clinical suspicion of infection found that HPV-51 was negatively associated with HPV-16 [42
In this large sample of men with an HPV prevalence of 50%, we examined type-type associations, even for rarer HPV types, using semi-Bayesian methods that incorporated a shrinkage factor. We chose shrinkage methods because they allowed us to include all 4 vaccine types and numerous potential confounders in each model. In addition, these methods serve as an adjustment for multiple comparisons [49
] and reduce the number of spurious associations as compared with maximum likelihood methods [50
]. There is, however, a trade-off between precision and bias in methods that incorporate a shrinkage factor [40
]. We therefore conducted sensitivity analyses to understand the effect of our statistical model choices and to compare our results with other studies that used maximum likelihood estimation [42
]. Different methods to analyze our data resulted in wider confidence intervals and, in some cases, nonsignificant ORs <1.0. However, all methods resulted in the same conclusions regarding the lack of evidence of negative associations between prevalent HPV type infections in men. As in previous studies [42
], we made multiple comparisons across outcomes, which can increase the possibility of falsely concluding that certain HPV types are associated. By including all 4 vaccine types in each of our models, the number of comparisons was reduced compared with other pairwise analyses [24
In the present study, we could only determine if the number and type of HPV infections were independent. The causes of nonindependence may be differences in host susceptibility, the distribution of HPV genotypes among female sexual partners, primer competition during PCR detection, or molecular interactions between HPV types that inhibit infection with other HPV types (type competition). Although we were able to control for several measured confounders, it is likely that there is residual confounding by other unmeasured behavioral risk factors, immunological differences between men [7
], and HPV-type exposure from female sexual partners. In populations where negative associations between HPV types are observed, molecular studies and prospective studies of couples are needed to address whether these associations are likely due to true biological type competition or competition for consensus primers.
Our study has additional limitations to consider when interpreting the findings. We detected 103 single HPV-X infections. If HPV-X represents multiple untyped infections, our reported prevalence of multiple infections and the number of HPV types detected are an underestimation. The likely impact of untyped HPV infections on the conclusions regarding HPV-type competition is minimal because inferences are mainly drawn from type-specific analyses. In addition, the prevalence and the type distribution of HPV has been shown to differ by circumcision status; thus, the results from this cohort of uncircumcised men may not be generalizable to circumcised populations [6
]. No participants received HPV prophylactic vaccination, so we could only observe differences between individuals who were naturally HPV uninfected and infected with HPV-16, 18, 6, and 11. Given the cross-sectional design, it is also not known whether infection with specific HPV types affects HPV acquisition or HPV persistence of additional HPV types.
With the recent approval of HPV vaccination in men [33
], these data fill an important gap in our knowledge of the distribution and associations between HPV types in multiple infections among men. Future prospective studies in populations pre- and postvaccination are needed to observe the natural history of multiple infections and assess the long-term potential for HPV type replacement.