In this cross-sectional analysis of cervical histology in HIV+/HPV+ Rwandan women we found that immunosuppression as measured by CD4 count was significantly associated with CIN3+ only in women who harbored carcinogenic HPV genotypes other than HPV16. We and others have reported that HPV16 prevalence and incidence are not associated with CD4 count 
, suggesting that HPV16 may be more able than other HPV genotypes to avoid immune surveillance in immune competent women. Immunosuppression may differentially affect the risk of progression to CIN3 and cancer of non-HPV16 carcinogenic HPV genotypes versus HPV16, perhaps mediated directly through the likelihood of HPV viral persistence. However, we had inadequate numbers (65) of women testing positive for HPV16 to allow direct analysis of the association of CD4 count with CIN3+ in HIV+ women with HPV16 genotype.
Compared to testing positive for non-HPV16 carcinogenic HPV, CIN3+ was inversely associated with testing positive for only non-carcinogenic HPV and positively associated with testing positive for HPV16. Notably, the fraction of women with CIN3+ who tested positive for HPV16 was less than expected (33% in this study versus 50–60% in other studies) and the strength of association and magnitude of the odds ratio for the association of HPV16 with CIN3+ were smaller than anticipated. This is further albeit indirect evidence that the relative effects of HPV16 in the development of cervical cancer may be less enhanced in HIV-infected compared to uninfected women, consistent with our prior report 
. The other AIDS-defining cancers, Kaposi's Sarcoma and large B-cell lymphoma, have demonstrated dramatically decreased incidence since the advent of effective antiretroviral therapy (ART) 
, which reverses much of the immune compromise of HIV infection. However, the HPV-related cancers, invasive cervical cancer and anal carcinoma, have in most reports not decreased in incidence in the era of effective ART 
. This may be in part due to the lack of impact of immune suppression with HPV16 natural history, which in most populations causes at least 50% of cervical cancers.
We found that recent malarial infections had a significant positive association with CIN3+ in HIV+/HPV+ women. To our knowledge, this is the first report of any evidence that malarial infections might contribute to the risk of cervical precancer and cancer. The mechanisms by which malarial infections could increase the risk of CIN3+ are unknown, but might include additional challenges to the already compromised immune system of HIV+ women. HIV, HPV and malarial infections are highly prevalent in much of Africa, and any increased risk of CIN3+ associated with malaria could result in a substantial attributable risk. However, given the limited sample size in this analysis, this observation needs to be confirmed in other populations.
Several factors previously identified as associated with cervical precancer and cancer in general populations of women were also identified here in HIV+/HPV+ Rwandan women. We found that multiple pregnancies (7 or more) were associated with increased risk of CIN3+, independent of number of reported sexual partners. High parity has been reported as a risk factor for cancer in populations of unknown HIV serostatus 
; the mechanism by which parity increases the risk is unknown but may be related to endogenous hormones causing further eversion of the squamocolumnar junction and/or tissue damage and inflammation resulting from birthing. A higher number of sexual partners was also associated with CIN3+, perhaps reflecting greater likelihood of multiple HPV exposures and thus stochastically to greater opportunity for HPV persistence and progression to CIN3+.
Our study has several limitations. It is a cross-sectional analysis and temporal or causal relationships cannot be inferred. The lack of follow-up does not allow observation of progression with specific HPV types. An additional limitation of this study is that HPV+ women without clinical indication of abnormal cytology or visual lesions were not colposcoped and therefore not biopsied. Thus visually unapparent, HPV+ high-grade lesions could have been missed 
, resulting in verification bias. We expect that this may have muted the strength of association by misclassifying some cases of CIN3+ as controls. Potential misclassification of women as HPV+ may have resulted from the use of CVL rather than tissue to determine HPV infection, thus eliminating from the population women who were in fact HPV infected. It may also have led to over-representation of non-carcinogenic HPV genotypes that preferentially infect the vagina 
. Further, cross-sectional detection of HPV is only a proxy for lifetime level of exposure. The smaller magnitude of the association of CIN3+ with CD4 count may not have been sufficient to achieve statistical significance because of the small number of women with HPV16. We also could not distinguish prevalent from incident infection. This may lead to underestimating the association of HR-HPV with CIN3 and cancer since prevalent infections represent a mixture of incident and persisting infections, the latter of which are on the causal pathway to cervical precancer and cancer.
In conclusion, we found that CIN3+ was inversely associated with CD4 cell count in HIV+ women co-infected with non-HPV16 carcinogenic HPV. The implication of these data is that a greater proportion of cervical precancer and cancer may be attributable to carcinogenic HPV genotypes other than HPV16 in HIV+ women compared to HIV-negative women but larger studies are needed to evaluate this directly. We also found some evidence that malarial infections may increase the risk of CIN3+; this warrants confirmation in larger population-based studies.