Progression and regression of incident cervical HPV 6, 11, 16 and 18 infections in young women

Cervical cancer is the second most common cancer in women worldwide[1] and is caused by human papillomavirus (HPV) infection[2]. Organized screening using the Pap test has reduced cervical cancer rates in industrialized countries[3]. Because Pap testing does not address the root cause of cervical cancer (i.e., HPV infection), screening shifts the clinical burden of HPV disease from treating cancers to managing a large number of precancerous lesions[4, 5]. It has been estimated that approximately 400,000 new cases of cervical intraepithelial neoplasia (CIN) are diagnosed in the United States each year, with annual costs of approximately $2.3 billion for cervical cancer screening and $700 million for CIN treatment[6]. The diagnosis and treatment of CIN can lead to anxiety concerning cancer risk and sexual functioning[7, 8].

Prophylactic vaccines targeting HPV 6/11/16/18[9] and HPV 16/18 infections[10] have recently been shown to reduce the incidence of CIN due to these types. Policymakers worldwide are also evaluating technologies designed to enhance the diagnosis of CIN such as HPV tests and liquid-based cytology[11]. Data on the natural history of HPV infection are therefore critical for the evaluation of these and other emerging technologies.

The present study describes the type-specific progression and regression of incident HPV 6, 11, 16 and 18 infections, along with the observed distribution of these HPV types in CIN grade 1–3 lesions. A unique feature of the present study with respect to prior analyses of HPV infection natural history [12–15] is the correlation of HPV type observed in the incident infection with that detected in cervical lesion tissue specimens. Although the detection of CIN-1 lesions following HPV infection have received less attention than CIN-2/3 due to their lower oncogenic potential, a recent US study reported that more than half of women with incident CIN-1 were treated in clinical practice[16]. Policy models of HPV disease and costs have thus typically considered sub-clinical HPV infections (without clinically detectable CIN-1) as a distinct health state from progression to clinically detectable CIN-1 and we follow this dichotomy in the present study[17, 18]. In this regard, a study of women with active HPV 6, 11, 16 and 18 infections undergoing colposcopy found that CIN-1-3 lesions were detected in 10%, suggesting that most HPV infections are not synonymous with clinically detectable CIN[19].

Among 1,203 women in the placebo arm of the trial who underwent cervical swab HPV PCR testing at enrollment, the prevalence of individual HPV types was as follows: HPV-6 (3.2%), HPV-11 (0.5%), HPV-16 (6.8%), HPV-18 (2.6%). The combined baseline prevalence of HPV 6, 11, 16 and 18 infections was 12.1%.

This analysis has described type-specific progression, persistence and regression of incident HPV 6, 11, 16 and 18 infections, along with the observed distribution of these HPV types in CIN-1-3 lesions. To our knowledge, this is the first analysis to examine the type-specific detection of histologically confirmed CIN-1 following incident HPV infections, and to correlate the HPV type observed in incident HPV infections with that subsequently detected in cervical lesion tissue specimens in estimating progression rates. The perspective adopted is particularly useful for informing HPV disease natural history models used in policy evaluations of emerging technologies[17, 18, 33].

A salient inference from this analysis is that women infected with a given HPV type are often co-infected or subsequently infected with other HPV types which may be responsible for subsequent cervical lesions. This was especially true for CIN-1, where the majority of detected lesions following incident HPV 6, 11, 16 or 18 infections did not correlate with the incident infection HPV type, however was also observed for nearly one-third of CIN-2/3 cases.

Prior studies of the natural history of incident HPV infections have generally not performed HPV typing of CIN lesion tissue specimens and have thus estimated results based on all lesions observed, regardless of HPV type, following detection of a given infection[12, 15]. This perspective can be clinically useful for assessing the overall risk of being subsequently diagnosed with cervical neoplasia for women with HPV infections of a specific type. For instance, Koutsky et al. reported a 24 month risk of CIN-2/3 following incident HPV-6 or -11 infection of 26%[13]. This compares to a figure of 1.9% observed over the same time period in the present study, based on HPV typing of lesion tissue specimens. Thus it is possible that women with incident HPV-6 and -11 infections are clinically at relatively high-risk for CIN-2/3 compared to other women,[15] but that in the majority of cases, the lesions are caused by other HPV types. In general, the exclusion of disease testing positive for HPV types discordant from the incident infection would be expected to yield lower estimates of CIN risk following type-specific infection in the present study, compared to previous analyses. Correlation of HPV typing between infections and lesions can be particularly helpful for modeling studies evaluating HPV vaccines and tests[17, 18], where risks associated with individual HPV types or groups of types are typically of interest.

The reporting of data for clinically detected CIN-1 as a distinct endpoint represents a second difference in analytic method compared to most prior HPV natural history studies, which have focused upon CIN-2/3 development [13–15]. Although CIN-2/3 represents a more proximate precursor to cervical cancer, CIN-1 is also often treated in clinical practice in the US, ¹⁶and results in substantial healthcare costs[6].

In the present study, the initial biopsy result showing evidence of CIN was used for the purposes of denoting progression of infection to clinically detected CIN-1. These women were not eligible to further contribute to the endpoints of CIN-2 or -3 should they have further progressed, and some women with CIN-1 were subsequently treated, thereby preventing progression. In most recent HPV natural history models[17, 18], individual transition probabilities have been utilized to model rapid progression from HPV infection directly to CIN-2 or -3, more deliberate progression of infection to CIN-1, and regression of HPV infection. Additional health state transitions have been used to model progression from CIN-1 to CIN-2, and regression of CIN-1. Data in this study were analyzed in a manner to inform each of the described transitions originating from HPV infection. Other studies describing progression and regression rates for CIN-1 have been reviewed elsewhere [34–36].

Based on this perspective, risks of progression of incident HPV infection to CIN-2/3 from the present analysis are not directly comparable to those reported in prior natural history studies where CIN-1 was not evaluated as a separate endpoint [13–15]. All else equal, relatively lower risks of progression to CIN-2/3 over time would be expected in the present study, as cases of CIN-1 that otherwise may have progressed to CIN-2/3, and been counted as CIN-2/3 endpoints in prior analyses, were truncated at the point of the initial CIN-1 diagnosis. Indeed, a study by Winer et al. reported a 36 month risk of progression of HPV-16 and -18 infections to CIN-2/3 (regardless of HPV type) of 27.2%, well above the 14.0% figure estimated in the present analysis[14].

However, a common finding across studies has been the relatively high risk of progression of HPV-16 and -18 to cervical neoplasia compared to other HPV types [13–15], with nearly one-third of women newly infected with these types subsequently diagnosed with CIN in the present study. Approximately 85% of HPV-6 and -11 infections were observed to regress without resulting in a CIN diagnosis, however these HPV types cause 90–100% of condyloma acuminata, the classic form of genital warts[37, 38].

This study adds HPV typing of 192 CIN-1 and 53 CIN-2/3 lesion tissue specimens to previous US data for 190 CIN-1 and 299 CIN-2/3 lesions respectively[30, 31, 39–43]. The distribution of HPV types 6, 11, 16 and 18 in cervical lesions was generally similar to that observed in aggregate across prior studies. However, a small proportion of CIN-2/3 lesions also tested positive for HPV-6 alone (5.7%) or in combination with HPV-16 (1.9%). Since testing was not conducted for types other than HPV 6, 11, 16 and 18 it is possible that other types were also present in combination with HPV-6. No CIN-2/3 lesions positive for HPV-6 have been reported in prior US studies, however a large US study of nearly 400 invasive cervical cancer tissue specimens by Schwartz et al. revealed a prevalence of HPV-6 alone in 1.8% of cases, and HPV-6 in combination with other types in 3.1% of cases[44]. The presence of HPV-6 in CIN-2/3 lesions has not been examined in prior international reviews and represents an area for further research[45].

A limitation of this study is that although liquid-based cytology was conducted at 6 month intervals, with rigorous follow-up of suspected abnormalities, and women were referred for colposcopy at the conclusion of the study, the detection of some CIN lesions may have been missed or delayed. Thus, progression rates to CIN may be conservative. A prior study of HPV natural history by Winer et al. performed colposcopy at regularly scheduled visits on all women and reported relatively higher risks of progression to CIN-2/3[14]. However, as described previously, because HPV typing was not conducted for lesion tissue specimens, and progression over time to CIN-1 was not separately evaluated, all else equal, the risk of progression of specific HPV types to CIN-2/3 would be expected to be somewhat liberal for disease modeling purposes. In the absence of a single study overcoming the limitations of both analyses, in our own HPV infection modeling work[46], we have chosen to average estimates obtained from the Winer and present study to account for the divergent effects of these limitations.

Also, as previously noted, testing data were not available in this study population for additional HPV types beyond HPV 6, 11, 16 and 18. Thus, it was not possible to fully characterize multi-type HPV infections in lesions where present or explore the natural histories of other HPV types.

Finally, women in the clinical trial were required to have 5 or fewer lifetime sexual partners at enrollment. In the sub-sample of women examined in this analysis that developed incident HPV 6, 11, 16 or 18 infections over the course of follow-up, the median number of lifetime sexual partners at the time of infection was five. There are not published nationally representative US data on the lifetime number of sexual partners among women developing incident HPV infections from which to make comparisons of external generalizability. However, in a further investigation of this potential limitation, we found through Cox regression analysis that the risk of developing CIN among women infected with HPV 6, 11, 16 or 18 did not vary by the lifetime number of sexual partners at the time of infection (hazard ratio = 1.04; 95% CI 0.98–1.11).

Over the past two decades there has been substantial research demonstrating the causal link between HPV infection and cervical neoplasia[47], and the relatively high proportion of cervical cancers associated with certain HPV types, particularly types 16 and 18[48]. With the arrival of new technologies, such as HPV tests and vaccines targeting specific HPV types, there has also arisen a need to accurately model the natural history of infection for individual or groups of HPV types for policy evaluations[9, 10, 17, 18]. Epidemiologic studies to date have generally not been specifically designed to evaluate HPV infection or disease natural history in a manner ideal for infectious disease modeling. This study has contributed additional data for modeling outcomes of incident HPV 6, 11, 16 and 18 infections. However, further studies with rigorous disease ascertainment would be helpful, particularly for incident infections caused by other HPV types, and the progression and regression of incident cervical neoplasia caused by specific HPV types.