Baseline assessment of prevalence and geographical distribution of HPV types in Chile using self-collected vaginal samples

The availability of highly effective human papillomavirus (HPV) vaccines has led to a discussion about their incorporation in public health vaccine programs [1, 2]. A baseline characterization of HPV types in cervical cancers and also in the general population is required before deciding whether to introduce the vaccine. With this information on the population genotype prevalence, together with regional specific genotype distribution in cervical lesions and cervical cancer, the potential impact of the vaccine can be estimated. In a previous study of HPV prevalence in cervical specimens of a population-based sample of women, we explored the individual risk factors for low-risk and high-risk HPV infections, confirming that in Chile the main risk factor was the number of sexual partners for both high and low risk HPV [3]. This study aims to describe the geographical distribution of the HPV types in the national community. Epidemiologic studies of the prevalence of HPV in the community have typically used cervical samples obtained by a gynaecological examination [4]. Most women find this exam to be uncomfortable [5]. Additionally, in countries where Papanicolaou (PAP) screening is not widely accepted, studies based on gynaecological sampling may not represent the general population. For community surveillance, public health officers will need non-invasive methods that are acceptable to a representative sample of women. One option is the self-obtained vaginal sample. Various well-designed trials have demonstrated that HPV detection using self-obtained vaginal sample is a reasonable surrogate for the identification of HPV types that are infecting the cervix, with moderate to excellent agreement with the detection of HPV from cervical samples obtained by clinicians [6–10]. Agreement was better for the high-risk HPVs [6, 8, 11, 12]. Studies comparing acceptability of methods demonstrated that most women preferred self-vaginal sampling [5, 8, 13–19]. We assessed the prevalence and distribution of HPV types throughout Chile using self-obtained vaginal sampling in the community. The National Health Survey, ENS 2003, the first prevalence study of chronic diseases, provided a population-based national sample of non-institutionalized people aged 18 and above [20]. This is one of the first studies with self-obtained vaginal sampling in a health survey for chronic conditions in the adult population. The aims were to assess the magnitude of the problem and to explore the association of HPV prevalence with some regional characteristics.

Chile is a long (6,435 km length) and narrow (200 km wide) country (Figure 1) with a climate that goes from the driest desert in the world, with a mean temperature of 18.1°C and 0 mm of rain/year in the extreme north (Region 1) to a cold, windy and rainy (3,500 mm yearly) climate in the far south (Region 12). The population in the northern regions resides mostly in urban areas, with a racial admixture of Hispanic and Andean Indians. Its main economic activity is copper mining. Most of the Chilean population lives in the central regions. The metropolitan area (Region 13) is located between Regions 5 and 6; the climate is temperate with an average of 15°C and an average annual rainfall of 150–200 mm. Its main economic activity is industry and wine production. The population is an admixture of Spanish with native Mapuche Indians. The southern regions are mountainous, with many lakes and rivers, and large rural population. Average rainfall is 1,500–2,000 mm/year, with a temperature range of 2°C-23°C. This area hosts the largest ethnic Mapuche population and is the poorest in the country. Its main economic activity is agriculture and livestock production. The far southern regions have a cold climate and their principal activity is fisheries; its native population is the lowest in the country.

Surveillance of adult health status and its risk factors has become an essential tool for public health planning and evaluation. To date this surveillance has focused on cardiovascular diseases and its risk factors and, therefore, this is one of the first surveys to incorporate HPV testing as one of the priority problems for adult health surveillance. The overall prevalence of any HPV was strikingly similar to a recent US national survey using similar methods; however we did note an interesting contrast between high-risk and low-risk HPV prevalence and regional variability that was not previously reported [27].

In the multivariate analysis, residing in the north of Chile was a risk factor for high-risk HPV infection. The relative contribution of high risk HPV to the total burden of HPV infection was almost 3 times higher in the north than the national average while in the south it was almost half the national average [OR 2.8 (95% CI, 1.4–5.7) and 0.6 (95% CI, 0.3–0.98), respectively]. The relative excess of high-risk types in the north appears to be associated with a younger age and with variations in factors associated with sexual behaviour. While we could not explore sexual behaviour directly, secondary data shows that northern regions have higher rates of smoking, higher birth rates and higher rates of HIV while they are less rural and have less poverty than the southern regions (data not presented). On the other hand, the excess of low- risk HPV in the south appears to be associated with older age, poverty and rurality. Other authors have also described a diverse epidemiological profile for low and high risk HPVs [4, 28]. To explain the differences in the relative distribution of high and low risk HPV, some authors have suggested an antagonism between high-risk and low-risk HPVs. In a serologic case-control study, Luostarinen, et al. [29] describe a statistically significant null association with cervical cancer risk among women who are seropositive to both HPV 6/11 (OR = 1.0) and HPV 16, relative to HPV 6/11 alone (OR = 2.2) or HPV 16 alone (OR = 5.5). Similarly, Wheeler, et al. [30] observed that women who had no HPV had a five times higher probability of acquiring HPV 16 in the following two years than women who carried low-risk HPVs (authors' calculation from Table 2 of Wheeler 2006). These reports, coupled with our ecologic observational data of an inverse correlation between low and high risk HPV infections, not entirely controlled by age adjustment, point to possible immunological interactions between the cervicovaginal HPV flora that require further investigation.

The age-specific prevalence of HPV were the same in the current survey and in our previous study, with high-risk HPV prevalence peaking below 25 years old while low-risk HPV increased slowly after age 50 [3]. The increase in the prevalence of low-risk HPV in older ages has been reported in some studies in Latin America and Spain [31–33] and confirmed in a recent meta-analysis [34], whereas others have found a flat age curve for low-risk HPV [25, 35, 36]. Some authors have attributed the excess of low-risk HPV in older women to the elimination of high-risk HPV by the screening programs [35]. It is intriguing that while low-risk HPV incidence at younger ages is half as frequent and in some reports is undetectable twice as fast as high-risk HPV [37–39], it increases in frequency at older ages both in screened and unscreened women [40]. A recent study by Wheeler found the same 2 year clearance rate for low-risk (36%), high-risk except 16 (41%) and HPV 16 (34%) (calculated from Table 3 in Wheeler 2006). The possibility that low-risk HPV in adult populations results from frequent reactivation or re-infection should be considered in view of these data.

When comparing the HPV types found in these vaginal samples with the published data on HPV types isolated from cervical samples in Chile's general population [3] we found a remarkable coincidence in the distribution of high-risk HPV types (Fig 5) in vaginal and cervical swab samples. However, we found no similar correlation in the most frequent low-risk types (Figure 6). HPV DNA positivity was assessed in the cervical sample using a general primer-mediated GP5+/6+ polymerase chain reaction (PCR) and an Enzyme Immune Assay (EIA), to detect 36 HPV types (Ferreccio 2004), while in the current study, as already described, DNA was extracted and typed using PGMY consensus PCR and line blot assay to 37 HPV genotypes. However, these methods have been shown to have similar test characteristics, with difference in the detection of only 3 types: 35, 53 and 61 [41]. These results suggest that while self-collected vaginal swabs may result in higher low-risk HPV prevalence estimates compared to cervical swab samples, the concordance for high-risk, clinically relevant HPV is high [9, 12, 42].

Self-sampling was highly acceptable for Chilean women, as has been repeatedly demonstrated in different cultural settings in Canada [18], USA [16], Germany [13], Mexico [5], Brazil [19], Africa [8] and China [43]. Collectively, these data suggest that vaginal self-collected swabs could be effectively used for primary cervical cancer screening via HPV DNA testing. This application was particularly useful for those women who had never been screened, 70% of whom agreed to self-sample in our survey.

There are several caveats in the interpretations and generalizations based on this population-based survey of HPV type-specific prevalence. The relatively low response rate may bias the results. Nevertheless, since the non-participants are weighted, part of the bias may have been removed. We estimated the bias by correcting the estimates based on propensity scores, but the estimated prevalence changed minimally (Vives A. manuscript in preparation), therefore we believe that these biases do not significantly change our inferences.

The most important limitation in estimating the real HPV burden is the fact that we are measuring HPV at a single point in time, with no information about the historical exposure of these women to HPV. Winer and colleagues [44] estimated a cumulative HPV incidence of 32% after 2 years of follow-up among young college women. HPV seroprevalence (cumulative lifetime exposure estimate) have been generated from a population survey [45, 46], with an average 15% – 25% cumulative exposure to HPV 16 alone. Compared to the 3.2% HPV 16 DNA observed in this population, clearly cervicovaginal HPV DNA point prevalence is not a valid marker of ever having been exposed to the virus.

Self-obtained vaginal sampling is an adequate method for monitoring the burden of HPV in the community. Coupling this sample collection with routine census procedures allows high-risk areas to be identified in order to prioritize interventions and to carry out routine surveillance to assess the impact of the interventions (e.g., HPV prophylactic vaccination) over time.