Data on human papillomavirus (HPV) type distribution in women with invasive cervical cancer (ICC) and its precursor lesions are essential to predict the potential worldwide impact of new prophylactic vaccines against HPV16/18,1, 2 as well as to determine priorities for the inclusion of HPV types in future HPV vaccines and HPV-based screening tests.
A standardised pooled analysis of 3,607 ICC cases3 and a wider meta-analysis of 10,058 ICC cases4 both confirmed that a majority of worldwide ICC cases are associated with HPV16/18. They also suggested some geographical variation in the importance of specific HPV types,3, 4 although data were limited or missing from many regions in Africa and Asia.
A further meta-analysis in 4,338 high-grade squamous intraepithelial lesions (HSIL) showed that the most common HPV types in HSIL were broadly similar, but not identical, to those in ICC.5
The purpose of the present publication is to update previous meta-analyses of HPV type distribution in ICC and HSIL with studies published between January 2002 and January 2006, including many from previously under-studied regions, and to identify remaining worldwide epidemiological data gaps prior to HPV vaccine implementation.
Material and methods
The detailed methods used for this meta-analysis of type-specific HPV prevalence have been reported previously, and are similar for both ICC and HSIL.4, 5 In brief, Medline was employed to search for citations published from January 2002 to January 2006 using the following MeSH terms: “cervical cancer”, “cervical intraepithelial neoplasia”, “HPV”, “human”, “female” and “polymerase chain reaction”. Additional relevant references cited in retrieved articles were also evaluated. Included studies had to meet all of the following criteria: (i) use of polymerase chain reaction (PCR)-based technology to detect HPV DNA, (ii) inclusion of at least 20 cases of ICC or HSIL and (iii) reporting of type-specific prevalence for at least one HPV type other than 6, 11, 16 and 18.
For the purposes of this study, ICC refers both to squamous cell carcinoma (SCC)/unspecified histology (85% of included cases) and adeno/adenosquamous carcinoma (ADC, 15%). HSIL refers both to cytologically-detected lesions as classified by the Bethesda system (44% of included cases), and those reported as histologically diagnosed cervical intraepithelial neoplasia (CIN)2 (17%), CIN3 (37%), or carcinoma in situ (2%).
For each included study, the following key information was extracted: country of sample; sample size, distribution of cancer cases by histological type (SCC/unspecified or ADC), HPV DNA source (fresh/fixed biopsies or exfoliated cells), PCR primers used to detect HPV-positive samples, and type-specific and overall prevalence of HPV DNA. These data are presented study by study in the Appendices. Studies were additionally classified into 17 geographical regions.
Crude type-specific prevalence is presented for 19 HPV types, including the 18 most common HPV types as identified by the previous ICC meta-analysis,4 namely HPV6, 16, 18, 31, 33, 35, 39, 45, 51, 52, 56, 58, 59, 66, 68, 70, 73 and 82. HPV11 was also considered on account of its inclusion in currently available quadri-valent HPV vaccines.2 Each HPV type was evaluated independently of all others. All studies provided information on HPV16, but for other types, prevalence was estimated only among those studies testing for the HPV type in question, and thus denominators of prevalence estimates vary by type. Type-specific prevalence thus includes that in either single- or multiple-type HPV infections.
Type-specific HPV prevalence was compared between HSIL and SCC by prevalence ratios, adjusted for continent (Africa, Asia, Europe, North America, Oceania, South/Central America), with corresponding 95% confidence intervals (CIs).
One hundred thirty ICC and 85 HSIL studies met inclusion criteria, including a total of 14,595 ICC and 7,094 HSIL cases (Table I). This constituted a gain of 4,537 ICC and 2,756 HSIL cases from previous meta-analyses.4, 5 The proportional gain in cases was greatest for Africa (from 609 to 1,339 ICC cases) and Asia (from 3,091 to 5,652 ICC cases, Table I). Thirty-eight percent of ICC cases came from Asia, 30% from Europe, 10% from South/Central America, 9% from North America, 9% from Africa and 3% from Oceania. The equivalent regional proportions were 19%, 49%, 12%, 15%, 4% and 1%, respectively, for HSIL cases (Table I).
Table I. Geographic Distribution of Studies and Cases with Type Specific Human Papillomavirus DNA Typing for Invasive Cervical Carcinoma (ICC) and High-Grade Squamous Intraepithelial Lesions (HSIL)
Country for which additional ICC cases have been gained since Clifford et al., 2003 [ref.4].
Country not previously represented with ICC cases in Clifford et al., 2003 [ref.4].
Country for which HSIL data only is available.
Continents do not add up to total due to multi-centric studies.
Overall HPV prevalence in ICC was 87%, ranging from 86% to 94% by region. The 8 most common HPV types in ICC are shown by continent in Figure 1a. HPV16 was the most common type (ranging from 52% in Asia to 58% in Europe) and HPV18 the second most common type (ranging from 13% in South/Central America to 22% in North America) in ICC cases from all continents studied. HPV16/18 prevalence was thus 70% overall, and varied from 65% in South/Central America to 76% in North America. The next most common HPV types in ICC were also the same in each continent, namely HPV31, 33, 35, 45, 52 and 58 (with the slight exception of Europe, where HPV56 was the eighth most common type instead of HPV52), although their relative importance differed somewhat by continent. HPV58 and 52 prevalence was notably higher in ICC cases from Asia (5.6% and 3.8%, respectively). HPV types other than these 8 accounted individually for no more than 2% of ICC cases from any continent. Appendix I summarises the prevalence of the 19 most common types and the prevalence of multiple infections in ICC (although the frequency of multiple infections depends largely on the number of HPV types tested for within a given study).
The breakdown of HPV type-specific prevalence data by histological type of ICC was available for 9,494 SCC and 1,949 ADC cases. Among these cases, overall HPV positivity was higher in SCC (90%) than in ADC (85%). The 8 most common HPV types in ICC are compared by histological type in Figure 2. HPV16 was significantly under-represented in ADC (33%) compared with SCC cases (55%), as were HPV31, 33, 52 and 58. Conversely, HPV18 was significantly over-represented in ADC (37%) compared with SCC (13%).
Overall HPV prevalence in HSIL was 85%, ranging from 78% in Asia to 88% in Europe. Combined HPV16/18 prevalence in all HSIL cases was 52%. Overall, the 8 most common HPV types in HSIL (Fig. 1b) were largely similar to those in cervical cancer (Fig. 1a), except for a noticeable absence of HPV45. HPV16 was the predominant type in HSIL from all continents studied, varying from 34% in Asia to 52% in Europe. The prevalence of the 19 most common types and multiple infections (see note on multiple infection prevalence above) in HSIL is summarised by study and by region in Appendix II.
HPV type-specific prevalence among 7,094 HSIL cases was formally compared with that among 9,494 SCC cases using prevalence ratios (Table II). HPV16 was significantly less prevalent in HSIL (45.3%) than in SCC (55.2%) (SCC:HSIL prevalence ratio 1.30, 95% CI: 1.26–1.34), as were both HPV18 (1.76, 95% CI: 1.58–1.95) and HPV45 (1.54, 95% CI: 1.20–1.98). Conversely, HPV types 31, 33, 35, 39, 52, 56, 58, 68 and 73 were each 2- to 3-fold more prevalent in HSIL than ICC, and HPV types 6, 11, 51, 66, 70, and 82 were 5- to 10-fold more prevalent in HSIL than ICC.
Table II. Comparison of Human Papillomavirus (HPV) Type Distribution in Squamous Cell Carcinoma (SCC) Versus High-Grade Intraepithelial Lesions (HSIL)
Type-specific prevalence includes that in single or multiple infections.
Prevalence ratio adjusted for continent. CI = confidence interval.
Figure 3 maps the availability of HPV-typed ICC and HSIL cases by country, and highlights the limited amount of ICC and HSIL data from most countries in Africa, Central/Eastern Europe, and Western and South-Central Asia. No data were available from Melanesia/Micronesia/Polynesia, and were very limited for the Caribbean.
This present study represents a 50% gain in information on HPV type distribution in ICC over a previous meta-analysis.4 This gain in information was particularly high for Africa and Asia, making overall estimates of HPV type distribution from these continents more robust and representative than before. Overall, 70% of ICC cases were associated with either HPV16 (55%) or 18 (15%) infection. The 6 next most common types, namely HPV31, 33, 35, 45, 52 and 58 accounted for an additional 18% of cases.
Whilst these overall findings are similar to previous estimates,3, 4, 5 interpretations by continent have altered somewhat. With the accumulation of additional published data, differences by continent became less pronounced than those noted in previous analyses,3, 4, 5 with the estimates for the proportion of ICC cases attributable to HPV16/18 in Africa, Asia and South/Central America all increasing from the previously reported 59–64%.4 However, the most relevant variation by continent in HPV type distribution noted by the previous meta-analysis4 remained evident, namely that although HPV16/18 are the most common types in ICC from all continents, the HPV16/18 proportion is higher in Europe, North America and Oceania (74–77%) than in Africa, Asia and South/Central America (65–70%).
An additional effect of the gain in data from less developed countries is that the 6 most important types in ICC after HPV16/18, namely HPV31, 33, 35, 45, 52 and 58, now appear to be the same in all continents. The consistency in the 8 most common types is therefore clearer than in previous estimates,3, 4 likely due to a reduction in random fluctuation. Future generations of prophylactic vaccines would benefit by including high-risk types other than 16 and 18. However, if it is not possible to include all 6 additional types in next generation vaccines for ICC prevention, then there do remain some differences by continent in the priorities for HPV types. In particular, the prevalence of HPV58 and 52 is relatively high in ICC from Asia. This pattern was also seen in HSIL, and from a previous meta-analysis of low-grade squamous intraepithelial lesions.7
Data on HSIL have greatly increased since a previous report5 and the 8 most common HPV types in HSIL were broadly similar to those in ICC. However, the HSIL data from less-developed countries remain scarce as HSIL detection requires active cervical screening, which is rare in such settings. Despite these limitations, it is clear that HPV type distribution in HSIL is not entirely representative of that in ICC. In agreement with similar previous reports,5, 8 HPV types 16, 18 and 45 were significantly over-represented in ICC compared to HSIL, and all other high-risk types were significantly under-represented, suggesting type-specific differences in propensity to progress from HSIL to cancer. Alternatively, HPV18-positive HSIL may be preferentially missed by cervical screening programmes.9 Thus, whilst data on HPV type distribution in HSIL is relevant to predicting the impact of HPV16/18 vaccines on reducing screen-detected HSIL, it underestimates the impact on ICC.
Limitations associated with cross-sectional meta-analyses of HPV type-specific prevalence include variation in HPV type-specific sensitivity of different PCR protocols used in the included studies,10 the fact that many studies did not type for a broad range of HPV types, and the lack of standardisation of cytohistological diagnoses across studies. Thus, the observed overall HPV DNA prevalence of 87% in ICC is lower than the 99.7% found by gold standard HPV detection techniques.11 Furthermore, type-specific prevalence estimates include those in both single- and multiple-type HPV infections. This means that multiple infections count at least twice when summing the prevalence for more than one type, and that some types, e.g., HPV6 and 11, may exist exclusively in multiple infections. These limitations have been previously described in detail.4
We assumed that the proportion of ADC in the included studies, which varied from 4% in Africa to 32% in North America, was representative of underlying ICC in the continent. Previously identified differences in HPV type distribution by histology, namely HPV16 being more prevalent in SCC than in ADC and HPV18 being more prevalent in ADC than in SCC,4 were clearly apparent from this analysis. Differences by histological type may contribute to some of the heterogeneity in HPV type distribution by continent and also to some of the differences between ICC and HSIL, given that HSIL, but not ICC, includes, by definition, only squamous cell lesions.
Although this study is the broadest summary of HPV type distribution in ICC and HSIL worldwide to date, the cases included in the overall estimates were far from being geographically representative of the worldwide cervical cancer burden. More than half of all HPV-typed ICC cases came from North America, Eastern Asia and Northern Europe alone. Data were relatively scarce from South-Central Asia, including India, where approximately one-quarter of all worldwide cervical cancer occurs.6 Furthermore, while data has increased for Africa as a whole, they remained missing for vast regions of the continent, where cervical cancer incidence rates are estimated to be among the world's highest.6
As more data are accumulated, it is reassuring to observe that HPV16/18 account for at least two-thirds of ICC in all continents, suggesting that it is not essential to collect data from every country. On the other hand, local decision makers may seek assurance of the relevance of a HPV16/18 vaccine to their local ICC burden, especially as data aggregation may also mask some intra-continental heterogeneity. In Asia for example, the proportion of ICC due to HPV16/18 in the few studies from South-Central Asia appears as high as that in Europe and North America, but is lower in the studies from Eastern Asia. For the comparison of smaller geographical groupings, HPV type-specific information is given for individual studies and by region in the Appendices.
In summary, this meta-analysis suggests that a prophylactic vaccine against HPV16/18 has the potential to prevent more than two-thirds of worldwide ICC and half of HSIL. These proportions may be even higher if cross-protection against other high-risk HPV type infections, as recently reported for one of the two HPV16/18 vaccine candidates,1 also proves to be relevant for preventing cancer and HSIL associated with these types.
We thank Ms. Trudy Perdrix-Thoma for editorial work on the manuscript, and Ms. Annie Arslan for her data management assistance.
Table APPENDIX I. Type-Specific Human Papillomavirus (HPV) Prevalence among Women with Invasive Cervical Cancer, by Study and Region
Table APPENDIX II. Type-Specific Human Papillomavirus (HPV) Prevalence among Women with High-Grade Cervical Intraepithelial Lesions, by Study and Region