Conceived and designed the experiments: AC AB DG MAP. Performed the experiments: AC AB DG MAP. Analyzed the data: AC AB DG LG MAP. Contributed reagents/materials/analysis tools: AC AB DG LG MAP. Wrote the paper: AC AB DG MAP.
The authors have read the journal's policy and have the following conflicts: The study was supported by an independent grant from GlaxoSmithKline and the Institute for Clinical Effectiveness and Health Policy - IECS. The authors have declared that no other competing interests exist. This does not alter the authors' adherence to all the PLoS ONE policies on sharing data and materials.
Cervical cancer is a major public health problem in Latin America and the Caribbean (LA&C), showing some of the highest incidence and mortality rates worldwide. Information on HPV type distribution in high-grade cervical lesions (HSIL) and invasive cervical cancer (ICC) is crucial to predict the future impact of HPV16/18 vaccines and screening programmes, and to establish an appropriate post-vaccinal virologic surveillance. The aim was to assess the prevalence of HPV types in HSIL and ICC in studies in LA&C.
We performed a systematic review, following the MOOSE guidelines for systematic reviews of observational studies, and the PRISMA statement for reporting systematic reviews and meta-analyses. Inclusion criteria were at least ten cases of HSIL/ICC, and HPV-type elicitation. The search, without language restrictions, was performed in MEDLINE, Cochrane Library, EMBASE, LILACS from inception date to December 2009, proceedings, reference lists and consulting experts. A meta-analysis was performed using arc-sine transformations to stabilize the variance of simple proportions. Seventy-nine studies from 18 countries were identified, including 2446 cases of HSIL and 5540 of ICC. Overall, 46.5% of HSIL cases harbored HPV 16 and 8.9% HPV18; in ICC, 53.2% of cases harbored HPV 16 and13.2% HPV 18. The next five most common types, in decreasing frequency, were HPV 31, 58, 33, 45, and 52.
Study's limitations comprise the cross-sectional design of most included studies and their inherent risk of bias, the lack of representativeness, and variations in the HPV type-specific sensitivity of different PCR protocols.
This study is the broadest summary of HPV type distribution in HSIL and ICC in LA&C to date. These data are essential for local decision makers regarding HPV screening and vaccination policies. Continued HPV surveillance would be useful, to assess the potential for changing type-specific HPV prevalence in the post-vaccination era in Latin America.
Human papillomavirus infection (HPV) is one of the most common sexually transmitted diseases worldwide
It is now recognized that virtually all cervical cancers (both the squamous and adenocarcinoma histological types) and their precursor lesions are causally related to cervical infections through at least 14 oncogenic HPV genotypes (16, 18, 31, 33, 35, 39, 45, 51, 52, 56, 58, 59, 66 and 68)
We performed a systematic review, following the Meta-analysis Of Observational Studies in Epidemiology (MOOSE) guidelines
A search, without language restrictions, was performed on the main international and regional literature databases MEDLINE; EMBASE; CINAHL; NHS R&D Health Technology Assessment Program; ClinicalTrials.gov; LILACS; Cab International Global Health; Pascal Biomed; generic and academic internet search and meta-search engines; and the specialized register of the Cochrane Gynecological Cancer Group from its inception date to December 2009. Databases containing regional proceedings or congress's annals, doctoral theses and experts were also consulted.
The Medline, LILACS, and EMBASE search strategy is available at the
Any descriptive epidemiological study with individual-level data was considered. Participant subjects were women from LA&C countries, in studies of cervical cancer/HSIL associated with HPV. The inclusion criteria were a) to inform at least ten cases of HSIL or ICC, b) confirmed by biopsy, and c) HPV-type elicitation. We excluded those papers that undoubtedly failed to meet the aforementioned inclusion criteria. Studies using both polymerase chain reaction (PCR)-amplified and non-amplified genotyping methods were included. There were no restrictions on PCR primers' utilization. HPV DNA tissue sources included fixed or fresh biopsies and/or exfoliated cells. Outcome measures included global and type-specific HPV prevalence. Two attempts of email contact with the author were made in order to recover missing data.
Two reviewers assessed the methodological quality of studies independently. Discrepancies were solved by consensus of the whole team. Observational studies or control arms of randomized controlled trials were assessed by a checklist of essential items stated in STROBE
Pairs of reviewers independently abstracted the following key information: country where the samples were drawn, setting, population, sample size, study design, age, study year, distribution of cases by histological type, type of cervical specimen and PCR primers, type-specific and overall prevalence of HPV infection, reported duration of HPV infection, and quality score. Data on HPV-specific prevalence was extracted independently for squamous cell carcinoma (SCC) and for adeno- and adenosquamous carcinoma. Each study, or regional components of a study, was classified by the following criteria: 1) geographical region (Central America/Mexico/The Caribbean or South America) 2) income level as defined by the Gross Nation Income (GNI) World Bank Classification (lower-middle income, upper-middle income, high income), 3) tissue source (exfoliated cells, fixed biopsies, fresh biopsies, combined), and 4) genotyping method (Southern blot, Dot blot, FISH and In Situ Hybridization), PCR 1 (PCR MY09/11 or Consensus primers), PCR 2 (PCR SPF, GP5/6, E6, E7 and others) and PCR 3 (PCR MY and GP performed together).
HPV prevalence data was expressed as a percentage of all cases tested for HPV. Multiple infections were separated into constituent types, thus type-specific prevalence represents both single and multiple infections. For HPV type-specific prevalence, only studies testing for a particular HPV type contribute to the analysis for that type, and therefore sample size varied between the type-specific analyses. In order to perform a meta-analysis with prevalence data, we first transformed proportions into a quantity (the Freeman-Tukey variant of the arcsine square root transformed proportion)
One must consider that each HPV type proportion is a pooled estimate of only those studies reporting the particular HPV type. Hence, each proportion has its own denominator and must be considered regardless of the other types. Thus, cumulative point estimates do not sum to 100%. DerSimonian-Laird weights for the random effects model
We hypothesized the following possible sources of heterogeneity: age, risk factors of HPV and/or HSIL/cervical cancer, country, geographical region, income level by the Gross National Income (GNI) World Bank Classification, type of cervical lesion, type of tissue source and type of genotyping method used. With the available data we could perform pre-designed subgroup analyses considering the country where the study was carried out, the geographical region, the income level of the country according the Gross National Income (GNI) World Bank Classification, the type of genotyping method and the tissue source. Additionally, we applied a meta-regression analysis in order to further study the possible sources of heterogeneity and to get the adjusted prevalence. Publication bias was unlikely as assessed by funnel plots although this type of bias is unlikely to occur in prevalence studies (data not shown). No ethical approval was required for this study.
The present Systematic Review and Meta-analysis met the PRISMA statement requirements (See
Overall, 1452 citations were retrieved from the search strategy. After the assessment (
We considered 114 sub-studies for the analysis, including seven country-level sub-studies from Bosch 1995
HSIL/ICC cases came mainly from Brazil (23.7%), Argentina (19.0%), and Mexico (17.9%). The HSIL and ICC prevalence, and ICC∶HSIL prevalence ratio by type are presented in
HPV TYPE | HSIL | CANCER | CANCER∶HSIL | ||
N° of patients | Prevalence % | N° of patients | Prevalence % | Prevalence | |
(N° of Studies) | (95% CI) | (N° of Studies) | (95% CI) | ratio | |
|
|
|
|||
|
1749 (36) | 82.5 (77.3–87.1) | 3435 (43) | 89.0 (84.3–92.9) | 1.08 |
|
1415 (29) | 4.2 (2.2–6.7) | 2274 (32) | 1.7 (0.9–2.8) | 0.4 |
|
1414 (29) | 2.4 (1.3–3.8) | 2274 (32) | 1.3 (0.5–2.5) | 0.54 |
|
2327 (49) | 46.5 (41.3–51.7) | 5463 (60) | 53.2 (49.1–57.2) | 1.14 |
|
2194 (45) | 8.9 (6.3–11.8) | 4962 (56) | 13.2 (11.0–15.6) | 1.48 |
|
1785 (36) | 8.0 (6.0–10.4) | 3903 (45) | 7.5 (5.5–9.8) | 0.94 |
|
1722 (35) | 6.5 (4.7–8.5) | 3821 (42) | 4.3 (3.2–5.5) | 0.66 |
|
1228 (24) | 3.0 (1.9–4.4) | 2332 (31) | 2.0 (1.3–2.7) | 0.67 |
|
885 (20) | 2.4 (1.5–3.5) | 1977 (27) | 1.8 (1.3–2.4) | 0.75 |
|
1077 (24) | 3.9 (2.8–5.2) | 3389 (37) | 4.6 (3.5–5.7) | 1.18 |
|
1013 (21) | 3.7 (2.1–5.7) | 2131 (30) | 2.1 (1.1–3.3) | 0.57 |
|
1152 (25) | 4.9 (2.9–7.4) | 2544 (34) | 3.2 (2.1–4.4) | 0.65 |
|
892 (19) | 2.4 (1.5–3.4) | 2155 (28) | 1.2 (0.8–1.7) | 0.5 |
|
1197 (26) | 8.7 (6.0–11.9) | 2564 (34) | 3.0 (2.1–4.1) | 0.34 |
|
954 (21) | 1.9 (1.2–2.9) | 2199 (30) | 1.6 (1.1–2.2) | 0.84 |
|
926 (20) | 1.8 (1.1–2.8) | 2095 (28) | 1.1 (0.7–1.6) | 0.61 |
|
619 (14) | 1.3 (0.6–2.3) | 1864 (23) | 0.5 (0.3–0.9) | 0.38 |
|
1479 (32) | 11.6 (7.6–16.2) | 3177 (34) | 7.5 (5.0–10.4) | 0.65 |
|
1431 (29) | 16.8 (12.9–21.2) | 2090 (27) | 12.6 (8.7–17.2) | 0.75 |
The comparison of HPV type-specific prevalence cancer and HSIL cases is illustrated by
In the 52 sub-studies included in the HSIL systematic review, 16 were performed in Mexico or Central America and 36 in South America. Overall, a total of 2446 patients' samples were analyzed with a median of 47.5 specimens in each sub-study (range 6 to 130). Most data came from cross-sectional studies (n = 39) while seven came from case-control studies, four from cohort studies/prospective follow up, one from a nested case-control study, one from a before-after study and one from a randomized controlled trial. Mean age of women was 40.4±7.6 years old.
Any HPV in HSIL was found in a pooled proportion of 82.5% (95% CI 77.3–87.1%; I2 = 86.4%) of samples, while prevalence of HPV16 was 46.5% (95% CI 41.3–51.7%; I2 = 84.6%) and prevalence of HPV18 was 8.9% (95% CI 6.3–11.8%; I2 = 80.0%) (
Subgroups | HSIL | CERVICAL CANCER | ||||||
HPV TYPE: 16 | HPV TYPE: 18 | HPV TYPE: 16 | HPV TYPE: 18 | |||||
N patients(studies) | Prevalence(95% CI) | N patients(studies) | Prevalence(95% CI) | N patients(studies) | Prevalence(95% CI) | N patients(studies) | Prevalence(95% CI) | |
|
||||||||
|
||||||||
Argentina | 502 (12) | 48.5 (36.7–60.3) | 490 (11) | 16.9 (9.8–25.4) | 1013 (10) | 59.5 (51.3–67.5) | 1013 (10) | 17.6 (12–24.1) |
Barbados | - | - | - | - | 21 (1) | 71.4 (47.8–88.7) | - | - |
Belize | 15 (1) | 46.7 (21.3–73.4) | 15 (1) | 0 (0–0) | - | - | - | - |
Bolivia | - | - | - | - | 49 (1) | 34.7 (21.7–49.6) | 49 (1) | 4.1 (0.5–14) |
Brazil | 466 (13) | 52.7 (45.6–59.6) | 466 (13) | 9 (5–14.1) | 1269 (13) | 53.2 (42.9–63.3) | 1269 (13) | 15.8 (8.9–24.2) |
Chile | 95 (3) | 18.5 (5.8–36.3) | 73 (2) | 5.9 (0.2–26.2) | 420 (4) | 51.8 (29.7–73.5) | 420 (4) | 9.5 (4.2–16.7) |
Colombia | 241 (3) | 56.7 (31.2–80.4) | 209 (2) | 4.9 (1.7–29.5) | 450 (4) | 46.7 (35.9–57.7) | 450 (4) | 7.5 (3.7–12.6) |
Costa Rica | 130 (1) | 43.1 (34.4–52) | 130 (1) | 7.4 (2.8–15.4) | 35 (1) | 45.7 (28.8–63.4) | 35 (1) | 17.1 (6.6–33.6) |
Cuba | 45 (1) | 31.1 (18.2–46.6) | 45 (1) | 6.3 (0.8–20.8) | 45 (1) | 57.8 (42.2–72.3) | 45 (1) | 6.7 (1.4–18.3) |
Ecuador | 32 (1) | 81.3 (63.6–92.8) | 32 (1) | 4.5 (0.9–12.7) | 47 (1) | 80.9 (66.7–90.9) | 47 (1) | 4.3 (0.5–14.5) |
Honduras | 81 (1) | 35.8 (25.4–47.2) | 81 (1) | 6.9 (3.2–12.7) | 104 (1) | 43.3 (33.6–53.3) | 104 (1) | 10.6 (5.4–18.1) |
Jamaica | 66 (1) | 24.2 (14.5–36.4) | 66 (1) | 6.7 (1.4–18.3) | - | - | - | - |
Mexico | 405 (9) | 48.5 (35.5–61.6) | 405 (9) | 6 (3.1–9.7) | 1021 (14) | 54.9 (47.6–61.9) | 840 (13) | 12.8 (9.7–16.2) |
Nicaragua | 175 (2) | 28.8 (22.4–35.7) | 108 (1) | 6.7 (1.4–18.3) | 136 (2) | 38.1 (17–61.9) | 19 (1) | 5.3 (0.1–26) |
Panama | - | - | - | - | 255 (2) | 41.6 (31.3–52.2) | 73 (1) | 15.1 (7.8–25.4) |
Paraguay | 74 (1) | 41.9 (30.5–53.9) | 74 (1) | 1.4 (0–7.3) | 154 (2) | 61.3 (33.9–85.2) | 154 (2) | 7.2 (1.8–15.7) |
Peru | - | - | - | - | 198 (1) | 55.6 (48.3–62.6) | 198 (1) | 12.6 (8.3–18.1) |
Suriname | - | - | - | - | 246 (2) | 42.2 (29.4–55.7) | 246 (2) | 16.3 (12–21.2) |
|
||||||||
Central America and Mexico | 917 (16) | 41.7 (33.8–49.8) | 850 (15) | 6.3 (4.6–8.3) | 1617 (22) | 51.7 (45.6–57.8) | 1116 (18) | 12.5 (10.1–15.1) |
South America | 1410 (33) | 48.9 (42.2–55.5) | 1344 (30) | 10.5 (6.6–15.1) | 3846 (38) | 54.0 (48.6–59.2) | 3846 (38) | 13.3 (10.4–16.5) |
|
||||||||
Lower middle income | 714 (10) | 43.6 (32.8–54.8) | 615 (8) | 5.5 (2.3–10.1) | 1429 (15) | 49.4 (42.6–56.2) | 1312 (14) | 9.5 (7.2–12) |
Upper middle income | 1613 (39) | 47.3 (41.5–53.2) | 1579 (37) | 9.8 (6.8–13.2) | 4013 (44) | 54.1 (49.2–58.9) | 3650 (42) | 14.8 (11.9–18) |
High income | - | - | - | - | 21 (1) | 71.4 (47.8–88.7) | - | - |
We found a pooled prevalence of HPV18 in HSIL of 16.9% (95% CI 9.8–25.4%; I2 81.2%) in Argentina, 9.0% (95% CI 5.0–14.1%; I2 = 66.0%) in Brazil, and 6% (95% CI 3.1–9.7%; I2 = 50.6%) in Mexico. HPV prevalence according to subgroups of geographic region and by GNI World Bank Classification are shown in
Subgroups | HSIL | CERVICAL CANCER | ||||||
HPV TYPE: 16 | HPV TYPE: 18 | HPV TYPE: 16 | HPV TYPE: 18 | |||||
N patients(studies) | Prevalence(95% CI) | N patients(studies) | Prevalence(95% CI) | N patients(studies) | Prevalence(95% CI) | N patients(studies) | Prevalence(95% CI) | |
|
||||||||
Hybridization techniques |
494 (8) | 37.1 (31.6–42.7) | 427 (7) | 8.2 (1.6–19.3) | 998 (15) | 47.7 (39.1–56.4) | 816 (14) | 12.0 (9–15.4) |
PCR 1 |
948 (23) | 48.2 (39.7–56.7) | 882 (20) | 7.6 (6–9.4) | 1355 (18) | 58.5 (51.2–65.7) | 1174 (17) | 11.3 (7.5–15.7) |
PCR 2 |
560 (12) | 42.9 (33.5–52.7) | 560 (12) | 7.5 (4.2–11.6) | 2618 (19) | 49.9 (42.8–56.9) | 2480 (17) | 14.9 (10.2–20.3) |
PCR 3 |
292 (5) | 57.7 (39.7–74.6) | 294 (5) | 16.6 (4.7–33.7) | 420 (6) | 62.4 (51.9–72.4) | 420 (6) | 16.9 (11.7–22.9) |
|
||||||||
Exfoliated cells | 1330 (26) | 44.7 (38.4–51.1) | 1251 (24) | 6.5 (4.3–9.2) | 914 (16) | 58.4 (52.3–64.4) | 914 (16) | 12.2 (8.4–16.5) |
Fixed biopsies | 805 (13) | 43.4 (31.4–55.7) | 586 (12) | 13.2 (6.3–22.3) | 2352 (30) | 52.4 (46.2–58.6) | 2149 (28) | 14.6 (10.9–18.8) |
Fresh biopsies | 266 (7) | 50.5 (36.1–64.7) | 266 (7) | 9.1 (4.5–15.2) | 1592 (9) | 50.7 (42.3–59) | 1411 (8) | 8.8 (6.3–11.8) |
Combined | 32 (1) | 78.1 (60–90.7) | - | - | 605 (5) | 46.5 (25.6–68) | 488 (4) | 16.3 (10.2–23.3) |
*Southern blot, Dot blot, FISH and In Situ Hybridization.
**Polymerase Chain Reaction MY09/11 or Consensus primers.
Polymerase Chain Reaction SPF, GP5/6, E6, E7 and others.
Polymerase Chain Reaction MY and GP performed together.
In the 62 sub-studies included in the ICC systematic review, a total of 5540 patients' samples were analyzed with a median of 56 specimens in each study (range 14 to 750). Most data came from cross-sectional studies (n = 52) while 10 came from case-control studies, and one nested case-control study. Mean age of women was 41.1±7.0 years old.
Any HPV in cervical cancer was found in a pooled proportion of 89.0% (CI 84.3–92.9%; I2 = 94.0%) of the samples, while the prevalence of HPV16 was 53.2% (CI 49.1–57.2%; I2 = 88.5%) and the prevalence of HPV18 was 13.2% (CI 11.0–15.6%; I2 = 81.1%) (
When we analyzed the prevalence of HPV18 in ICC samples, we found a pooled prevalence of 17.6% (95% CI 12–24.1%; I2 = 65.7%) in Argentina, 15.8% (95% CI 8.9–24.2%; I2 = 92.8%) in Brazil, and 12.8% (95% CI 9.7–16.2%; I2 = 47.6%) in Mexico. HPV prevalence by the geographic region and by GNI World Bank Classification are shown in
We also applied a meta-regression analysis adjusting by GNI World Bank Classification, Geographic region, genotyping method and HPV tissue source to obtain adjusted estimates. There were no statistically significant differences for HPV16 in cancer and in HSIL. For HPV18, the statistically significant difference were seen for HSIL when the tissue source was fixed biopsies (compared to exfoliated cells) and when MY and GP performed together were used compared to Hybridization techniques and for cancer when the tissue source were Polymerase Chain Reaction SPF, GP5/6, E6, E7 and others compared to Hybridization techniques. However the adjusted prevalence, by the means of each variable and considering the SE of the meta-analysis, remained stable: HPV16 in HSIL women 45.7% (CI 95% 42.9–48.5%) HPV18 in HSIL 8.7% (7.2–10.3%); HPV16 in ICC 55.3% (52.5–58.1%); and HPV18 in ICC 13.4 (11.5–15.3).
Funnel plots showed no evidence of publication bias (data not shown).
Data on the geographic distribution of HPV type in HSIL and ICC are crucial for estimating the impact of HPV vaccines on cervical cancer and cervical screening programs.
This review brings representative estimations of HPV type distribution from the LA&C region. Since multiple HPV genotyping techniques have been included, varying sensitivities of the techniques considered might impact the HPV type-specific prevalence reported
In 2003, Smith et al.
Data on ICC has greatly enriched previous reports; we increased the number of Latin American cases included from 3,010 considered by the last published meta-analysis
Some intra-regional variations of the most common HPV types have been observed, although these apparent differences may happen simply by random fluctuation and/or a lack of sample representativeness of certain countries. For ICC, Mexico, Central America and the Caribbean showed a slightly lower HPV16/18 prevalence than South America (64.2% vs. 67.3% respectively). Particularly, Argentina shows the highest prevalence rate for HPV16/18 in both HSIL (65.4%) and ICC (77.1%). It is interesting to point out that the 12 Argentine studies incorporated samples from women of different provinces of the country, including aboriginal communities (Quechua
In 11.6% of HSIL and 7.5% of ICC, HPV detection resulted positive, but the viral type could not be identified (“other type”); these cases most likely represent the failed detection of known types (almost certainly different than HPV 16 and 18) rather than infections of yet-undiscovered types.
In this review, multiple-type HPV infections were detected in 16.8% of HSIL and 12.6% of ICC, although the frequency of multiple infections depends largely on the number of HPV types tested for within a given study. The attribution of ICC etiology to HPV types is increasingly complicated by the rising prevalence of multiple co-existing types. It was suggested that infections with multiple HPV types seem to act synergistically in cervical carcinogenesis
HPV18 and 16 had the highest ICC∶HSIL prevalence ratio in our studies, as found in Smith et al. meta-analysis
As more data is accumulated, it is supportive to observe that HPV16/18 accounts for two-thirds of ICC in LA&C. The proportion of ICC cases potentially averted by the current approved vaccines may be even higher than the aforementioned one if cross-protection against non-vaccine high-risk HPV types (like HPV31 and 45) is found to be clinically effective in reducing the incidence of ICC and HSIL caused by these genotypes. The information given by this work would be also useful in LA&C for the evaluation of polyvalent vaccines (currently in development) for the prevention of ICC associated to more than eight or nine high-risk HPV types.
Limitations of our meta-analysis include the cross-sectional design of the included studies and their inherent risk of bias, lack of representativeness, the HPV type-specific prevalence variation and HPV type-specific sensitivity of different PCR protocols
Although information on the histological type of ICC was collected, its discrimination was not always clear and the data came mostly from SCC. For this reason we presented only global data of ICC.
This study is the broadest summary of HPV type distribution in HSIL and ICC in LA&C to date, and it has included the majority of American countries which have the highest cervical cancer burdens in the region and worldwide. The presented information may be of importance for local decision makers to consider the cervical cancer prevention as a whole, taking into account the relevance of vaccination and updating screening strategies using type-specific high-risk HPV-DNA-based tests. This work comes available at the time some Latin American and Caribbean countries are evaluating the HPV vaccine introduction in their National Vaccination Schedules, in the frame of the Pan American Health Organization purchase using revolving fund, which makes vaccines affordable. Continued surveillance of HPV types in HSIL and ICC as HPV vaccines are introduced would be useful, to assess the potential for changing type-specific HPV prevalence in the post-vaccination era in Latin America.
(DOC)
(DOC)
(DOC)
(DOC)
(DOC)
We deeply thank Dr. Eduardo Franco, Dr. Nubia Muñoz, Dr. Jorge A. Gómez, Dr. Mark Schiffman and other collaborators for their advice and support. We are also indebted to Dr. Virginia Alonio for her critical review of the manuscript, Jonathan Willner for language editing and librarian Daniel Comandé for his efforts.