Cervical cancer is the second most common cancer in women worldwide but the first in most developing countries, where 80% of cases occur.1 Molecular epidemiologic evidence accumulated over the past 20 years clearly indicates that certain HPV types are the central and probably necessary cause of cervical cancer.2, 3, 4 Over 100 HPV types have been identified, and about 40 are known to infect the genital tract.5
To assess the geographic distribution of HPV types in invasive cervical cancer and the risk linked to the various HPV types, the IARC carried out an international survey of cervical cancer specimens in 22 countries4, 6 and a multicenter case-control study in 11 countries around the world.7, 8, 9, 10, 11, 12, 13, 14, 15
These studies were carried out using the same protocol and sensitive PCR assays for the detection of HPV DNA in a central laboratory.
HPV prophylactic vaccines for the most common types (HPVs 16 and 18) are now being tested in various populations with promising results as they have been shown to be safe and to induce high titers of neutralizing antibodies that prevent HPV infection and cervical intraepithelial neoplasia (CIN).16, 17, 18 Although there is some evidence of cross-reactivity among certain HPV types,19 this appears to account for <1% of antibody reactivity, indicating that protection to infection is type-specific. Effective HPV vaccines should then be polyvalent and should contain the HPV types responsible for most cervical cancers. It is thus of great importance to assess if there are significant differences in the geographic distribution of HPV types causing cervical cancer. If this is the case, their description would provide the rationale for considering tailoring vaccine products to the HPV types prevalent in a given geographic area vs. producing a common vaccine suitable to the worldwide distribution of HPV types. Previous studies have revealed some geographic differences in HPV type distribution, but the size of individual studies and the validity of the PCR testing systems used precluded delineation of clear patterns.4, 6–15, 20
HPV screening using a cocktail of 13 high-risk HPV types, the hybrid capture 2 (HC2) system, has been recognized by the U.S. Food and Drug Administration (FDA) regulatory office as useful for triage of the atypical squamous cells of undetermined significance (ASCUS) cytology and for primary screening in women over 30 years of age adjunct to cytology; it is being introduced into clinical management protocols (http://www.fda.gov/bbs/topics/NEWS/2003/NEW00890.html). It is thus relevant to evaluate the screening ability of testing cocktails as predicted by the HPV types included and the opportunities for region-specific testing systems.
We report here the pooled results of all cervical cancer cases included in the IARC studies to assess the geographic distribution of HPV types in invasive cervical cancer and its implications for cervical cancer prevention and screening regarding the composition of prophylactic vaccines and screening cocktails.
MATERIAL AND METHODS
International Biological Study on Cervical Cancer (IBSCC).
From 1989 to 1992, a total of 1,050 women with incident, histologically confirmed cervical cancer were recruited in 32 hospitals from 22 countries. More details are given in the original publication based on 932 women included in the first analysis6 and in the subsequent histology and HPV reassessment performed in 55 of the 66 cases that tested initially HPV-negative.4 For the present analysis, 24 women were excluded because their specimens could not be reassessed or because they lacked basic information. A total of 206 women were included both in the IBSCC and in the multicenter case-control studies and were counted once. In case of divergent results, those from the multicenter case-control study were used. The IBSCC contributed thus a total of 702 cases for this pooled analysis.
Multicenter case-control study.
From 1985 to 2000, 11 case-control studies were conducted in 10 countries in Africa, the Americas, Europe and Asia (Table I). The studies were population-based in Colombia and Spain and hospital-based in the other countries. Briefly, cases were 2,905 women with incident, histologically confirmed cervical cancer who had not received previous treatment. Detailed methods of case and control selection are described in the original papers.7, 8, 9, 10, 11, 12, 13, 14, 15
Table I. Overall HPV Prevalence by Study Region and Histology
Women were interviewed at the hospitals by trained interviewers using a standardized questionnaire to elicit information on sexual behavior, reproductive history, contraceptive practice, smoking habits, screening history and various measures of socioeconomic status. The questionnaire used in the case-control studies was more detailed than the one used in the IBSCC. After interview, all women had a pelvic examination performed by a gynecologist or a nurse. In the IBSCC, biopsy specimens were obtained from the tumor and kept frozen at −20°C to −70°C.4 In the case-control studies, 2 cervical scrapes were taken in most women. They were used to prepare a Pap smear and the remaining cells stored in PBS and kept at −70°C. A tumor biopsy kept frozen at −70°C was also taken in most cases.2
The slides from which diagnoses were made were reviewed by expert pathologists as described in the original papers.
HPV DNA detection and typing
PCR-based methods were used for the detection and typing of HPV DNA. Primers for a fragment of the β-globin gene were used to assess the quality of the target DNA.
In the IBSCC, HPV DNA was detected in 2 stages. In the first stage, the L1 primers MY09 and MY11 and 26 specific probes were used [HPVs 6, 11, 16, 18, 26, 31, 33, 35, 39, 40, 42, 45, 51, 52, 53, 54, 55, 56, 58, 59, 66, 68, 73 (PAP 238a, MM9), 82 (W13b), 83 (PAP 291) and 84 (PAP 155)].6 In the second stage, paraffin-embedded tissue from 66 biopsies found to be HPV DNA-negative in the first stage were retested using type-specific E7 primers for 14 high-risk HPVs (HPVs 16, 18, 31, 33, 35, 39, 45, 51, 52, 56, 58, 59, 66 and 68) and consensus GP5+/6+ and CP I/II primers.4 Paraffin-embedded tissue was used for these 66 cases because the remaining tissue of the original frozen biopsies had been fixed in formalin.
For the case-control studies, HPV DNA in cervical scrapings (exfoliated cells) and biopsy specimens was detected blindly in one central laboratory using PCR-based assays. Detailed protocols for these assays have been described as follows. The relatively insensitive version of PCR primers for the L1 gene, MY09/MY11, was used in the Colombian and Spanish studies.21 These data were excluded when comparability across geographic regions or age-adjusted estimates are presented. The GP5+/6+ general primer system was used in the remaining studies.22 To detect specific HPV types, oligohybridization methods were used for types 16, 18, 31, 33, 35, 39, 45, 51, 52, 54 and 56. A second round was performed for types 6, 11, 26, 34, 40, 42, 43, 44, 53, 58, 59 and 68. A third round was performed for types 57, 61, 66, 70, 72, 73, 81 (CP8304), 82 (W13B/MM4 and IS39 subtypes), 83 and CP6108. Specimens that were positive on assays with GP5+/6+ but did not hybridize with any of the 33 probes were coded as HPV type X.
For specimens from patients negative for both β-globin and HPV DNA or positive for an unknown type (HPV X), DNA was extracted from the cell pellets and retested as described above. For detection of HPV in biopsy specimens, the sandwich method, which provides histologic evidence of cancer in the previous and subsequent tissue sections adjacent to that used for HPV testing, was used; and an additional HPV type-specific PCR was performed, using a set of E7 primers, for 14 HPV types (16, 18, 31, 33, 35, 39, 45, 51, 52, 56, 58, 59, 66 and 68) that were formerly designated as oncogenic.4
As shown in Table I, countries included in the study were grouped into 5 regions: northern Africa, sub-Saharan Africa, Central/South America, south Asia and Europe/North America, the latter combined because of their similar incidence rates of cervical cancer and the low number of cases included from North America.
The 95% confidence intervals (CIs) for proportions were estimated by standard methods.
A generalized linear Poisson model was fitted to the data on viral type and geographic region to detect statistically significant departures from the independent model of no association. Standardized residuals were used to identify the region–HPV type combinations that were responsible for the heterogeneity (those in which residuals were >1.96). For each genotype–region combination detected in the Poisson model, a 2 × 2 table was created and Pearson's χ2 test or, when appropriate, Fisher's exact test was used to assess differences in HPV type-specific prevalence in one region compared to the other regions combined. To correct for multiple statistical comparisons, statistical significance for these tests was set at the 2-sided 0.005 level.
To test for trends in the prevalence of a specific HPV type with age, the Mantel-Haenszel linear trend test was used, with statistical significance set at the 2-sided 0.05 level.
HPV types that were detected in 10 or fewer women were combined and labeled as “HPV other” unless otherwise specified.
To assess the theoretical impact of different HPV vaccine compositions on cervical cancer prevention, we estimated the number of cervical cancer cases attributed to each genotype in 5 world regions. We used estimates of the number of incident cervical cancer cases in women 15 years of age and older published in Globocan1 and the region- and type-specific HPV distribution derived from this study. A woman with multiple infections was assigned in proportional fractions to each genotype but counted only once.
The impact of theoretical cocktail compositions on HPV DNA detection using the GP5+/6+ PCR system was assessed by estimating the number of HPV-positive cases and controls recruited in the IARC multicenter case-control study that would be detected by including different numbers of HPV types in the screening cocktail. These compositions included (i) the 7 most frequent HPV types detected in the present study; (ii) the 13 types included in the current HC2 system (16, 18, 45, 31, 52, 33, 58, 35, 59, 51, 56, 39 and 68); (iii) a 15-type cocktail that included the previous 13 types plus HPVs 73 and 82, recently classified as high-risk types in the IARC multicenter case-control study;2 and (iv) an 18-type cocktail that included the previous 15 types plus HPVs 26, 53 and 66 classified as probable high-risk types by the IARC study.2
All protocols were cleared by the IARC and local ethics committees, and informed, written consent was obtained from all participants.
A total of 3,607 women with cervical cancer were included (702 from the IBSCC and 2,905 from the multicenter case-control study). Of these, 3,216 (89.2%) provided specimens for HPV DNA testing; 3,012 of these had squamous cell carcinoma and 204 had adenocarcinoma or adenosquamous carcinoma. Of the 3,216 specimens tested for HPV DNA, 131 (4.1%) were negative for both β-globin and HPV DNA and excluded from the analysis, leaving 3,085 specimens adequately tested for HPV DNA; all analyses are based on these cases (Table I).
Overall HPV DNA prevalence
The overall HPV DNA prevalence was 92.5% (95% CI 91.6–93.4), but if data from Spain and Colombia (whose specimens were tested with a less sensitive PCR assay) are excluded, this figure rises to 96.2% (95% CI 95.4–97.0). In squamous cell carcinoma, it ranged from 83.6% in Europe/North America to 96.5% in south Asia. The lower prevalence in Europe/North America and Central/South America is essentially explained by the prevalence observed in Spain (80.1%) and Colombia (74.4%). In adenocarcinoma, HPV prevalence ranged from 93.3% in south Asia to 100% in sub-Saharan Africa and Europe/North America, but the numbers of cases in these 2 regions were small (Table I).
HPV type-specific prevalence
In 2,633 (92.2%) of the 2,855 specimens positive for HPV DNA, a single HPV type was detected, and in 222 (7.8%), 2 or more types were identified. Of those with multiple infections, 71% had HPV-16 or HPV-18 (Table II). We estimated the HPV distribution either as a single or as part of a multiple infection of each HPV type over the total number of HPV-positive cases in the study. Thirty different types were detected, and the most common genotypes, in descending order of frequency, were HPVs 16 (57.4%), 18 (16.6%), 45 (6.8%), 31 (4.3%), 33 (3.7%), 52 (2.5%), 58 (2.3%), 35 (2.2%), 59 (1.5%), 56 (1.3%), 39 (1.2%), 51 (1.1%), 73 (0.6%), 68 (0.5%), 66 (0.5%), other HPV types (1.9%) and untyped HPV (4.4%).
Table II. Distribution of Single and Multiple HPV Infections by Histology
Single HPV infections
HPV HR (high risk), material not available for further typing.
Table II shows the distribution of HPV types by histologic type of cervical cancer. HPV-16 was detected more often in squamous cell carcinoma (54.4%) than in adenocarcinoma (41.6%) (p < 0.001). The same was observed for the phylogenetically related HPV-52 (2.3% vs. 0%, respectively; p = 0.04). Conversely, HPV-18 was more common in adenocarcinoma (37.3%) than in squamous cell carcinoma (11.3%) (p < 0.001).
HPV type-specific prevalence by study region is shown in Table III, after excluding cases from Spain and Colombia. HPV-16 was the predominant type in all regions, ranging from 47.7% in sub-Saharan Africa to 69.7% in Europe/North America. HPV-18 was the second most common type worldwide, with prevalence ranging from 12.6% in Central/South America to 25.7% in south Asia. HPV-45 was the third most common type in Africa, south Asia and Europe/North America, while HPV-31 was the third most common type in Central/South America. In northern Africa, HPVs 33 and 31 were, respectively, the fourth and fifth most common types, while in sub-Saharan Africa, HPVs 33 and 58 ranked fourth and HPV-56 fifth. The fourth and fifth most common types were, respectively, 45 and 33 in Central/South America, 52 and 58 in south Asia and 31 and 56/68 in Europe/North America.
Table III. Prevalence1 of HPV Types2 in all Cases of Cervical Cancer by Study Region
Excludes studies in Spain and Colombia (see Material and Methods for details). Percent relative to HPV+ women.
As single or multiple HPV infections.
Significantly increased (p ≤ 0.005).
Significantly decreased (p ≤ 0.005).
“Others” includes HPV types 40, 42, 53, 54, 55, 83 and 84.
The HPV-16 prevalence was statistically higher than expected in northern Africa and Europe/North America and statistically lower than expected in south Asia. For HPV-18, a decreased prevalence was observed in Central/South America and an increased prevalence was found in south Asia. HPV-45 detection was higher than expected in sub-Saharan Africa. HPV-31 prevalence was higher in Central/South America and lower in south Asia. Other statistically significant differences were found for HPV types 39 and X, as shown in Table III.
To explore the possibility of time trends in the HPV type-specific prevalence in invasive cancer, cases were stratified into 3 age groups (Table IV). The prevalence of HPVs 33, 52, 58 and 39 increased significantly with age, while that of HPV-45 decreased significantly with age. For HPV-18, the prevalence also tended to decrease with age, but this decrease was statistically significant only for adenocarcinoma.
Table IV. Prevalence1 of HPV Types2 in all Cases of Cervical Cancer by Age
Age ≥ 50
Not including 2 studies of carcinoma in situ. Percent relative to HPV+ women.
As single or multiple HPV infections.
“Others” includes HPV types 40, 42, 53, 54, 55, 83 and 84.
We estimated the number of cases that can be attributed to each HPV type in 5 world regions (Fig. 1) and overall (Fig. 2), after taking into account the estimated region-specific HPV genotype distribution and the number of incident cervical cancer cases (see Material and Methods). As shown in Figure 1, HPVs 16 and 18 are the 2 types accounting for most of the cases in each world region. The proportion of cases attributed to these 2 types are 63.9% in sub-Saharan Africa, 78.9% in northern Africa, 65.0% in Central/South America, 73.5% in south Asia, 71.5% in Europe/North America and 70.7% in all regions combined (Fig. 2).
As shown in Figure 2, there are 7 genotypes that each account for more than 10,000 cases of cervical cancer: 16, 18, 45, 31, 33, 52 and 58. These 7 types account for 87.4% of all cervical cancer cases worldwide.
Table V shows the predicted number and percentage of cases of invasive cervical cancer and controls from our study that would have been identified as HPV-positive by different cocktail compositions (i.e., those including 13, 15, 18 or 7 genotypes) using the GP5+/6+ PCR system. No substantial gains in the number of HPV-positive cases and controls detected were observed from adding 2 or 5 types to the current HC2 13-type cocktail. However, reducing the number of types from 13 to the most frequent 7, a substantial loss in sensitivity would be observed among cases (74 would be classified as HPV-negative, a 5.5% reduction) and controls (25 would be classified as HPV-negative, a 2.0% reduction).
Table V. HPV DNA Detection Using the GP5+/6+ PCR System with Different HPV Cocktail Compositions in Invasive Cervical Cancer Cases and Controls Recruited in the IARC Multicenter Case-Control Study
Our pooled analysis based on 3,085 cases of cervical cancer from 25 countries around the world provides the most homogeneous set of cases to determine geographic variation and prevalence of HPV types. All studies used similar protocols and, more importantly, HPV DNA detection was done in central laboratories using the 2 PCR-based assays that were state of the art at the time. A total of 30 different HPV types were detected, but 2 types predominated: 57.4% of cases were associated with HPV-16 and 16.6%, with HPV-18. These 2 types, as either single or multiple infections, were detected in 74% of all cases. The next 5 most prevalent types, accounting for 19.6% of cases, were, in descending order of frequency, 45, 31, 33, 52 and 58. HPV-16 and the phylogenetically related HPV-52 were more common in squamous cell carcinoma than in adenocarcinoma, whereas HPV-18 was more common in adenocarcinoma.
Concerning geographic variation, HPVs 16 and 18 were the first and second most common types in all regions, respectively, but showed significant differences in distribution. The highest prevalence of HPV-16 was observed in Europe/North America and in northern Africa and the lowest, in sub-Saharan Africa and south Asia. HPV-18 was 2 times more common in south Asia than in Central/South America. The third most common type was HPV-45 in Africa, south Asia and Europe/North America but HPV-31 in Central/South America.
The 10 most common HPV types identified in our pooled analysis were the same types reported in a recent meta-analysis.20 However, some differences were observed. Our estimated overall HPV prevalence in the various regions using the GP5+/6+ system was >95%, while estimates in the meta-analysis ranged 83–89%. These differences reflect the heterogeneity in the type of specimens and the lower sensitivity of the PCR systems used in many of the studies included in the meta-analysis.
Although we used the 2 most sensitive PCR assays at the time that our studies were conducted, some limitations of our results should be considered. The relative sensitivities for overall HPV DNA detection of the 2 systems are highly comparable. However, the typing system used in these studies for the GP5+/6+ system was less sensitive than that of the MY09/11 system.22 Overall, the MY09/11 system has been reported to be more robust in detecting multiple infections than GP5+/6+. GP5+/6+ detects only 47% of samples with multiple types compared to 90% detected by MY09/11. Thus, there are some differences in the identification of certain HPV types. The GP5+/6+ PCR primers are relatively inefficient in the amplification of HPV types 53 and 61 compared to MY09/11 PCR, whereas MY09/11 PCR is inefficient in the amplification of HPV-35.23
Around 75% of our samples were tested using GP5+/6+, which may have resulted in a small percentage of multiple infections being categorized as single infections. Future studies using more sensitive techniques for the detection of multiple types, such as blot line assays, will refine these estimates.
The number of adenocarcinomas studied was rather small (close to 200), which precluded analysis of regional variation of HPV types. The number of cases studied from Africa and Europe/North America was about half of the number studied from Central/South America and south Asia, and large regions such as China were not included. Thus, the prevalence of HPV-58, e.g., which has been reported to be high in China,24 Korea25 and some regions of Africa,26 may be underestimated in our study.
Our results have important implications for the prevention of cervical cancer in relation to the composition of prophylactic HPV vaccines and screening cocktails for high-risk HPV types.
An HPV-16/-18 vaccine, like the ones currently being tested,17, 18 would theoretically prevent 71% of cervical cancers worldwide (Fig. 2); but its impact with regard to the percentage of cases theoretically prevented would be higher in Asia (73.5%) and Europe/North America (71.5%) than in Africa (67.7) or Central/South America (65.0%).
A vaccine that included the 7 most common HPV types shown in Figure 2 (16, 18, 45, 31, 33, 52 and 58) would prevent 87.4% of cervical cancers worldwide: 90.6% in northern Africa, 88.8% in south Asia, 86.5% in sub-Saharan Africa, 85.8% in Europe/North America and 84.9% in Central/South America (data not shown). These estimates assume that the prevention of HPV-16 and HPV-18 infections is unlikely to increase the risk of persistent infection by other HPV types, as has been suggested by Liaw et al.27
Another approach would be to consider that the vaccine is effective for cervical cancer cases linked to multiple types when all the types are included in the vaccine. Under this assumption, the global percent prevention would decrease from 71% to 68.3% for an HPV-16/-18 vaccine and from 87.4% to 85.8% if the 7 most common HPV types were included.
Concerning HPV cocktails for screening and management, our results indicate that the HC2 cocktail, which currently includes the 13 HPV types listed in Table V, would have detected as HPV-positive 91.6% of the cancers. Adding HPVs 73 and 82, classified as high-risk types in the IARC study,2 would have resulted in 8 additional cases detected, for a gain in sensitivity of 0.6%. Adding 3 HPV types, classified as probable high-risk types2 (i.e., HPVs 26, 53 and 66), would have resulted in 8 additional cases detected, for a gain in sensitivity over the 13-type cocktail of 1.2%.
Conversely, a system with either 15 or 18 HPV types would have detected one additional HPV-positive control, for a gain in sensitivity of 0.1%. These fractions would have minute variations if the HPV types added reflected the small region-specific variation observed.
However, reducing the screening cocktails to the 7 most common types in cervical cancer (HPVs 16, 18, 45, 31, 33, 52 and 58) would result in a loss of sensitivity of 5.5% among cases and of 2.0% among controls (i.e., 74 additional cases and 25 additional controls classified as HPV-negative).
In general terms, the HPV type distribution in cases is skewed enough so that modifications of the current HC2 system would have consistently a minor impact in the performance of the screening test. These modifications are largely irrelevant if compared with the impact of achieving widespread screening coverage of the population or ensuring proper follow-up of HPV-positive women. In the presence of HPV mass immunization campaigns, screening may have to continue until more definite information is available on duration of protection, proportion of nonrespondents and protection of women who are only partially vaccinated. In these populations, screening protocols would likely focus on HPV testing as the primary screening method and use cytology as the method of triage for HPV-positive women.
In conclusion, a vaccine containing the 7 most common HPV types would prevent about 87% of cervical cancers worldwide, with little regional variation. Generating a vaccine with 7 HPV types would be technically feasible; however, the production cost may be high, and the protection conferred by the less common HPV types may be difficult to demonstrate clearly.
The impact of modifying the number of HPV types in the screening cocktail tests would be small and probably irrelevant for screening programs.
We thank Dr. P.J.F. Snijders (Vrije Universiteit Medical Center, Amsterdam, the Netherlands), who performed the HPV DNA testing in the original studies. We acknowledge Ms. C. Rajo, who was responsible for the secretarial workload. We thank the following researchers and local principal investigators: A. Daudt (Hospital of the Clinics of Porto Alegre, Brazil), A. Schneider (Friedrich Schiller University, Jena, Germany), A. Vila Tapia (Regional Clinical Hospital, Ministry of Health, Concepción, Chile), A.M. Jansen (School of Public Health, Johns Hopkins University, Baltimore, MD), A.R. Teyssie (National Institute of Microbiology, Buenos Aires, Argentina), B. El Gueddari (Institut National d'Oncologie, Rabat, Morocco), C. Greer (Chiron Corporation, Emeryville, CA), C. Navarro (Health Council, Murcia, Spain), C. Ngelangel (University of the Philippines, Manila, the Philippines), C. Santos (Maes Heller Cancer Research Institute, Lima, Peru), C. Wheeler (University of New Mexico, Albuquerque, NM), E. Alihonou (Université Nationale du Bénin, Cotonou, Bénin), E. de los Rios (National Oncology Institute, Panamá City, Panamá), H. Cherif Mokhtar (Centre Hospitalo-Universitaire, Setif, Algeria), H.R. Wabinga (Makerere Medical School, Kampala, Uganda), I. Izarzugaza (Euskadi Cancer Registry, Vitoria Gasteiz, Spain), J. Eluf Neto (São Paulo University, São Paulo, Brazil), J. Kaldor (National Centre in HIV Epidemiology and Clinical Research, Darlinghurst, Australia), J. Peto (Institute of Cancer Research, Belmont, UK), J.L. Rios-Dalenz (Cancer Registry of La Paz, La Paz, Bolivia), J.N. Kitinya (Muhimbili Medical Center, Dar es Salaam, Tanzania), L.A. Tafur (University of Valle, Cali, Colombia), L.C. González (Delegation of Social Welfare, Salamanca, Spain), L.M. Puig Tintoré (Hospital Clínic i Provincial, Barcelona, Spain), M. Gili (Cátedra de Medicina Preventiva y Social, Seville, Spain), M. Koulibaly (Centre National d'Anatomie Pathologique, Conakry, Guinea), M. Manos (School of Public Health, Johns Hopkins University, Baltimore, MD), M. Santamaria (Navarra Hospital, Pamplona, Spain), M. Sherman (Johns Hopkins Medical Institutions, Baltimore, MD), M. Torroella (National Cancer Institute, Havana, Cuba), M.H. Schiffman (National Cancer Institute, Bethesda, MD), N. Aristizabal (Cali, Colombia), N. Ascunce (Breast Cancer Prevention Centre, Pamplona, Spain), P. Alonso de Ruiz (General Hospital of Mexico, Mexico City, Mexico), P. Ghadirian (Hôtel-Dieu de Montréal, Montréal, Canada), P. Viladiu (Cancer Registry of Catalunya, Barcelona, Spain), P.A. Rolón (Laboratorio de Anatomía Patológica y Citología, Asunción, Paraguay), R. Kurman (Johns Hopkins Medical Institutions, Baltimore, MD), S. Bayo (Institut Nacional de Recherche en Santé Publique, Bamako, Mali), S. Chichareon (Prince of Songkla University, Hat-Yai, Thailand), Sarjadi (Diponegoro University Medical Faculty, Semarang, Indonesia), V. Moreno (Institut Català d'Oncologia, Barcelona, Spain) and W. Zatonski (M. Sklodowska-Curie Memorial Cancer Centre, Warsaw, Poland). N.M. received a grant from the Spanish Ministerio de Educación, Cultura y Deporte for short-term visiting scientists (SAB2000-0261).