HUMAN PAPILLOMAVIRUS INFECTION AND INVASIVE CERVICAL CANCER IN PARAGUAY

Transcription

1 Int. J. Cancer: 85, (2000) 2000 Wiley-Liss, Inc. Publication of the International Union Against Cancer HUMAN PAPILLOMAVIRUS INFECTION AND INVASIVE CERVICAL CANCER IN PARAGUAY Pedro A. ROLÓN 1, Jennifer S. SMITH 2, Nubia MUÑOZ 2 *, Stefanie J. KLUG 3, Rolando HERRERO 2, Xavier BOSCH 4, Fernando LLAMOSAS 1, Chris J.L.M. MEIJER 5 and Jan M.M. WALBOOMERS 5 1 Laboratorio de Anatomia Patologica y Citologia, Asunción, Paraguay 2 Unit of Field and Intervention Studies, International Agency for Research on Cancer, Lyon, France 3 Department of Medical Information, Biometry and Epidemiology, Ludwig-Maximilians-Universität München, Munich, Germany 4 Institut Català d Oncologia, Barcelona, Spain 5 Free University Hospital, Amsterdam, The Netherlands HPV types 16 and 18 have been categorized as human carcinogens based on their strong associations with cervical cancer in previous case-control studies. Recent IARC studies in the Philippines, Thailand and Morocco show strong associations between invasive cervical cancer and less common HPV types, including HPV 31, 33, 45, 51, 52 and 58. We present results of a further IARC case-control study conducted in Asunción, Paraguay, to examine the association between specific HPV types and invasive cervical cancer as well as risk factors other than HPV. One-hundred thirteen incident histologically confirmed invasive cervical cancer cases and 91 age-matched hospital controls were recruited. A standardized questionnaire was administered to investigate known and suspected risk factors for cervical cancer. For HPV status determination, cervical biopsy specimens from case subjects and exfoliated cervical cells from control subjects were obtained. HPV DNA was ascertained using a GP5 /6 PCR-based assay capable of detecting more than 33 HPV types. Overall HPV prevalence was 97% in the cervical cancer cases and 20% in the control subjects. As a single infection, HPV 16 was the predominant type with a prevalence of 48% among case subjects and 5.5% among control subjects. Significant associations with the risk of cervical cancer were detected as follows: any HPV type (OR 114; 95% CI: ); HPV 16 (OR 910); HPV 18 (infinite OR); HPV 31 (OR 110); HPV 33 (OR 261); HPV 45 (OR 129); and HPV 58 (OR 36). In the multivariate model, risk factors other than HPV significantly associated with cervical cancer risk were a higher number of lifetime sexual partners, lower educational status and never having had a Pap smear. Strong associations were found between invasive cervical cancer and specific HPV types 16, 18, 31, 33, 45 and 58. Int. J. Cancer 85: , Wiley-Liss, Inc. Cervical cancer is the second most common cancer among women in the world and in developing countries. In the early 1990s, the worldwide age-standardized incidence rates of cervical cancer ranged from 2.6/100,000 women in Qidong, China, to 67.2/100,000 women in Harare, Zimbabwe (Parkin et al., 1997). Paraguay, a South American country with an estimated population of 4.8 million, had a notably high cervical cancer incidence rate of 41.1/100,000 in This is the second highest cervical cancer incidence rate in South America after Bolivia yet higher than in Brazil, Ecuador and Colombia. (IARC, 1998) Certain human papillomavirus (HPV) types have been demonstrated to be the primary etiological agents for cervical cancer (IARC, 1995). HPV types 16 and 18 have been categorized as human carcinogens based on their strong associations with cervical cancer (Mun oz et al., 1992; IARC, 1995). IARC-conducted invasive cervical cancer case-control studies in the Philippines (Ngelangel et al., 1998), Thailand (Chichareon et al., 1998), Brazil (Eluf-Neto et al., 1994) and Morocco (Chaouki et al., 1998) confirmed the strong associations between invasive cervical cancer and HPV types 16 and 18 and showed similar strong associations with less common HPV types, including HPV 31, 33, 45, 51, 52 and 58. Further data on the association between these less common HPV types and invasive cervical cancer risk are currently needed. We present results of an IARC study of invasive cervical cancer conducted in Asunción, Paraguay. Our report examines the association between invasive cervical cancer and specific HPV types as well as other risk factors of invasive cervical cancer. MATERIAL AND METHODS Subjects Between December 1988 and May 1990, 116 patients with invasive cervical cancer and 101 controls were recruited from the Cancer Hospital and the Clinical Hospital in Asunción, Paraguay. Women between 18 and 85 years of age were invited to participate. Inclusion criteria for case subjects included having a newly diagnosed, histologically confirmed invasive cervical cancer, having no previous cervical cancer treatment, and being mentally and physically capable of providing reliable responses. An expert pathologist confirmed all histological diagnoses. Women without cervical cancer and having had no prior history of treatment with conization or hysterectomy were selected as controls from outpatient clinics at the same hospitals where case selection occurred. Controls were selected using broad stratified age matching to the cases within 10-year age groups. Exclusion criteria for the controls included diseases associated with known risk factors for cervical cancer, including diagnoses of ano-genital tract cancers (vulva, vagina, rectum), tobacco-related diseases (e.g., lung cancer, coronary heart disease) and cancer of the breast, endometrium, ovary or colon. Data and specimen collection All consenting participants were administered a standardized questionnaire by trained interviewers on socio-demographic characteristics, sexual behavior, reproductive and contraceptive history, smoking and Pap smear screening history. A pelvic examination and Pap smear were performed on consenting participants. For HPV DNA detection, cervical biopsy specimens from case subjects and cervical exfoliated cells from control subjects were obtained. Cervical biopsy specimens were kept frozen at 70 C without additives. The stage of disease was coded according to the International Federation of Gynecology and Obstetrics. Cervical exfoliated cells were collected from control subjects by sampling the ectocervix with 2 wooden spatulas and the endocervix with 2 cytobrushes. Following Pap smear preparation, remaining cervical specimens were placed in tubes with PBS, centrifuged and stored Grant sponsor: European Community; Grant number: CI F[CD]; Grant sponsor: International Agency for Research on Cancer, Lyon, France; Grant number: FI/92/3-2 PAR; Grant sponsor: Preventiefonds, The Netherlands; Grant number: *Correspondence to: Chief, Unit of Field and Intervention Studies, International Agency for Research on Cancer, 150, Cours Albert Thomas, F Lyon, France. Fax: munoz@iarc.fr Received 14 June 1999; Revised 14 September 1999

2 HPV AND CERVICAL CANCER IN PARAGUAY 487 at 70 C until shipping. Protocols were cleared by the IARC Ethical Review Committees and by the Ethical Committee of the local hospitals in Asunción, Paraguay. HPV detection techniques HPV DNA detection was performed using the HPV GP5 /6 primer mediated PCR assay, targeting a small fragment of the L1 gene, in combination with HPV type-specific oligoprobes (Jacobs et al., 1995). To analyze the quality of target DNA for PCR purposes, cervical specimens were tested by PCR using -globin gene-specific primers. Cervical specimens were subjected to HPV DNA genotyping, which was carried out as follows: Briefly, a first screening was performed to determine the overall presence of HPV using a general primer GP5 /6 mediated PCR, which permits the detection of a broad spectrum of sequenced and still-unsequenced genital HPV types at the subpicogram level (Jacobs et al., 1995). HPV positivity was assessed by low-stringency Southern blot analysis of the PCR product with a cocktail probe of HPVspecific DNA fragments (Walboomers et al., 1992). Subsequently, GP5 /6 PCR was repeated on positive samples in triplicate to generate sufficient products for further typing. PCR products from independent reactions were pooled to generate identical Southern blots for parallel and successive hybridization rounds for typing using internal type-specific oligoprobes (Jacobs et al., 1995). The strategy of melting/rehybridizing the filters was essentially the same as described earlier (de Roda Husman et al., 1994). The first hybridization round was performed for HPV types 16, 18, 31, 33, 35, 39, 45, 51, 52, 54 and 56, respectively. The second round included HPV types 58, 59, 68, 6, 11, 26, 34, 40, 42, 43 and 44. The third round was performed for HPV 57, 61, 66, 70 (equivalent to CP-141-7), 72 (equivalent to CP 4137), 73 (equivalent to MM9), IS39 and MM4 (related to HPV 51), MM7, CP 6108 and CP The latter 3 types are phylogenetically related to the low-risk types HPV 61 and 62 (Myers et al., 1996). Type-specific oligonucleotides (30 mers) were selected on the basis of sequence information from the HPV sequence database (Myers et al., 1996) after alignment analysis with the Clustal program (PC/Gene, Release 6.7, Intelligenetics, Mountain View, CA). Samples that were GP5 /6 -positive but could not be identified by the above-mentioned probes were considered as HPV X, uncharacterized HPV types. MY09 and MY11 primers (Manos et al., 1989) were also used for 42 cases that were originally -globin- or HPV-negative for GP5 /6 primers. To further investigate the presence of HPV DNA among cases, we recut biopsy specimens using an HPV type-specific PCR that amplifies a small fragment of the E7 open-reading frame of 14 high-risk HPV types (for 11 cases that were either negative or positive for low-risk type HPVs or HPV X using the L1 primers) (Walboomers et al., 1999). For case subjects, the sandwich method (i.e., snap-frozen tissues were cut in a way so that the outer sections were used for histologic analysis and the inner sections were used for PCR testing) was used in which the quality of the tissue was controlled as described earlier (Walboomers et al., 1999). We took special precautions to minimize false-positive results in the PCR, as has been described in detail elsewhere (Walboomers et al., 1992). Statistical analyses Standard methods were used to estimate the associations between invasive cervical cancer and specific HPV types and other risk factors. For exploratory analyses, cross-tabulations and Pearson s correlation coefficients between variables were calculated. Statistical significance for differences in dichotomous variables was tested using a chi-squared test. A 2-sided linear trend was used for ordered categorical variables. Two categorical variables (age at first intercourse and age at first pregnancy) were highly correlated (correlation coefficient 0.79); age at first intercourse was more strongly associated with cervical cancer and included in the final model. Odds ratios (ORs) and 95% confidence intervals (CIs) were calculated as approximations of relative risks by unconditional, multiple logistic regression (STATA; Stata Press, College Station, TX). ORs for the association between cervical cancer and specific HPV types were adjusted for age (categorized into 4 age groups: 40, 40 49, and 60 years of age). The adjusted HPV type-specific associations were calculated by comparing women who had a specific HPV type (either by having a single infection with this specific HPV type or a multiple infection containing this specific type) with women who were HPV-negative. HPV attributable fractions (AFs) were calculated according to standard methods, AF p(r 1)/(1 p(r 1)) where p HPV DNA prevalence in the population, approximated by the prevalence in the control group, and r relative risk, approximated by the OR. For the analysis of risk factors other than HPV for cervical cancer, a multivariate model including HPV and the variables associated with risk in the univariate analysis was prepared. RESULTS One-hundred sixteen patients with cervical cancer and 101 controls were recruited. Three case subjects and 6 control subjects were excluded due to inadequate biological specimens. Only specimens that were -globin PCR-positive or were -globin-negative and HPV DNA-positive were included in the analysis; 4 control specimens that were -globin- and HPV-negative were not included in the analysis. Thus, 113 patients with cervical cancer and 91 controls were included in our study. The median age of case subjects (49 years) was higher than, although not statistically different from, that of the controls (45 years). According to histological classification, 106 women had squamous cervical carcinoma and 7 women had adenocarcinoma/ adenosquamous carcinoma. Both types were combined for this analysis because our previous studies show that their risk factors are similar and because the number of adenocarcinomas was too small to conduct separate analyses. Clinical data on the stage of the cancers were available for 109 (96.5%) of the case subjects, and the stage distributions were as follows: 19 case subjects (17.4%) in stage I, 41 case subjects (37.6%) in stage II, 47 case subjects (43.1%) in stage III and 2 case subjects (1.8%) in stage IV. Concerning the 11 cases that were retested with E7 primers, the original classification with L1 primers for these cases was the following: 3 were positive for low-risk HPVs (HPV 11, 26 and 73), 3 were classified as having HPV type X and 5 of them were negative. Results from the retesting of the 3 cases in which low-risk HPV types were detected indicated that one sample with low-risk type HPV 73 was dually infected with high-risk HPV type 59; another sample with low-risk type 11 was dually infected with HPV X (based on a hybridization signal found with GP5 /6 ); and the third sample originally classified as HPV 26 had to be reclassified as -globin-negative and HPV-negative. Two of the 3 HPV X types were further characterized as HPV 18 and HPV 33. Two of the 5 HPV-negative cancers were classified as HPV positive (HPV 16 and HPV 18), while the 3 remaining samples had insufficient material left. Table I shows the prevalence of HPV stratified by cervical cancer case and control status. The overall HPV prevalence was 97% in patients with cervical cancer and 20% in the control subjects. A total of 18 different HPV types were detected in the cervical cancer cases and 11 different types were found in the controls. As a single infection, HPV 16 was the predominant type among both case (47.8%) and control (5.5%) subjects. Among HPV-positive subjects, 49.5% of the case subjects and 27.8% of the control subjects had HPV 16 DNA detected as a single infection. The prevalence of multiple HPV infections was 18.6% among case subjects and 3.3% among control subjects, corresponding to 19.3% and 16.7% of all HPV infections, respectively. Approximately half of the HPV-positive case subjects with multiple infections had

3 488 ROLÓN ET AL. TABLE I HUMAN PAPILLOMAVIRUS (HPV) TYPES IN WOMEN WITH CERVICAL CANCER AND IN CONTROL WOMEN IN PARAGUAY HPV types Number Cervical carcinoma % of total Control subjects % of total HPV-positive 1 Number % of total % of total HPV-positive Negative Positive for any HPV Total N N 18 HPV HPV HPV HPV HPV HPV HPV HPV HPV HPV HPV HPV X HPV 11 X HPV HPV HPV HPV HPV HPV HPV HPV HPV HPV 33 W13B (93) HPV HPV HPV HPV HPV (MM9) Total multiple types Percentage of HPV types is calculated relative to the total HPV-positive and indicates percentage of positive cases in which a specific type or combination of types was detected. simultaneous infections with HPV 16 and 18. All women with multiple infections had at least one high-risk HPV type (16,18, 31, 33, 45, 58 or 59), with the exception of one case with low-risk type 11 and HPV X. For women with single HPV infections, a specific HPV type could not be determined for one HPV-positive cervical cancer subject and one control subject; these specimens were labeled as HPV X. There were no cervical cancer cases who had HPV types 6, 34, 40, 42, 43, 44, 51, 54, 57, 66, 68, 70, 72, IS39, MM4, MM7, CP6108 or CP8304. Thus, among case subjects, the prevalence of high-risk types, low-risk types and HPV X were 94.7%, 0% and 1.8%, respectively. Among control subjects, there was no clear trend in the prevalence of HPV DNA by age, although women over age 60 had a non-significantly higher HPV prevalence when compared with women under age 60 (data not shown). All HPV types detected in older women were high-risk types (16, 18, 31, 33, 39, 45, 52, 58, 59). Among control subjects overall, high-risk HPV types were more common (17.6%) than low-risk HPV types (1.1%) or HPV X (1.1%). HPV positivity among controls was not clearly associated with educational status, number of pregnancies, age at first intercourse, Pap smear history or reported lifetime number of sexual partners (data not shown). HPV positivity was higher among control women who ever used contraceptives (27.6%) when compared with never users (14.6%) and among control women who had never smoked (20.3%) when compared with smokers (12.5%). However, these differences in HPV prevalence were not statistically significant. Table II summarizes the association between specific HPV types and the risk of invasive cancer. The most prevalent types of HPV infection (single or within a multiple infection) were HPV 16, 18, 33, 45, 31 and 58 among case subjects and HPV 16, 58, 31, 33 and 45 among control subjects. ORs were calculated for HPV types identified in a minimum of 4 case subjects; ORs for HPV types 35, 39, 52, 59, 61 and X were not determined due to too few HPV infections. For any HPV type, the age-adjusted OR for invasive cervical cancer was and the AF was 95.8%. Strong ageadjusted associations were noted for HPV 16 (OR 909.8), HPV 18 (infinite OR), HPV 31 (OR 110.1), HPV 33 (OR 260.5), HPV 45 (OR 129.0) and HPV 58 (OR 35.6). The risk of having cervical cancer for women with multiple infections (OR 302; 95% CI: ) was greater than for those with single infections (OR 119; 95% CI: ), although this difference was not statistically significant (data not shown). In the age-adjusted univariate analyses (Table III), the following factors were associated with an increased risk of invasive cervical cancer: younger age at first intercourse, having no education, a higher number of pregnancies, having more than one lifetime sexual partner and never having had a Pap smear. Conversely, ever having used oral contraceptive pills was associated with reduced risk. The hospital of diagnosis or a history of past smoking were not found to be significant (data not shown). Factors that were significantly associated with an increased risk of invasive cervical cancer after adjustment for HPV DNA in the multivariate model included having no education (OR 20.3), never having had a Pap smear (OR 26.7) or having had more than one sexual partner (OR 6.8). DISCUSSION No case-control study of cervical cancer has yet been conducted in Paraguay, a notably high-risk country for cervical cancer. Our study shows strong associations between invasive cervical cancer and HPV types 16 and 18 and also with less common HPV types 31, 33, 45 and 58. Despite the small sample size, the odds ratios for

4 HPV AND CERVICAL CANCER IN PARAGUAY 489 TABLE II NUMBER AND PERCENTAGE OF HUMAN PAPILLOMAVIRUS (HPV)-POSITIVE CASE AND CONTROL SUBJECTS AND ODDS RATIOS (ORS) AND 95% CONFIDENCE INTERVALS (95% CI) FOR AN ASSOCIATION BETWEEN CERVICAL CANCER AND HPV TYPES HPV types Number Cervical cancer cases HPV prevalence % Number Control subjects HPV prevalence % OR a (95% CI) Negative (referent) 1.0 (referent) Positive for any HPV (36 361) Total HPV HPV b HPV HPV HPV HPV Other types (11, 26, 35, 39, 52, 59, 61, 73 MM9, X) a Adjusted for age. b Not adjusted for age. Risk factors TABLE III OTHER RISK FACTORS FOR CERVICAL CANCER Case subjects (N 113) N (%) Control subjects (N 91) N (%) OR 1 (95% Cl) OR 2 (95% Cl) Age at first intercourse: (15.2) 34 (39.5) (31.3) 36 (41.9) 1.9 ( ) 0.8 ( ) (53.6) 16 (18.6) 7.8 ( ) 0.8 ( ) Education: Ever 47 (42.3) 82 (94.3) 1 1 Never 64 (57.7) 5 (5.8) 27.7 ( ) 20.3 ( ) Number of full-term pregnancies: (8.2) 23 (28.0) (19.1) 36 (43.9) 1.5 ( ) 1.4 ( ) 6 80 (72.7) 23 (28.1) 9.3 ( ) 0.8 ( ) p for trend 0.05 Lifetime number of partners: 1 37 (33.6) 67 (76.1) (66.4) 21 (23.9) 7.4 ( ) 6.8 ( ) Oral contraceptive use: Never 11 (13.3) 29 (34.5) 1 1 Ever 72 (86.8) 55 (65.5) 0.3 ( ) 0.4 ( ) Pap smear history: Ever 30 (26.8) 79 (88.8) 1 1 Never 82 (73.2) 10 (11.2) 23.9 ( ) 26.7 ( ) 1 Odds ratio adjusted by age. 2 Odds ratio adjusted by age, human papillomavirus and all other variables in the Table. the associations of invasive cervical cancer with specific HPV types in Paraguay are consistent with those observed in Thailand, the Philippines and Morocco (Chichareon et al., 1998; Ngelangel et al., 1998; Chaouki et al., 1998). The age-adjusted OR for cervical carcinoma associated with HPV positivity overall was 114 in Paraguay. Recent IARC data based on the retesting of 66 initially HPVnegative cervical cancer cases revealed that, after correction for histology and integration of HPV affecting the L1 region, the prevalence of HPV among 1,000 cervical cancer biopsy specimens worldwide is 99% (Walboomers et al., 1999). The determined HPV positivity among cervical cancer cases in this study was 97%. Our results from Paraguay are thus consistent with the notion that the occurrence of HPV-negative invasive cervical cancer is negligible and may be non-existent. Among cervical cancer cases, HPV 16 infection alone was the most prevalent single HPV infection (48%) in Paraguay. More than half of the cases (66%) had infection with either HPV 16 alone, 18 alone or both. Thus, the elimination of these 2 HPV types could lead to significant reductions in the incidence of cervical cancer in this population. Among control subjects, HPV DNA positivity did not appear to generally decrease with age; this may be due to a small number of younger age women in our study. Older women (over age 60) had a higher HPV positivity; although these results are based on small numbers, they are consistent with an increase in HPV prevalence among older women previously reported (Muñoz et al., 1996; Herrero et al., 1999). The overall prevalence of HPV among control subjects in Paraguay (20%) is consistent with that reported from other high-risk countries such as Brazil (17%) and Colombia (17%), Thailand (16%) and Morocco (21%), yet higher than that in the Philippines (9%) and Spain (6%), countries with a lower incidence of cervical cancer. The high rate of cervical cancer in Paraguay in comparison with other countries could be attributable to this high prevalence of HPV infection and the general lack of available and effective cytological screening. The prevalence of HPV infections with dual or multiple types was 16.7% among HPV-positive control subjects in Paraguay, higher than that observed in the IARC studies in Brazil (0%), the Philippines (14.3%), Thailand (9.8%) and Morocco (5.3%) but lower than the prevalence of multiple HPV infections in other studies of cervical neoplasia in younger women (Ho et al., 1998; Kalantari et al., 1997). Due to the slightly higher prevalence of multiple infections among control subjects in Paraguay, the ageadjusted HPV type-specific associations for invasive cervical cancer were recalculated after excluding women having multiple HPV types. All HPV type-specific odds ratios were similar in magnitude and remained statistically significant [any HPV type (OR 118.8; 95% CI: )] (data not shown). Our results do not indicate a statistically significant higher risk of invasive cancer among

5 490 ROLÓN ET AL. women with multiple HPV infections when compared with women with a single HPV infection, indirectly indicating that multiple HPV types do not act synergistically in cervical carcinogenesis (Chaouki et al., 1998). For HPV DNA detection, cervical biopsy specimens from case subjects and exfoliated cells from all control subjects were tested. Previous concern has been raised that there could be a potential higher detection of HPV among cervical cancer cases compared with controls due to these differences in the type of specimens used for HPV ascertainment. However, a validation study indicates that the prevalence of HPV DNA detection among controls should not be significantly underestimated by using cervical exfoliated cell specimens without biopsy specimens. This validation study compared the prevalence of HPV DNA from exfoliated cells with paired biopsy specimens among women with normal cervices undergoing hysterectomy for reasons other than cancer. The HPV prevalence was almost identical in biopsies and exfoliated cells when comparing HPV test results (de Sanjosé et al., 1999). HPV DNA ascertainment in the cases included using MY09 and MY11 primers in 42 cases that were originally -globin- or HPVnegative with GP5 /6 primers. An E7 PCR primer was also used for 11 cases that were HPV-negative, low-risk HPV or HPV X using L1 primers. Although these primers were used for this selected group of cases and not for the corresponding control group, we believe that HPV DNA detection rate among control women should not be notably underestimated because the GP /6 PCR-based assay is very sensitive up to the level of 1 10 femtograms (Jacobs et al., 1995) and because the HPV prevalence among controls was high (19.8%), as expected. In our ongoing work, we used E7 primers to reanalyze specimens that were initially classified as HPV-negative with G5 /6 primers in a small number of HPV-negative cases and HPV-negative control women exhibiting normal cytology. The prevalence of HPV DNA positivity increased among the cases but did not change among the control group. Thus, any potential overestimation of the ORs for invasive cervical cancer due to the different HPV ascertainment methods in these selected cases should be of minor relevance, if any. In addition, after excluding the additional E7 HPV DNA results, the recalculated HPV type-specific odds ratios were not significantly different from those presented in Table II (data not shown). There are limitations in the design of our study. As a hospitalbased case-control study, there is a potential bias related to the selection of the hospital-based controls. A proportion of the controls in this study was primarily seeking cervical cancer screening services. Participating controls may have been more educated, may have had fewer lifetime sexual partners and may have had a higher likelihood of a previous Pap smear than the general population from which the cases were ascertained. We have compared the educational level of women who served as controls in our study with women aged surveyed in the 1990 National Paraguay Demographic and Health Survey (DHS, 1991). Control women in our study with an average age of 45 years had the following distribution of educational attainment: 5.8% had no education, 66.7% had completed primary school and 27.6% had received secondary or higher educational attainment. Women aged included in the DHS survey had a lower educational level than the control women in our study: 20.5% had 0 2 years of schooling, 38.3% had 3 5 years of schooling, 22.8% had completed primary school and 18.4% had completed secondary school or a higher educational level. The control women in our study were characterized by high parity, with a median number of 4 children. Women in the DHS survey were also characterized by a high fertility rate of 4.7 births per woman. However, since this is a not population-based study, controls should be representative of the same population from which the cases were ascertained rather than representative of the population. The significant increased risk of cervical cancer associated with never having had a Pap smear, a lower educational status and a greater number of sexual partners may be partially the result of a potential control selection bias. However, our results are consistent with previous studies using HPV PCR-based assays demonstrating a significant increased risk of invasive cervical cancer in relation to lower educational attainment (Chaouki et al., 1998), a higher number of sexual partners or similar indices (Peng et al., 1991; Chichareon et al., 1998; Chaouki et al., 1998) and a previous history of cytological screening (Chaouki et al., 1998; Ngelangel et al., 1998; Eluf-Neto et al., 1994). With a small sample size, our study has limited power to investigate HPV cofactors or progression factors using a stratified analysis of HPV-positive cases (N 108) and controls (N 18). HPV infection plays a central role in the etiology of cervical cancer in Paraguay, with HPV types 16, 18, 31, 33, 45 and 58 contributing the highest risks. Our data provide additional evidence to consider HPV types 31, 33, 45 and 58 as human carcinogens. Our results are consistent with previous international casecontrol studies, providing further evidence that global cervical cancer control strategies should focus on improving the access to and quality of cytological screening services, while envisioning the use of HPV vaccines and HPV screening for the prevention and detection of specific high-risk HPV-type infections. ACKNOWLEDGEMENTS The authors thank Mr. R. Pol and Mrs. N. Fransen for technical assistance, Ms. D. Magnin for the handling of the specimens and Dr. M. Blettner for her useful comments. REFERENCES CHAOUKI, N., BOSCH, F.X., MUÑOZ, N., MEIJER, C.J.L.M., EL GUEDDARI, B., EL GHAZI, A., DEACON, J., CASTELLSAGUÉ, X. AND WALBOOMERS, J., The viral origin of cervical cancer in Rabat, Morocco. Int. J. Cancer, 75, (1998). CHICHAREON, S., HERRERO, R., MUÑOZ, N., WALBOOMERS, J.M.M., JACOBS, M., BOSCH, F.X., DEACON, J., SANTAMARIA, M., CHONGSUVIVATWONG, V., MEIJER, C.J.L.M. AND WALBOOMERS, J.M.M., Risk factors for cervical cancer in Thailand: a case-control study. J. nat. Cancer Inst., 90, (1998). DE RODA HUSMAN, A.M., WALBOOMERS, J.M.M., MEIJER, C.J.L.M., RISSE, E.K.J., SCHIPPER, M.E.I., HELMERHORST, T.J.M., BLEKER, O., DELIUS, H., VAN DEN BRULE, A. AND SNIJDERS, P. Analysis of cytomorphologically abnormal cervical scrapes for the presence of 27 mucosotropic human papillomavirus genotypes using polymerase chain reaction. Int. J. Cancer, 56, (1994). DE SANJOSÉ, S., BOSCH, F.X., MUÑOZ, N., CHICHAREON, S., NGELANGEL, C., BALAGUERÓ, L., JACOBS, M.V., MEIJER, C.J.L.M. AND WALBOOMERS, J.M.M., Screening for genital HPV: results from an international validation study on HPV sampling techniques. Diagn. dol. Pathol., 8, (1999). DHS: DEMOGRAPHIC and HEALTH SURVEYS, ENCUESTA NACIONAL DE DE- MOGRAFIA Y SALUD, 1990, Institute for Resource Development/Macro Systems, Inc., Columbia, MD (1991). ELUF-NETO, J., BOOTH, M., MUÑOZ, N., BOSCH, F.X., MEIJER, C.J.L.M. AND WALBOOMERS, J.M.M., Human papillomavirus and invasive cervical cancer in Brazil. Brit. J. Cancer, 69, (1994). HERRERO, R., HILDESHEIM, A., BRATTI, C., SHERMAN, M., HUTCHINSON, M., MORALES J., BALMECEDA, I., GREEDBERG, M.D., ALFARO, M., BURK, R.D., WACHOLDER, S.& SCHIFFMAN, M.H. (1999) A population-based study of all grades of cervical neoplasia in rural Costa Rica. J. Natl Cancer Inst. (in press) HO, G.Y.F., KADISH, A., BURK, R., BASU, J., PALAN, P, MIKHAIL, M., S. ROMNEY, HPV 16 and cigarette smoking as risk factors for high-grade cervical intra-epithelial neoplasia. Int. J. Cancer, 78, (1998). IARC, Globocan I, Cancer Incidence and Mortality Worldwide, (IARC Cancer Base No. 3, International Agency for Research on Cancer, INTER- NATIONAL AGENCY FOR RESEARCH ON CANCER, Lyon (1998). IARC, Monographs on the evaluation of the carcinogenic risks to humans, Vol. 64, Human papillomaviruses. International Agency for Research on Cancer, Lyon (1995). JACOBS, M.V., DE RODA HUSMAN, A.M., VAN DEN BRULE, A.J.C., SNIJDERS,

6 HPV AND CERVICAL CANCER IN PARAGUAY 491 P.J.F., MEIJER C.J.L.M. AND WALBOOMERS, J.M.M., Group-specific differentiation between high- and low-risk human papillomavirus genotypes by general primer-mediated PCR and two cocktails of oligonucleotide probes. J. clin. Microbiol., 33, (1995). KALANTARI, M., KARLSEN, F., JOHANSSON, B., SIGURJONSSON, T., WAR- LEBY, B. AND HAGMAR, B., Human papillomavirus findings in relation to cervical intraepithelial neoplasia grade: a study on 476 Stockholm women, using PCR for detection and typing of HPV. Hum. Pathol., 28, (1997). MANOS, M.M., TING, Y., WRIGHT, D.K., LEWIS, A.J., BROKER, T.R. AND WOLINSKY, S.M., The use of polymerase chain reaction amplification for the detection of genital human papillomaviruses, molecular diagnostics of human cancer. Cancer Cells, 7, (1989). MUÑOZ, N. AND 17 OTHERS, The causal link between human papillomavirus and invasive cervical cancer: a population-based case-control study in Colombia and Spain. Int. J. Cancer, 52, (1992). MUÑOZ, N. AND 14 OTHERS, Risk factors for HPV DNA detection in middle-aged women. Sex. Trans. Dis., 23, (1996). MYERS, G., HALPERN, A., BAKER, C., MCBRIDE, A., WHEELER, C. AND DOORBAR, J.(EDS.) Human papillomavirus compendium: a compilation and analysis of nucleic acid and amino acid sequences. Los Alamos National Laboratory, Los Alamos, NM (1996). NGELANGEL, C., MUÑOZ, N., BOSCH, F.X., LIMSON, G.M., FESTIN, M.R., DEACON, J., JACOBS, M., SANTAMARIA, M., MEIJER, C.J.L.M. AND WAL- BOOMERS, J.M.M., The causes of cervical cancer in the Philippines: a case-control study. J. nat. Cancer Inst., 90, (1998). PARKIN, D.M., WHELAN, S.L., FERLAY, J., RAYMOND, L. AND YOUNG, J., Cancer incidence in five continents, Vol. VII. IARC Sci. Publ., 143, International Agency for Research on Cancer, Lyon (1997). PENG, H., LIU, S., MANN, V., ROHAN, T. AND RAWLS, W., Human papillomavirus types 16 and 33, herpes simplex type 2 and other risk factors for cervical cancer in Sichuan Province, China. Int. J. Cancer, 47, (1991). WALBOOMERS, J.M.M., JACOBS, M.V., MANOS, M.M., BOSCH, F.X., KUM- MER, J.A., SHAH, K.V., SNIJDERS, P.J.F., PETO, J., MEIJER, C.J.L.M. AND MUÑOZ, N., Human papillomavirus, a necessary cause of invasive cervical cancer worldwide. J. Pathol., 189, (1999). WALBOOMERS, J.M.M, MELKERT, P., VAN DEN BRULE, A.J., SNIJDERS, P.J., MEIJER, C.J.L.M., The polymerase chain reaction for human papillomavirus screening in diagnostic cytopathology of the cervix. In: C.S. Herrington and O. McGee (eds.), Diagnostic molecular pathology: a practical approach, pp , Oxford University Press, Oxford (1992).