We collected a random sample of 17
949 women aged 20-29 years from the general population in Copenhagen using the central personal registry. Every citizen in Denmark has a unique 10 digit identification number (CPR number), which is universally used in the public administration. These identification numbers, which comprise information on sex and date of birth, are registered in the computerised central personal registry. The register is updated daily and contains information on vital status and migration, including the current address. We invited all eligible women to a study clinic established at one of the university hospitals in Copenhagen. Recruitment was from May 1991 to January 1993. We included 11
088 women in the study, all of whom gave informed consent. A detailed description of the enrolment procedure is provided elsewhere.7
The study was approved by the local ethics committee.
Examination at enrolment
At enrolment all 11
088 women were interviewed personally by specially trained female nurses. The nurses collected data on demographic variables, smoking, reproductive background, contraception, sexual habits, previous sexually transmitted diseases, and history of cervical smear tests. The participants also had a gynaecological examination, in which we carried out a smear test and obtained endo-ectocervical cells for detection of HPV DNA. All swabs were placed in a tube with TE-buffer (10 mM Tris-HCl and 1 mM EDTA, pH=8.0). In addition, all participants gave two blood samples. All biological material was kept at −80°C until tested.
Examination at follow up
In October 1993 we invited the entire cohort for a second examination. Initially, the cohort was linked to the central personal register using the CPR number as key identifier. We traced all the women in the cohort using this register and retrieved information on vital status and current address. We invited the women to participate in the second phase in the same order as they were originally enrolled in the study. During the following 18 months (that is, until January 1995) 8656 women (78%) underwent this second examination. Women were interviewed about suspected risk factors for cervical cancer, focusing on the time between enrolment (first examination) and follow up (second examination). We also did a smear test and took cervical swabs for HPV testing (placed in phosphate buffered saline with 0.05% methiolate) and two blood samples from each woman using the same procedure as at the initial examination, all biological material being stored at −80°C.
Passive follow up
We also had the cohort under passive surveillance for occurrence of abnormal cytology. In a high proportion of Danish counties, all cytological and histological diagnoses are registered in a computerised pathology register (the smears taken in the present study were also registered in the pathology register). In November 1995 we linked the original cohort of 11
088 women to the pathology register files, and all women were traced in the register. Although the Danish Board of Health recommends cervical smear testing every three years, many women tend to get screened more often.8
By means of the pathology register we were able to get information about all such examinations on every woman in our study since their first smear test and up to the date of the register linkage.
We excluded women with a history or current evidence of cervical neoplasia. The figure shows the different exclusions for the entire cohort. We excluded 11 women in whom cervical neoplasia had been diagnosed in the first nine months (the time was chosen to ensure comparability with another study4
) to avoid inclusion of potentially prevalent cases in the study (four had atypical squamous cells of undetermined significance, four had low grade lesions, three had high grade lesions). After all exclusions 10
177 women remained in the follow up study.
Identification of potential cases
A total of 428 potential cases developed in the cohort. Of these, we identified 329 at the second examination. The linkage with the pathology register resulted in 99 more women with an incident diagnosis of lesions on the uterine cervix that qualified them as potential cases in the time period between nine months after the first examination (that is, enrolment in the cohort) and November 1995. The cytological diagnoses covered a spectrum from “non-specific viral changes, not further specified,” “koilocytosis,” and “atypia” to dysplasia (mild, moderate, severe) and carcinoma in situ.
Review and confirmation of case diagnoses
From the files of the pathology register we identified the microscopy number on all abnormal smear results (and biopsies if taken) as well as on the enrolment smear and smears taken and diagnosed as negative during follow up from every potential case. The smear samples and biopsy slides were subsequently located and retrieved from the respective pathology departments. This material was reviewed in a masked fashion by one pathologist (GP) using the Bethesda nomenclature system.9
In cases of discrepancy between the original diagnosis and the review diagnosis, another pathologist (MES), who was unaware of any of the two previous diagnoses, reviewed the material. In most cases there was agreement at the first review, and in the remaining cases there was agreement between two of the three pathologists (GP, MES, PAP).
In cases where both cervical smear result and biopsy contributed to the diagnosis, the more severe diagnosis formed the basis of the final diagnosis. Among 428 potential cases, 58 were downgraded to normal in the review procedure, and 370 (86%) had a confirmed diagnosis of incident atypical squamous cells of undetermined significance or cervical neoplasia. This covered 40 with atypical squamous cells, 165 with low grade squamous intraepithelial lesions, and 165 with high grade squamous intraepithelial lesions (figure). Histological examination results to confirm the diagnosis were available in 136 (83%) high grade cases and 60 (35%) low grade cases. None of the smear results that were originally negative at enrolment were upgraded at the review procedure.
We randomly selected a sample of 1000 from the 10
758 women in the entire cohort who had cytologically normal results at enrolment. We retrieved the smear samples taken at enrolment and during follow up from the files of the pathology departments, and they were reviewed by one pathologist (PAP). In cases of discrepancy between the original diagnosis and the reviewed diagnosis, another pathologist blindly reviewed the smear (MES).
We excluded 39 women from the subcohort because of previous cervical neoplasia (n=38) or abnormal cytology detected within nine months of enrolment (n=1). During follow up, 40 women had an abnormal smear test result, and we included them in the group of potential cases. This left 921 women without any history of cervical neoplasia (that is, no history of cervical neoplasia before enrolment and no abnormal cervical cytology during follow up). At the review procedure none of the enrolment or follow up smears was upgraded.
Final study population
We excluded cases diagnosed later than three months after the follow up examination. This time limit was chosen so the HPV status at the follow up visit would still reflect the status at diagnosis. In the analyses including HPV status at follow up, we excluded cases that were diagnosed before the follow up examination and in which cervical biopsies or surgical treatment (cone) had been carried out (two with atypical squamous cells, 13 with low grade lesions, 19 with high grade lesions). For us to define women in the subcohort as “cytologically normal” we considered that they had to have a normal cervical smear result at or after the follow up examination. On the basis of these restriction criteria, we excluded 15 women with atypical squamous cells, 46 with low grade lesions, 51 with high grade lesions, and 265 controls from the analyses. We excluded four other women with low grade lesions, two with high grade lesions, and three controls because their cervical swabs were inadequate for HPV analysis. Thus, the final study population comprised 252 incident cases (25 with atypical squamous cells, 115 with low grade lesions, and 112 with high grade lesions) and 653 cytologically normal women (see figure). Among the cases, 191 (76%) women were identified at the second examination and 61 women with an incident diagnosis of cervical neoplasia were identified from the pathology register linkage.
HPV DNA detection
The cervical samples were analysed by the general primer GP5+/6+ mediated polymerase chain reaction-enzyme immunoassay method.10
Briefly, we added 10 μl of the crude cervical cell suspension to the polymerase chain reaction mixture (10 mM TRIS HCl, pH 8.3; 50 mM KCl; 3.5 mM MgCl2; 1 unit of thermostable DNA polymerase (Amplitaq, Perkin Elmer Cetus, Norwalk, CT); 200 μmol of each dNTP; and 25 pmol of each primer (GP5+ and biotinylated GP6+)). We incubated the mixture for five minutes at 94°C for DNA denaturation, followed by 40 cycles of amplification with a polymerase chain reaction processor (Biomed, Theres, Germany). Each cycle included a denaturation step to 94°C for one minute, an annealing step to 40°C for two minutes, and a chain elongation step to 72°C for 90 seconds. To ensure a complete extension of the amplified DNA we prolonged the final elongation step by four minutes.
We analysed the biotinylated GP5+/6+ polymerase chain reaction products by enzyme immunoassay using HPV high risk (HR) and HPV low risk (LR) oligococktail probes to identify 14 high risk HPV types (HPV 16, 18, 31, 33, 35, 39, 45, 51, 52, 56, 58, 59, 66, 68) and six low risk types (HPV 6, 11, 40, 42, 43, 44). We also typed the high risk and low risk positive swabs individually using specific enzyme immunoassays. In addition, we analysed GP5+/6+ polymerase chain reaction products for the presence of other HPV types not identified by the high risk and low risk enzyme immunoassays; this was done with gel electrophoresis, followed by Southern blot analysis under low stringent conditions with a cocktail probe of different HPV types.11
We classified samples that were positive by this Southern blot analysis but negative by both high risk and low risk enzyme immunoassay as HPV X positive.
We investigated the associations between squamous intraepithelial lesions and HPV DNA detected at the two examinations by multiple logistic regression analyses performed separately for each type of lesion compared with the controls (subcohort). This corresponds to being either cytologically normal or having a specific case type in the full generalised logistic regression model considering all four outcome categories (normal, atypical squamous cells, low grade lesions, and high grade lesions) simultaneously, and makes the estimates directly comparable with case-control studies of any of the single adverse outcomes. We corrected all analyses for age at enrolment as a categorical variable, grouped in yearly intervals. The 95% confidence intervals were based on Wald's test performed on the log transformed odds ratios and back transformed.
We classified HPV types in relation to their association with cervical cancer. HPV types 16, 18, 31, 33, 35, 39, 45, 51, 52, 56, 58, 59, 66, and 68 were grouped together in a high risk (“oncogenic”) group, and HPV types 6, 11, 40, 42, 43 and 44 were placed in the low risk (“non-oncogenic”) group. The uncharacterised HPV types (HPV X) were grouped together with the low risk types. Women with multiple types were grouped according to the highest risk group.
We performed the statistical analyses both on the unrestricted study population (that is, 40 with atypical cells, 165 with low grade lesions, 165 with high grade lesions, and 921 without any history of cervical neoplasia), on the final study population (25, 115, 112, and 653 respectively) (figure), and with intermediate restriction criteria for cases and the subcohort (data not shown). The overall pattern of the results was the same irrespective of the restriction criteria; however, the strength of association in the analyses including HPV status at follow up increased with the severity of the inclusion criteria for the subcohort and decreased with the length of time between the second examination and the case diagnosis (data not shown). We have presented only those results concerning the final (restricted) study population.
Finally, we investigated whether the restrictions leading to the final study population were associated with different distributions of some potential confounders registered at enrolment (age, number of sexual partners, age at first intercourse, and use of oral contraceptives). Thus, we compared the women in the final study population with those who were excluded from the unrestricted study population within the control group and within the case groups, and the respective distributions were nearly the same (data not shown).