Study population and follow-up visits
Women were recruited and followed up between September 2002 and December 2006. A description of the study population at baseline is available elsewhere . Briefly, participants were recruited among women aged 12-24 years presenting themselves for health services at Naguru Teenage Information and Health Centre (NTIHC) located in the suburbs of Kampala City. Women were eligible for participation if they were sexually active, resided within 20-kilometer radius of NTIHC and had no plans to relocate for the duration of the study. Follow-up visits were scheduled between 6-12, 13-18, and 19-24 months from baseline. Twenty-nine women who turned up for visits after 24 months were not turned away, although these visits were unscheduled.
Gynaecological examination, collection of exfoliated cells and cytological assessment
During all visits, trained midwives explained the study aims and procedures, which included answering a questionnaire and having a pelvic examination to record visible abnormalities. In all visits after visual inspection of the vulva, a non-lubricated sterile speculum was inserted, and cervical exfoliated cells were collected. At the first visit at cohort enrolment, cervical exfoliated cells were collected with a sterile swab (Copan International) which was then placed in a labelled 15 mL holding tubes containing 5 mL of phosphate buffered saline (PBS), pH 7.2; samples were kept temporarily at 4°C for an average of 6 hours and then transferred to a freezer for storage at 20°C until shipment to the laboratory for HPV analysis. PBS is not a suitable solution for preservation of cells for liquid based cytology and consequently no liquid based cytology could be done with baseline samples. However, during follow-up, cervical exfoliated cells were collected and stored in a vial containing PreservCyt solution (Cytyc, Boxborough, MA) which is suitable both for liquid-based cytology and HPV testing. Briefly, for these follow-up samples, a broom style cytobrush (Cervex brush, Rovers Medical Devices B.V., Oss. The Netherlands) was inserted deep into the endocervical canal and rotated gently in a clockwise direction 5 times to collect cells from the endo- and ectocervix. The cytobrush containing cervical cellular material was then placed in a vial containing PreservCyt solution (Cytyc) and rinsed by pushing it to the bottom of the vial 10 times. The cytobrush was then discarded. The vials were closed and kept at room temperature until shipment in dry ice to DDL Diagnostic Laboratory, Voorburg, The Netherlands, for liquid based cytology and HPV testing. Cervical abnormalities in liquid-based cytology samples were read at the cytology department of the Slotervaart Hospital, Amsterdam, The Netherlands and classified according to the 2001 Bethesda Classification .
Before removal of the speculum, both for the baseline and follow up visits, visual inspection with acetic acid (VIA) and with Lugol's Iodine (VILI) was performed. A urine sample and 4 ml of blood (in heparinised tubes) were collected for pregnancy, HIV and syphilis testing, respectively. Non-pregnant women with external genital warts were treated with 2% podophylline paint. The treatment of pregnant women was deferred until after delivery. Women with vaginal discharge and cervicitis received a one-week syndromic treatment and were requested to ask their sexual partners to receive treatment.
Isolation of HPV DNA
Total DNA was isolated from 200 μl of the suspension containing the cervical cells by the MagNa Pure LC instrument (Roche Diagnostics, Almere, The Netherlands), using the Total NA isolation kit (Roche Diagnostics, Almere, The Netherlands). DNA was eluted in 100 μl of elution buffer, and 10 μl was used for each PCR reaction. Each run contained positive and negative controls to monitor the DNA isolation, PCR and HPV detection and genotyping procedures.
The short PCR fragment (SPF) 10 primer set was used to amplify a broad spectrum of HPV genotypes, as described earlier [11, 12]. Briefly, this primer set amplifies a small fragment of 65 base pairs (bps) from the L1 region of HPV. Reverse primers contain a biotin label at the 5' end, which enables capture of the reverse strand onto streptavidin coated microtiter plates. Captured amplimers are denatured by alkaline treatment and a defined cocktail of digoxigenin - labeled probes, detecting a broad spectrum of HPV genotypes. This method is designated HPV DNA enzyme immunoassay (DEIA), and provides an optical density value. The same (SPF) 10 amplimers were used to identify the HPV genotype by reverse hybridisation on a line probe assay (LiPA) containing probes for 25 different genotypes [(SPF) 10 HPV LiPA version 1, Labo Bio-medical Products, Rijswijk, the Netherlands]. HPV 16, 18, 31, 33, 35, 39, 45, 51, 52, 56, 58, 59 and 68/73 were considered high-risk types and 6, 11, 34, 40, 42, 43, 44, 53, 54, 66, 70 and 74 were considered low-risk types . Samples that were positive using the (SPF) 10 primer set but did not reveal any of the 25 aforementioned types, were provisionally classified as positive for HPV X and were subjected to a second round of testing, similar to the one already described, for 17 additional types including HPV 26, 30, 55, 61, 62, 64, 67, 69, 71, 82, 83, 84, 85, 87, 89, 90 and 91.
HIV-1 testing was performed at NTIHC following the National HIV Rapid Testing algorithm consisting of Determine rapid test (Abbot Diagnostics, Abbot Park, IL) as screening test, Statpak rapid test (ChemoBio Diagnostics, Systems, Inc. Medford, NY) as the confirmatory test and Unigold (Orgenics, Waltham, MA) as tie breaker . For quality assurance and quality control, all HIV-positive results were confirmed by ELISA (Cambridge Bioscience, Cambridge, UK) or PCR assays (Roche Molecular Systems, Pleasanton, CA) at Makerere Medical School. Women found to be HIV-positive were referred to health institutions offering treatment, care and support services.
Syphilis testing was performed at NTIHC using a commercially available standard Rapid Plasma Reagin (RPR) 18 mm circle card test (Quorum Diagnostics, Sacramento, CA). Samples with reactive RPR test results were referred to established laboratories for confirmation before treatment was offered to the women concerned and their partners.
All women were tested for pregnancy using a commercially available HCG dipstick pregnancy test (Cypress Diagnostics, Langdorpsesteenweg, Belgium), and the results were communicated immediately to each woman. By the end of the study 30 women were tested positive for pregnancy. The pregnant women were referred for free pre-natal counselling, evaluation and care within the same health care center. Only the pre-natal care services are being evaluated in our study, not the number of weeks of pregnancy or the decision if the pregnancy was or not to be kept.
Women consented to participation in the study. The women younger than 18 years of age were treated according to the Uganda guidelines for conducting research on human beings that allows minors to assent for participation without a parent or legal guardian agreement. These guidelines are approved by the ethical review committees at the Faculty of Medicine, Makerere University, the Uganda National Council of Science and Technology, and the International Agency for Research on Cancer.
Statistical analyses and definitions of outcomes
We analyzed incidence and clearance of HPV infection for all women who had at least one follow-up visit. Any HPV type was defined as a combined outcome if any HPV type was present. We defined HPV 16-related types as HPV 16, 31, 33, 35, 52, and 58 and HPV 18-related types as HPV 18, 39, 45, 59 and 68. Single and multiple infections were defined without considering uncharacterized HPV types (HPV X). We computed relative risk (RR) for HPV incidence and clearance using Poisson regression dividing the follow-up time in 6-month windows for risk time and events.
For incidence analysis, the risk time was measured for each specific HPV type or combined set of HPV types from the first time a woman tested negative until a positive test or the last visit.
For each of the groups of infections (low-risk, high-risk, HPV 16-related, HPV 18-related and any HPV) incidence was calculated for women free from all the infections in each group until infection of at least one of the HPV types in the group. Correspondingly clearance was calculated for all women infected with at least one of the HPV types in each group until the woman was free from infection of all the HPV types in each group. Thus, clearance will typically be a lot longer than for specific HPV types. This choice of groups of HPV types intend to measure an exposure assuming interchangeable risks within the group so a woman was not considered cleared from exposure of high-risk HPV if one high-risk infection was cleared while she was still infected with other high-risk type(s). Thus, clearance of any HPV was interpreted as being cleared from all HPV infections. Highly sexually active women as our study population would tend to have long clearance time even if each specific type is cleared fast.
Women positive for HPV X at baseline who tested positive for a known HPV type at follow-up visit were considered to have developed a new infection in the analysis of incidence but were excluded from the analysis of clearance as persistence of HPV X could not be known. For each HPV type, we fitted Poisson models including a time covariate only and models adjusting for HIV status, age at baseline (12-17, 18-20, 21-24 year), positivity for genital warts or syphilis (yes/no), and lifetime number of sexual partners (1, 2, 3, ≥4).
We evaluated prevalent cytological abnormalities only on follow-up visits as the medium used to collect baseline specimens was not suitable for cytological analysis. Subjects were classified as normal or low-grade squamous intraepithelial lesions (LSILs). No high-grade squamous intraepithelial lesions or invasive cervical cancer cases were diagnosed. Among women with a baseline HPV negative sample with valid follow-up cytological sample we evaluated incident cytological abnormalities according to type-specific HPV infections at follow-up (HPV 16 infected vs. non-infected, HPV 18 infected vs. non-infected, high-risk infected vs. non-infected, low-risk infected vs. non-infected, HPV 16-related types infected vs. non-infected, HPV 18-related vs. non-infected and infection with any HPV type vs. non-infected).
All tests of statistical hypotheses were made on the two-sided 5% level of significance with corresponding 95% confidence intervals. All analyses were made using the SAS software version 9.2 (SAS Corp). Procedure genmod was used to fit the Poisson regression.