The aim of this study was to determine whether there were any differences in HPV type prevalence by area of residence or ethnicity, and this aim guided our choice of methods. Engagement and recruitment of Indigenous women in remote areas, by providing the opportunity to participate meaningfully in improving their own health and that of their communities, was recognized as crucial to the success of the study. Therefore, the sampling strategy was guided by the use of methods that would include these women in a culturally appropriate way.
Consultation, ethics, pilot site, and development of culturally appropriate materials
In recognition of the need for research to benefit Indigenous communities, and for Indigenous people to be involved in the design, planning, undertaking and dissemination of Indigenous research, we identified collaborative research partners involved in frontline community care and service provision in the area of cervical cancer prevention in Indigenous communities. Prior to initiation of the study, we undertook consultation with Indigenous communities (including the establishment of an Indigenous Steering Committee), medical services, healthcare workers, public-health practitioners and servicing cytology groups, commencing in January 2005, including on-site workshops and subsequent staff training. Protocols and communication materials (including flipcharts and brochures) were developed to support the study, with local input at different sites to tailor the materials so that they would be locally and culturally appropriate as needed. A pilot study was conducted in central Australia between April and July 2005 to examine feasibility, communication materials and protocols, and laboratory techniques: it was endorsed by the Central Australian Human Research Ethics Committee. The multisite study was approved by 34 site ethics committees. Women were enrolled into the study between July 2005 and February 2008.
Study population, recruitment methods, consent and specimen collection
Women aged 18 to 40 years who were attending their usual healthcare provider for routine Pap smear cytology were invited to participate in the study. Some women aged under 18 years, who were assessed as mature minors and hence competent for giving consent without parental or guardian involvement and attending for a routine Pap test, were also enrolled by providers. Although most women in Australia see a general practitioner (primary-healthcare providers) for Pap-test screening, our primary focus was upon adequate recruitment of Indigenous women, and therefore we targeted Indigenous health providers and clinics, and equivalent service providers (community-based, no fee for service) for non-Indigenous women, to ensure comparability of Indigenous and non-Indigenous women in the study. The study sample size was determined based on the HPV prevalence estimates from a large community-based study in the USA of women aged 18-40 attending for a Pap smear and was powered (α = 0.05, 1-β = 0.90) to detect an absolute difference in HPV16/18 of ≤ +/-5% between Indigenous and non-Indigenous women, and between Indigenous women residing in remote areas and other Indigenous women. We determined the sample size necessary at each site according to the predominant client population of the site (Indigenous or non-Indigenous) and the remoteness category of the site according to the Accessibility/Remoteness Index of Australia (ARIA). The ARIA assigns a 'remoteness score' based on each locality's relative access to services, which is estimated by road distance (or other transport mechanism for islands) to the nearest population centers of several different sizes. These scores are grouped into five categories: major cities, inner regional, outer regional, remote, and very remote. Approximately two-thirds of the Australian population , but only one-third of the Indigenous population, live in major cities 
We recruited women in all states and one territory (Northern Territory) of Australia from sixteen Indigenous health services, eight family-planning services and ten community clinics. Owing to the incorporation of the study into routine clinical care, response rates (number enrolled divided by number approached) were not accurately recorded at all sites. Where response rates were recorded or estimated from the number of eligible (by age) women who had a Pap test taken during the recruitment period, these ranged from 46% to 87%. Recruitment of Indigenous women in some sites was challenging, and required further strategies for community engagement with the study to be developed[20, 21].
Women were asked to allow the cervical samples collected for Pap cervical cytology to be tested for HPV DNA also. Once fully informed consent was obtained, Pap smear samples were collected in the routine clinic settings. In addition to routine pathology specimen details, information confirming Indigenous status, most recent previous Pap-test result, pregnancy status, smoking status and hormonal contraceptive use was requested from participants. Following the testing of specimens, identifying details were used to report Pap cytology and high-risk (HR) HPV DNA results back to the clinician.
Most specimens were collected using a brush (Cervex; Rovers Medical Devices BV, The Netherlands) and a spatula. The sample was then smeared onto a glass slide and fixed with ethanol ready for conventional Pap cytology, and the remainder was rinsed into vials containing fixative solution (ThinPrep vials with PreservCyt fixative; Cytyc Corp., Boxborough, MA, USA). In a minority of samples, LBC was standard, and hence an aliquot was removed using good microbiological practices within a biohazard cabinet, prior to processing for cytology. All samples were transported to the Regional WHO LabNet for HPV located at the Royal Women's Hospital in Melbourne, Australia, for HPV DNA detection and genotyping.
Algorithm for feedback of results and follow-up by clinicians
Results of the HR HPV testing were returned to the woman's healthcare provider so that they could be conveyed back by trained staff to each participant in a culturally appropriate way. All women were managed according to National Health and Medical Research Council guidelines for the follow-up of abnormal Pap tests. Currently, HPV DNA testing in clinical practice in Australia is only recommended as a "test-of-cure" following local ablative or excisional therapy for precursor cervical cancer lesions of high-grade (2+) cervical intraepithelial neoplasia (CIN).
HPV DNA detection and genotyping
DNA was extracted from stored specimens (in PreservCyt) using a commercial isolation and purification system (MagNA Pure LC; Roche Molecular Systems, Alameda, CA, USAusing 1-ml aliquots of the preserved specimens as previously described[23, 24]. From a final volume of 100 μl of resultant DNA, 10 μl of extracted DNA was screened (AMPLICOR HPV test; Roche Molecular Systems) for detection of 13 HR HPV genotypes (types 16, 18, 31, 33, 35, 39, 45, 51, 52, 56, 58, 59 and 68). As this test detects only 13 HR HPV types, all samples testing negative for HR HPV by this test were also tested to exclude presence of other types, using a 20 μl aliquot of extracted DNA in a PGMY09/11-based HPV consensus PCR assay. A PCR-ELISA detection protocol was used, as described previously. All assays used incorporated the amplification of the β-globin gene as an internal control.
All samples that tested positive by the AMPLICOR or PGMY09/11 PCR test were genotyped (HPV Linear Array® (LA) Genotyping Test; Roche Molecular Systems), using 50 μl of extracted DNA, and following the manufacturer's instructions with minor modifications as previously reported[24, 27, 28]. LA identifies 37 genotypes: 6, 11, 16, 18, 26, 31, 33, 35, 39, 40, 42, 45, 51, 52, 53, 54, 55, 56, 58, 59, 61, 62, 64, 66, 67, 68, 69, 70, 71, 72, 73, 81, 82 (previously known as IS39), 83, 84 and 89 (previously known as CP6108) . There were 13 specimens that were weakly positive only to the PGMY09/11 PCR but negative on LA, and these specimens were classified as HPV negative.
Specimen contamination and carryover were prevented by use of barrier tips, prior division of all reagents into aliquots, and performance of pre- and post-PCR steps in different rooms specifically allocated for PCR. Negative and positive controls were processed with each run, and lack of signal in the negative control was used to monitor possible carryover.
The ARIA category was assigned using the postcode of each participant's residence. Pap-test results were reported by cytology laboratories (usual service providers) according to the Australian Modified Bethesda classification, 2004 . For the purposes of analysis, 'possible' low-grade squamous intraepithelial lesion (LSIL) Pap results were grouped LSIL and 'possible' high-grade squamous intraepithelial lesion (HSIL) Pap results were grouped with HSIL.
Age-adjusted prevalences were calculated by weighting the WHINURS sample to the relevant Australian Bureau of Statistics (ABS) population structure, by single year of age. Data were analyzed using SPSS software (version 17.0; SPSS Inc., Chicago, IL, USA), except for weighted confidence intervals, which were calculated using STATA software (StataCorp LP, College Station, TX, USA). Univariate analysis compared proportions using Pearson's χ2 tests (or Fisher's exact test when the expected count in any cell was < 5), and mean ages using t-test. Binary logistic regression was used for multivariate analysis. Interaction terms were created for predictor variables to check for effect modification. No significant interactions were identified. Because of the high correlation between living in a remote area and Indigenous status, the effect of area of residence was analyzed separately for the two groups of women, with the analysis of non-Indigenous women comparing residence in a major city versus non-major city.
Indigenous status and area of residence were variables of prior interest. Multivariate models also included other known determinants of HPV infection: age, smoking status, and use of hormonal contraception. We did not collect sexual history from participants. Although some women were pregnant, there was no association between pregnancy and HPV detection in univariate analysis, and hence this was not considered further.