The Caribbean populations consisted of 212 women from Tobago  and 99 women from Jamaica . The US population consisted of 88 women from Pittsburgh. All of the subjects from the three geographic areas were recruited from the general population and none were pregnant. The women from Tobago were recruited between July and September 2004 by means of posters, flyers, public service announcements on television and radio, word of mouth, and a series of cancer information sessions conducted throughout the island. The Jamaican women were recruited during visits to a family practice in Western Jamaica between January 2003 and October 2006. All Jamaican participants were recruited consecutively to avoid selection bias. The US study population was enrolled from May 2007–Aug 2008 as part of a nested cross-sectional study of the Epidemiologic STudy of HEalth Risk (ESTHER) project, (an ongoing women's study at the University of Pittsburgh). The ESTHER project's sample recruitment methodology attempted to address some of the biases related to use of a sample of convenience. The targeted recruitment for ESTHER was based on age, level of education, and race/ethnicity population distributions. Recruitment strategies included attending women focused events, news items and advertisements in women-focused news sources, and limited respondent driven sampling.
All subjects from Jamaica and Tobago combined are referred to in this study as Caribbean. The majority of subjects were African-Caribbean, although in Tobago, there were 19 subjects who defined their ethnicity as East Indian. The self-reported ethnic groups in the US population were classified as White, Black or other ethnic groups. One subject did not define her ethnicity and was included in the other ethnic group category.
Standardized protocols were implemented for sample collection, DNA extraction and HPV testing for all three sub-populations that were included in this analysis. A nurse or clinician collected cervical brush samples and the DNA was extracted from these samples using the Puregene DNA purification kit (Qiagen, Germantown, MD, USA). The HPV genotyping was performed on these samples using the Linear Array HPV Genotyping kit (Roche Diagnostics). All cervical samples were tested at the University of Pittsburgh. The assay involved amplification of samples by PCR using a master mix which contained biotin-labeled primers for the detection of the 37 most common HPV genotypes as well as the human beta-globin gene. The PCR products were chemically denatured and hybridized for 30 min at 53°C to linear array strips which contained specific and one cross-reactive oligonucleotide probe for the HPV genotypes as well as a high and low concentration of a beta-globin probe. The HPV genotypes were identified when visualized using a streptavidin-horseradish peroxidase conjugate and a substrate solution containing hydrogen peroxide and 3,3',5,5'-tetramethylbenzidine which yielded a blue precipitate at the positions where the hybridization occurred.
For six samples from the US population, beta-globin was not positive, which indicated either poor quality or low yield DNA samples. These samples were excluded. Therefore, HPV results were available for only 82 US samples. All of the samples from Jamaica were tested with the AMPLICOR HPV method  prior to testing with the Linear Array HPV Genotyping kit. Ten samples from the Jamaican population tested HPV-negative using our assay, possibly due to a lower sensitivity. Therefore, for our study, these samples were classified as HPV X.
The Linear Array protocol does not specifically detect HPV52. The cross-reactive probe in this assay detects HPV 33, 35, 52 and 58 combined. According to the manufacturer's protocol, samples that are negative for HPV 33, 35, and 58 individually, but positive for the cross-reactive probe are classified as HPV 52-positive. Samples that are positive for HPV 33, 35, and/or 58 individually, as well as the cross-reactive probe have an uncertain HPV 52 status. For our study these samples were considered negative for HPV 52.
HPV-risk classification was based on the epidemiological classification of HPV types that are associated with cervical cancer . Briefly there are fifteen high-risk HPV types: 16, 18, 31, 33, 35, 39, 45, 51, 52, 56, 58, 59, 68, 73, 82; three probable high-risk HPV types: 26, 53, 66 and twelve low-risk HPV types: 6, 11, 40, 42, 43, 44, 54, 61, 70, 72, 81, CP6108. For this analysis, we combined probable high-risk HPV types and high-risk HPV types into a single group. The HPV types that do not fall into any of these aforementioned categories were classified as undetermined risk.
All statistical analyses were performed using STATA SE (version 10), (StataCorp LP, College Station, TX). Age-adjusted prevalence rates and confidence intervals were calculated using logistic regression estimates of infection with any HPV type or any high-risk HPV type, adjusted for age. Observations with missing values of age were dropped from the analysis before estimation. The chi-squared test of proportions was used to calculate p-values for the unadjusted differences in proportions as well as, p-values for the differences in adjusted proportions. Although age and age at first sexual intercourse were both normally distributed, we rejected the hypothesis of equal variances. Therefore, the comparisons of mean age and age at first sexual intercourse were performed using two-sample t-tests on the equality of means adjusting for unequal variances.