Four pathology laboratories in North Carolina participated in this study: Carolinas Medical Center, Duke University Health System, Laboratory Corporation of America® Holdings, and the University of North Carolina Health Care. Combined, these laboratories serve a large majority of the population in North Carolina and cover a wide geographical distribution within the state. Each lab searched their pathology database to generate a complete list of CIN2, CIN3, AIS, and invasive cervical cancer cases diagnosed in women ages 18 years and older from January 1, 2001 to June 1, 2006. After stratifying cases by diagnosis (i.e., CIN2, CIN3/AIS, and invasive cervical cancer), a random sample of 25 cases was selected within each stratum (∼300 samples in total). One block representative of the histology from each of the selected cases was retrieved for type-specific HPV DNA testing. Demographic information retrieved from the pathology records that was linked to the de-identified case number included histologic diagnosis (CIN2, CIN3, AIS, SCC, ADC, adenosquamous cell carcinoma, and other invasive cervical cancer), and other data when available, including age at diagnosis, race, and insurance coverage.

Histopathology review: Serial sections were cut from each block using precautions to prevent PCR contamination between cases, including single-use disposable microtome blades, cleaning the microtome between cases, and direct transfer of sections for PCR from microtome to sterile tubes using new single-use applicators (no contact with waterbath). The first and last sections were stained with Hematoxylin and Eosin (H&E). Intervening sections were transferred into 2 mL conical screw cap tubes with tether caps, one 10-micron section or two 5-micron sections per tube (Simport, Beloeil, Canada). A study pathologist (ERU) reviewed the H&E stained sections to confirm that histology was representative of the diagnosis. Samples that did not have representative material were not processed.

DNA isolation: DNA was extracted and purified by a modified protocol using the DNeasy kit (Qiagen, Valencia, CA) as described elsewhere [7]. Briefly, sections were heated for 20 min at 120°C in 180 µl ATL lysis buffer, then incubated with Proteinase K overnight at 65°C and further purified with spin columns. DNA was eluted in a final volume of 100 µL and tested immediately or stored at −20°C. For every batch of 28 samples, a water blank was processed through all steps of extraction to serve as a “contamination control”.

HPV Genotyping Tests: All DNA extracts were tested with LA and LiPA HPV typing assays, both based on L1 consensus PCR and type-specific hybridization. The LA (Roche Diagnostics, Indianapolis, IN) uses biotinylated PGMY09/11 and β-globin primer sets and detects 37 individual HPV types (6, 11, 16, 18, 26, 31, 33, 35, 39, 40, 42, 45, 51, XR(52), 53, 54, 55, 56, 58, 59, 61, 62, 64, 66, 67, 68, 69, 70, 71, 72, 73, 81, 82, 83, 84, 89, IS39) along with β-globin as an internal control. The manufacturer's protocol was followed with the exceptions of using 10 µL extract in the 100-µL PCR reaction and automated hybridization and washing of the reverse line blot with the Beeblot instrument (Bee Robotics, Caernarfon, UK). Samples positive for the XR(52) probe on the LA HPV strip that were also positive for HPV33, 35 and 58 were further evaluated to confirm or exclude the presence of HPV 52 using an HPV 52 quantitative PCR assay with a threshold of 50 viral copies [8]. Samples negative for both β-globin and HPV were considered inadequate for evaluation.

LiPA (Innogenetics, Gent, Belgium) uses SPF10 primers and detects 28 HPV types (6, 11, 16, 18, 26, 31, 33, 35, 39, 40, 43, 44, 45, 51, 52, 53, 54, 56, 58, 59, 66, 68, 69, 70, 71, 73, 74, 81, 82) as well as a generic probe to detect other HPV types (HPV X) and a genomic control probe. The PCR and amplicon hybridization to the HPV genotyping strip (AutoBlot 3000H, MedTec, Buffalo, IL) followed the manufacturer's protocols. The manufacturer's algorithm for interpretation of typing result based on pattern of positive probes was followed. This algorithm includes reporting unequivocal detection of types as well as “possible types”. The ambiguity results from certain hybridization patterns and affects only types 39, 52, 54, 68, 69, and 71. Samples negative for both the genomic control probe and HPV were considered inadequate for evaluation.

Concordance between LA and LiPA: For specimens with adequate LA and LiPA results, we calculated kappa to assess congruence of the type-specific results and performed McNemar's test to identify differences in the marginal frequencies by typing assay. Because of the large number of comparisons, we used a threshold of p<.01 to assess significance of the McNemar's test. We performed these analyses for each HPV type present in both tests. We defined HPV 16, 18, 31, 33, 35, 39, 45, 51, 52, 56, 58, 59, 66, and 68, as high risk (HR) HPV types, HPV 26, 53, 67, 69, 70, 73, 82, 85, and IS39 as “Possible HR”, and HPV 6, 11, 32, 40, 42, 43, 44, 54, 55, 61, 62, 64, 71, 72, 74, 81, 83, 84, 87, 89 as low risk (LR) HPV [9], [10]. Concordance evaluations were restricted to unequivocal positive LiPA results and repeated with “possible type” LiPA results interpreted as positive.

We evaluated two algorithms for combining the LA and LiPA results. These were 1) parallel algorithm, in which results of both tests were considered for all samples (positive result = positive for either test) and 2) serial algorithm, in which the LiPA assay results were only considered for those samples that were inadequate or negative for any HPV type by LA. We calculated the type-specific prevalence and overall HR or LR prevalence by each algorithm. We also calculated the mean and median number of HPV types detected per woman for each algorithm.

HPV Type Distribution: We examined HPV type distribution stratified by grade of cervical diagnosis. For evaluating vaccine-type HPV prevalence, we examined cases positive for HPV16 or 18 within each diagnosis stratum. We used Pearson's χ² test for independence to evaluate bivariate associations between vaccine-type HPV prevalence and race, stratified by diagnosis category. The Cochran-Armitage test was used to examine age trends of HPV 16/18 positivity within each diagnosis grade for 3 age groups (<30, 30–39, > = 40). We examined bivariate associations between single and multiple infections and 1) disease grade, and 2) age group using Pearson's χ² test. Two-sided statistical tests were considered significant at the alpha level of 0.05. Data were analyzed using SAS (v 9.1, SAS Institute, Cary, NC, 2002).

As expected, the parallel algorithm yielded the highest prevalence of any HPV type (95.7%) as well as HR HPV (93.1%) (Table 2). Applying the serial algorithm resulted in nearly the same HR-HPV prevalence as the parallel algorithm (92.1%). However, the prevalence of a few types appeared somewhat lower by application of the serial algorithm, particularly HPV 16 (from 52.7% to 50.2%), HPV 33 (from 7.6% to 5.8%), and HPV 52 (from 13.4% to 8.3%). The median number of HR types per woman (one) did not differ by testing algorithm. The mean number of types per woman was slightly higher using the parallel algorithm (1.31) than for the serial (1.14). Based on the similarity of results using the parallel and serial algorithms, and the efficiency of only performing two tests in a subset of samples in future genotyping projects, we report the remainder of our results according to the serial testing algorithm. Of the 278 valid results, 248 (89.2%) are from LA.

Of the 278 specimens with typing results, 32.7% (n = 91) were CIN2, 31.3% (n = 89) were CIN3, and 30.9% (n = 86) were SCC. AIS comprised <1% of specimens (n = 2), and ADC accounted for only 2.5% (n = 7). The remaining 5 specimens were from subjects who were diagnosed with cervical cancer other than SCC and ADC (classified as ‘other invasive’), for a total of 98 (35.3%) invasive lesions. Proportionally higher CIN2 lesions (14.3%) were typed by LiPA than CIN3/AIS (12.4%) or invasive lesions (6.1%), but the difference was not statistically significant (p = .17).

Among women with race information (n = 204) in all diagnosis grades, about half were non-Hispanic white, a third were non-Hispanic black, and 10% were Hispanic. Median age increased with increasing severity of squamous lesions, from 27 years in CIN2 to 33 in CIN3 to 40 in ICC (p<.0001). Median age for AIS was higher at 41 years and highest for ADC (48 years).

Table 3 presents HPV DNA prevalence by diagnosis category. No HPV DNA was detected in 11 specimens, 9 of which were from invasive cancers. HR HPV was detected in 96.6% of CIN2, 95.5% of CIN3/AIS specimens and 85.7% of specimens with invasive cancer (data not shown). HPV type 16 was the most common in all diagnosis categories (except 7 ADC) and increased from 45.6% in CIN2 to 52.8% in CIN3/AIS to 56.3% in SCC specimens.

Vaccine type HPV 16 or 18 was detected in 59.2% of invasive cancer specimens. As shown in Figure 1, HPV 45 was the second most common type among invasive cases (9.2%), slightly more than HPV 18 at 8.2%. Prevalence of other phylogenetically related HPV types detected among invasive cases was 5.4% for HPV 33, 4.1% for HPV 35, 3.1% for HPV 52 and 1.0% for HPV 58. LR HPV types 6, 11, 40, 42, 54, 83, 84, and 89 were detected in approximately 10% of CIN2 lesions, mostly as a co-infection with HR HPV types except in two CIN2 (HPV 6 and 42) and one CIN3 (HPV 62) specimens (data not shown). No LR HPV types were detected in SCC, ADC, or other invasive cervical specimens.

Overall, HPV 16 or 18 was detected in 56.5% of specimens. The proportion of HPV 16/18-related lesions appeared to increase with disease severity, although this trend was not statistically significant (Table 4). We found no statistical difference in the proportion of 16/18-related lesions by race or by age category, overall or stratified by disease severity as shown in Table 4.

Figure 1 displays HPV type distribution by single and multiple type infection. Among SCC and ADC cases, 94.4% had only a single HR HPV detected; single infections were less common in CIN3 and AIS cases (86.4%) and lowest in CIN2 specimens (65.6%) (p<.0001). Multiple type infection decreased with increasing age group (58% in females <30, 27.7% in females aged 31–49, and 14.9% in females > = 50 years; p<.0001).