Prevalence and type distribution of HPV in cervical precancers and invasive cancers is not well-described in the United States 
. Using archived FFPE biopsy tissue from a large source population of females with histologically-confirmed precancerous and invasive cervical lesions, we examined HPV type prevalence and evaluated laboratory methods for HPV DNA detection in FFPE biopsy tissue. Specifically, we used a combination of LA and LiPA HPV genotyping assays to maximize HPV DNA detection in archived biopsy specimens.
Several studies have reported high concordance between the LA and LiPA assays for detecting HPV DNA in cervical cells 
. However, assay performance on cytology specimens are not directly comparable to that for FFPE specimens given the potential for DNA degradation in the latter. Fewer studies have investigated the use of commercial assays for HPV genotyping in FFPE specimens. In a study from Australia, Tan et al. compared use of LA vs LiPA for HPV genotyping of a small sample of vulval biopsy specimens, and concluded that LiPA has higher sensitivity to detect DNA in archived biopsy tissue 
. A larger study by Wheeler et al. conducted in a U.S. population, however, found high (>95%) crude agreement between the 2 assays for most HPV DNA types in cervical FFPE tissue 
. Consistent with the U.S. study, our data indicate that using a serial testing algorithm whereby only HPV-negative and inadequate samples by LA are tested with LiPA results in very little change in prevalence estimates of HR-HPV for most types compared to testing all specimens with both tests in parallel. Using a serial algorithm, we demonstrate greatly improved efficiency without compromise to sensitivity in detection or bias in detection of HR HPV types. If only LA negative or inadequate samples had been tested by LiPA, 238 tests could have been avoided.
As expected, HPV16 was the most commonly detected HPV type in all diagnosis categories, followed by HPV 52, 18, 31, 33, and 45. These findings are similar to that found in non-Hispanic white and Hispanic women with CIN3/AIS and invasive cervical cancer in the U.S. 
, as well as in other international settings 
. We found that HPV16 was more likely to occur as a single infection in CIN3/AIS and invasive cancers compared to CIN2. This could be due to differential clearance of HR HPV types, but the results may be confounded by age which we were unable to control for given the small sample size of the study 
. HPV16 and 18 were less commonly detected in CIN2 and CIN3/AIS than ICC lesions, and distribution of HR types differed by lesion grade. Among invasive cases, HPV45 was the second most common type (10%); HPV18 was third highest (8%) followed by 33 (6%). We found less HR HPV positivity in invasive cases compared to lower grade lesions. The reasons for this are unclear and could reflect differential specimen quality. However, the proportion of invasive cases associated with HR HPV (85.7%, 95% CI: 78.8–92.6) is similar to that found in other studies. For example, Wheeler et al. reported a 90.0% HPV positivity in U.S. women with invasive cervical cancer compared to 97.1% HPV positivity in those with CIN3 or AIS 
Our study has several limitations. First, our population was restricted to adult female residents of North Carolina and may not be generalizable to other U.S. populations. However, we obtained specimens from 4 large pathology laboratories that serve a large majority of residents of North Carolina to minimize selection bias. Second, because the study was cross-sectional in nature, it was neither possible to investigate the effects of aging on HPV persistence or type distribution, nor to examine the influence of HPV types on disease progression. Third, we had limited statistical power to investigate HPV coinfection patterns by disease grade. Fourth, we did not perform additional testing to validate typing results discrepant between LA and LiPA. We assumed that results positive by either method were correct, but some might actually be falsely positive. Concordance between LA and LiPA may differ depending on method of sample collection, extraction method, or disease prevalence.
In summary, we used specimens available from a pilot study to develop an efficient approach to HPV testing of archived FFPE specimens. Currently, the serial algorithm is being applied to a variety of other population-based studies of HPV type distribution in cancer and precancer specimens (CDC, unpublished data) 
. Our data also provide important information about the distribution of individual HPV types in cervical disease prior to HPV vaccine introduction in North Carolina. The impact of HPV vaccines in reducing the burden of cervical and other HPV-associated cancers is of greatest concern, but it may take many years to demonstrate population-level reductions in these diseases. Cervical precancer lesions (CIN2/3) take less time to develop and are the accepted proxy for monitoring the impact of vaccine on cervical cancer. North Carolina has a large rural population at higher risk of cervical cancer, and an annual cervical cancer incidence of 8.2 cases/100,000 women 
. Thus, these baseline results may be a useful contribution to monitoring primary and secondary cervical cancer prevention efforts in North Carolina.