Cervical cancers also can arise from glandular cells lining the endocervical canal. Premalignant changes in these cells are reflected cytologically as atypical glandular cells (AGC) or adenocarcinoma in situ (AIS). In the general population, AGC is an uncommon cytologic diagnosis, accounting for about 0.3% of Pap tests (9). However, AGC is associated with a high risk of preinvasive and invasive disease; paradoxically, most of these are squamous, including a 9% risk of low grade lesions, an 11% risk of high grade CIN, a 3% risk of histologic AIS, a 1% risk of endometrial hyperplasia, and a 5% risk of cancer (9). Current and past guidelines recommend colposcopy for all women with AGC and AIS, with further investigation dependent on colposcopic findings and the severity of the cytologic abnormality (10, 11).

Cervical adenocarcinoma and adenocarcinoma in situ have been linked to infection with the human papillomavirus (HPV) and like squamous disease should be more common among immunosuppressed women (12). Unfortunately, rates of cytologic AGC and AIS and the associated risk for glandular neoplasia have not been well studied among women with HIV. In part, this is because glandular lesions are less common than squamous lesions, and accruing sufficient numbers of HIV-seropositive women in screening programs across time has been challenging. Lehtovirta and colleagues found that atypical glandular cells (AGC) were found in 4% of HIV-seropositive women undergoing cytologic screening, but they studied only 108 women and did not describe histologic findings (13).

Every six months, participants had structured medical, social, and sexual histories taken by trained interviewers, followed by a physical examination that included gynecologic evaluation and Pap testing. According to study-wide protocol, single-slide Pap samples were obtained using a spatula and brush. HIV status was established by Western blot, and women who seroconverted during follow-up were classified according to visit-specific serostatus. Pap tests were interpreted centrally at Dianon (New York, NY, formerly Kyto or Kyto Meridien) according to the 1991 Bethesda system for classification of cervicovaginal cytology (16), updated in 2001. Glandular abnormalities were classified as atypical glandular cells, subclassified as not otherwise specified or favor neoplastic, as adenocarcinoma in situ, and as adenocarcinoma. All Pap smears were screened by two cytotechnologists blinded to HIV status, with 10% of all negative smears and all abnormal smears reviewed by a cytopathologist. Study protocol recommended referral for colposcopy for glandular abnormalities of any grade, though decisions on biopsy and treatment were individualized. At some sites, women were referred externally for colposcopy, especially early in the study, and women may have had Pap tests outside study centers that led to colposcopy. Although women were asked whether they had cervical biopsies externally in the six months prior to each visit, with affirmative answers leading to medical records retrieval, but we could not confirm negative responses. Histology results were interpreted locally and were not centrally reviewed.

Of the 3766 women in WIHS (2791 HIV+, 975 HIV−), 3522 women (2593 HIV+, 929 HIV−) were excluded because they did not have an index smear showing glandular abnormalities during the study period. One HIV seropositive woman was excluded because she had undergone hysterectomy before her index AGC pap. Another 95 women (78 HIV+, 17 HIV−) were excluded because they did not have colposcopy done within 6 month of index abnormal Pap. Finally, 29 women (25 HIV+, 4 HIV−) were excluded because they had no histologic results, leaving 119 women (94 HIV+, 25 HIV−) for analysis.

Histology and cytology results were tabulated. Differences in demographic and clinical characteristics between HIV seropositive and seronegative women were compared using Fisher’s exact test. Multivariate logistic regression models that incorporated Generalized Estimating Equation (GEE) models were used to compare the proportions of patient-visits in HIV seropositive and HIV seronegative group with glandular abnormality, adjusting for repeated measures across visits for the same women over time (17). The variables adjusted for were age (divided into 4 groups: <30, 30–40, 40–50, >50), ethnicity, smoking status, and whether the patient had been sexually active in the last six months. To further investigate the association of HIV serostatus and glandular abnormality, the model was refitted with an HIV serostatus variable, categorized as HIV seronegative, HIV seropositive/CD4 lymphocyte count <250/cmm, HIV seropositive/CD4 250–500, HIV seropositive/CD4 >500. A test for trend was applied: we parameterized HIV-serostatus and CD4+ lymphocyte count as a continuous variable with four levels for HIV seronegative women and HIV seropositive women with CD4 counts >500, 250–500, and <500/cmm and determined the significance of this variable in GEE models. The Wald test was used to determine the significance of all variables in these models (18).

Glandular abnormalities were found in 341 (0.7%, 95% CI, 0.6%–0.8%) smears from 244 (6%) women, including 264 (0.8%, 95% CI, 0.6%–0.9%) of 35057 smears from HIV seropositive women and 70 (0.6% (95% CI, 0.4%–0.8%) of 11,769 smears from seronegative women (P = 0.12 after adjustment for age, ethnicity, smoking, and sexual activity in the prior six months). The demographic characteristics of women with glandular abnormalities on Pap are presented in Table 1. The median age was 37 years among 198 HIV seropositive women and 39 years among 46 HIV seronegative women (P = 0.9). Compared to HIV seronegative women, seropositive women with glandular abnormalities were marginally less likely to be smokers and less likely to have multiple recent sexual partners.

Stratification by CD4 lymphocyte count showed the impact of immunosuppression on risk for glandular cytologic abnormality: glandular abnormalities were found in 93 (1.0%; 95% CI, 0.8%–1.3) of 9564 Pap tests among HIV-seropositive women with CD4 lymphocyte counts <250/cmm, 103 (0.8%; 95% CI, 0.6% –1.0%) of 13023 Paps among those with counts 250–500/cmm, 68 (0.6%; 95% CI, 0.4%–0.8%) of 12470 Paps among women with counts >500/cmm, and 70 (0.6%; 95% CI, 0.4%–0.8 of 11769 Paps among HIV-seronegative women (P for trend = 0.006). Seven women with glandular abnormalities did not have CD4 counts at the corresponding visit.

Despite a study-wide protocol recommending referral for colposcopy and despite intensive infrastructure at each site to promote compliance, colposcopy was documented to have been done within six months for only 148 (61%) of 244 index Pap tests with glandular abnormalities, 119 (60%) of 198 HIV seropositive and 29 (63%) of 46 HIV seronegative women. Endocervical curettings were obtained after 106 (43%) women with index abnormal Paps, including 81 (41%) HIV seropositive women and 25 (54%) seronegative women. Cervical punch biopsies were done for 76 (38%) index Paps, 60 (30%) in HIV seropositive women and 16 (35%) in seronegative women. Endometrial biopsies were done in only 19 (8%) index Pap cases, 13 in HIV seropositive and 6 in seronegative women. Excisional biopsies, including knife conization and loop excision, were done in 14 women. Overall, histologic findings were available for 119 (49%) of the 244 women within six months of an index Pap showing glandular abnormalities. Compliance improved as women were followed, as 212 (87%) of women with index glandular abnormalities had colposcopy and 192 (79%) had at least one biopsy at some subsequent point during the study.

In our study, the rate of biopsy-proven cervical cancer precursors among HIV seropositive women with glandular cytologic abnormalities was substantial. Among those with HIV, more than half had CIN or cancer, although only 21% had high grade disease. These results are consistent with rates reported in HIV seronegative women (9), and we did not find a difference in risk for high grade disease between HIV seropositive and seronegative women once glandular abnormalities were reported on Pap. While the 3% risk of cancer that we found among HIV seropositive women is small in absolute terms, it does reflect a substantial. Although WIHS is the largest U.S. cohort study of women with HIV, we had too few events to stratify histologic results by CD4 count; studies from other national cohorts may help to determine whether the most severely immunosuppressed women have an increased risk of biopsy-proven cervical disease after glandular Pap abnormalities.

For now, our results suggest that women with HIV and glandular abnormalities on Pap testing should be managed according to guidelines for the general population from the American Society for Colposcopy and Cervical Pathology (ASCCP)(10). These recommend colposcopy with endocervical curettage for all women, with endometrial assessment for those over age 35 years or with risk factors for endometrial cancer such as obesity or irregular vaginal bleeding. Risks for cervical disease and cancer are too high to justify either triage using HPV DNA testing without colposcopy or surveillance with serial cytology, as can be done for women with atypical squamous cells of uncertain significance (ASC-US). One exception to the management of glandular abnormalities according to ASCCP guidelines involves testing for high risk HPV types at initial colposcopy to minimize follow-up for women with negative initial evaluation. This strategy has not been tested in women with HIV, whose high prevalence of high risk HPV may limit the specificity of this approach.