The category “atypical glandular cells” (AGC) in the 2001 Bethesda System1 is an uncommon (approximately 0.5%) cervical cytology interpretation that occurs in routine cervical cancer screening. As noted in the Bethesda System at the time, “The finding of AGC is important clinically because the percentage of cases associated with underlying high-grade disease is higher than for ASC-US [atypical squamous cells of undetermined significance].
The current recommended management for AGC cytology by the American Society for Colposcopy and Cervical Pathology is immediate colposcopic evaluation with endocervical sampling for all subcategories of AGC cytology.2 In addition, endometrial sampling is recommended for all women aged 35 years or older and for women younger than 35 years with clinical indications of endometrial disease. Testing for high-risk human papillomavirus (HPV) is preferred at the time of colposcopy if it has not already been performed.
Studies have found AGC cytology to be strongly associated high-grade cervical and endometrial disease and cancer.3–8 However, most were relatively small in sample size, with most having approximately 100-200 women with AGC cytology3–8 and the largest having 662 cases,9 but did not include high-risk HPV testing. A recent literature review suggested that HPV DNA testing may be useful in differentiated the risk of cervical and endometrial cancer,10 based on a total of 661 women with AGC cytology and HPV DNA testing.
Given the differences in the natural history of cervical cancer11 and endometrial cancer,12 including the peak age of incidence, we hypothesized that both age and HPV status may be useful in differentiating the risks of the two cancers in women with AGC cytology. Specifically, we sought to quantify the age-specific and reproductive organ-specific cancer risk after an AGC cytologic interpretation in large clinic-based sample in which routine high-risk HPV testing is conducted.
Since 2003, Kaiser Permanente Northern California has implemented cotesting using high-risk HPV testing (Hybrid Capture 2; Qiagen Corporation, Gaithersburg, MD) and conventional cervical cytology in women aged 30 years or older for routine screening. Colposcopic evaluation was recommended for all women with AGC cytology, irrespective of the high-risk HPV results. Of 2,596 cytologic interpretations of AGC during that time period, 1,422 (54.8%) had retrievable high-risk HPV testing and histology results and were included in this analysis. Analysis of these data were approved by the Kaiser Permanente Northern California institutional review board and deemed exempt from human subjects review by the Office of Human Subjects Research of the National Institutes of Health.
Of those included in this cross-sectional analysis, 46.8%, 44.6%, and 8.7% were called AGC unqualified, AGC endocervical cells, and AGC endometrial cells, respectively. By comparison, of those excluded from the analysis because of missing histopathology or high-risk HPV testing results, 51.5%, 38.7%, and 9.9% of the AGC Pap results were called AGC unqualified, AGC endocervical cells, and AGC endometrial cells, respectively. The distributions of AGC unqualified, AGC endocervical cells, and AGC endometrial cells in the populations included and excluded from this analysis were statistically (P=.01) but not qualitatively different.
AGC subtype designations of endocervical cells or endometrial cells predicted cervical precancer/cancer or endometrial cancer, respectively (Table 1). However, neither did so perfectly, supporting the current recommendation that all initial workups be performed in the all-inclusive way for all AGC subtypes.2 Because almost half of AGC were not designated as either endocervical or endometrial (AGC not otherwise specified), and the reproducibility of this distinction is unknown, we considered them as one entity for assessing cancer risk. Notably, the reproducibility of AGC as a cytologic category is poor.13 The results of this analysis were similar when restricted to those women with AGC cytology unqualified. There were only minor differences in the distribution of histopathology in those women with either AGC endocervical or endometrial (combined) compared with those women with AGC unqualified (P=.02), suggesting that AGC not otherwise specified is an admixture of two AGC subtypes.
Cervical biopsy reports referable to abnormal screening tests in facilities served by the Kaiser Permanente Northern California Regional Laboratory were manually reviewed and entered into the database by Regional Laboratory personnel under the direction of one of the authors (B.F.). A random sample of 9.9% of all of the biopsy reports (for all cytologic interpretations and ages, not just for AGC) was subsequently rereviewed by the investigators to assess the accuracy of the data entry. Using cervical intraepithelial neoplasia grade 2 or more severe (CIN 2 or worse) compared with less than CIN 2 or worse as the cut point, classifications of abnormal histology in the database were found worthy of revision in 1.6% of cases, mostly in relation to CIN 1-2, which was reclassified as less than CIN 2 rather than CIN 2 or worse. Cervical histologic diagnoses also were categorized as CIN 2, CIN 2 or worse (ie, CIN 2, CIN 3, cervical adenocarcinoma in situ, cervical adenocarcinoma, cervical adenosquamous carcinoma, and cervical squamous cell carcinoma), CIN 3 or worse (excluding CIN 2), or cervical cancer.
Absolute risk and 95% confidence intervals (CIs) of different endpoint definitions were calculated. Two-sided Fisher exact tests were used to test for statistical differences between groups of women and for differences in risk for a given endpoint. Multinomial (polytomous) logistic regression14 was used to estimate odds ratios (ORs) and 95% CIs as a measure of association of age and high-risk HPV with two possible outcomes, cervical cancer or endometrial cancer (with a reference group of no cancer).
The age and racial/ethnic distributions of this subpopulation of women who had an AGC cytology in routine screening are shown in Table 2. The mean age, median age, and range of ages were 46.9, 46.0, and 30-87 years, respectively.
A total of 238 women with AGC cytology (16.7%, 95% CI 14.8-18.8%) were diagnosed with clinically significant disease: CIN 2 or worse, endometrial cancer, or other cancers (Table 3). This included 86 women (6.0%, 95% CI 4.9-7.4%) with any cancer and another 89 women (6.3%, 95% CI 5.1-7.6%) with CIN 3 or adenocarcinoma in situ. The risk of glandular disease (adenocarcinoma in situ, adenocarcinoma, adenosquamous carcinoma, or endometrial cancer) was 7.4% (95% CI 6.1-8.9%).
Approximately one quarter of women (24.3%) with AGC tested high-risk HPV positive; women diagnosed with endometrial cancer (5.3%) and women diagnosed with less than CIN 1 histology (11.1%) were the least likely to test high-risk HPV positive; women with CIN 3 histology (92.2%) and those with adenocarcinoma/adenosquamous carcinomas (90.0%) were the most likely to test high-risk HPV positive.
High-risk HPV test results and whether a woman aged 50 years or older compared with younger than 50 years (as a proxy for menopause) distinguished the risk for cervical precancer (CIN 2 or worse or CIN 3 or worse), cervical cancer, and endometrial cancer (Fig. 1). In women younger than 50 years (n=925), high-risk HPV-positive women had a 45.4% (95% CI 39.3-51.5%) risk of CIN 2 or worse, 32.7% (95% CI 27.1-38.7%) risk of CIN 3 or worse, and 6.3% (95% CI 3.7-9.9%) risk of cervical cancer. By comparison, high-risk HPV-negative women had only a 2.9% (95% CI 1.8-4.5%) risk of CIN 2 or worse (P<.001 compared with high-risk HPV-positive women). The risk of endometrial cancer was similarly low for high-risk HPV-positive (1.1%, 95% CI 0.2-3.2%) and high-risk HPV-negative women (1.5%, 95% CI 0.7-2.8%) (P=.8) younger than 50 years.
In women aged 50 years or older (n=497), high-risk HPV-positive women had 33.8% (95% CI 23.4-45.4%) risk of CIN 2 or worse, 22.1% (95% CI 13.4-33.0%) risk of CIN 3 or worse, and 10.4% (95% CI 4.6-19.4%) risk of cervical cancer. High-risk HPV-negative women had only a 0.5% (95% CI 0.1-1.7%) risk of CIN 2 or worse (P<.001 compared with high-risk HPV-positive women). The risk of endometrial cancer was much lower for high-risk HPV-positive women (0.0%, 95% CI 0.0-4.7%) compared with high-risk HPV-negative women (10.5%, 95% CI 7.7-13.8%) (P<.001).
We modeled the association of age and HPV test results with either cervical cancer (n=39; median age 47 years, mean age 49 years, and range 30-79 years) or endometrial cancer (n=92; median age=60 years, mean age 62 years, and range=34-88 years) outcomes (compared with not having either). For cervical cancer, being aged 50 years or older (OR 1.75, 95% CI 0.782-3.91) was not associated and testing high-risk HPV positive (OR 22.0, 95% CI 7.52-64.6) was positively associated with having cervical cancer. For endometrial cancer, being aged 50 years or older (OR 6.07, 95% CI 3.23-11.4) was positively associated and testing high-risk HPV positive (OR 0.247, 95% CI 0.0758-0.802) was negatively associated with having endometrial cancer.
We found that AGC cytology is a high-risk marker for cervical precancer, cervical cancer, and endometrial cancer. Perhaps not surprisingly, a greater proportion of cervical cancers associated with AGC cytology were adenocarcinoma/adenosquamous histology (10 of 29, 34%) than the expected proportion of approximately 20%15 (P=.03, two-sided binomial). Still, the majority of not only precancer but invasive cancer diagnosed as a result of an AGC cytology result is squamous and not glandular histology. Although a recent systematic analysis found similar risks of cervical precancer, cancer, and endometrial cancer associated with AGC cytology,8 the combined sample size was much smaller than the sample size in this analysis and did not provide age stratification. It was particularly noteworthy in this study that among women aged 50 or older, there was a 10% risk of endometrial cancer among HPV-negative women and 10% risk of cervical cancer among HPV-positive women.
The strengths and weaknesses of this study are both consequent to the difficulty in capturing data completely in any clinical practice evaluation. For example, 23.7% of women with AGC cytology had missing histologic diagnoses, which was a slightly lower percentage of women with AGC cytology missing histologic diagnosis than that in a similar clinical database analysis of AGC cytology by Zhao et al9; an additional 21.5% of women were missing high-risk HPV testing results.
Moreover, members with abnormal Pap results may choose to have their colposcopy at facilities not served by the Regional Laboratory, or with gynecologic providers outside of the Kaiser system (survey data indicate that approximately 1% of our Kaiser Permanente Northern California female members reported Pap tests outside of the Kaiser system in the preceding 3 years despite Kaiser membership). Members or their providers may be noncompliant with recommendations for colposcopy, or members may change insurance or lose coverage after an abnormal Pap result and before colposcopy. All of these factors decrease the number of cases available for evaluation, but none of them are recognized to be associated with increased risk of adverse histologic diagnoses, and as a consequence, those women for whom complete evaluation is available are thought to be a representative sample of the greater population.
In summary, the similarity in histology between atypical endometrial cells, atypical endocervical cells, and AGC not otherwise specified reemphasizes that all AGC should be reported primarily as AGC, regardless of whether further classification is offered, and evaluated in the same fashion. With the large clinical database from Kaiser Permanente Northern California, we were able to better quantify the cancer risk for the cervix and endometrium in women with AGC cytology. Although high-risk HPV testing is not currently recommended for triage of women with AGC to colposcopy,2 its use in combination with age may be used to distinguish between cervical and endometrial cancer risk and may thereby guide clinical follow-up and management. That is, everyone with AGC cytology needs diagnostic evaluation, but the focus of the diagnostic workup can be guided by HPV status. Especially in women aged 50 years or older with HPV-negative AGC cytology and a negative workup including an endometrial biopsy, clinicians should be aware that if malignancy has escaped detection, it lies in the endometrium, rather than the endocervix, and that additional evaluation, if undertaken, should be directed accordingly. In addition, for women aged older than 50 years with AGC who are diagnosed with adenocarcinoma where the primary site is uncertain (eg, the endometrial biopsy and the endocervical curettage are both positive), high-risk HPV testing can provide guidance to the clinician regarding probable site of origin, thereby guiding treatment decisions. Evaluation for women with HPV-positive AGC cytology should include multiple cervical biopsies, given the well-recognized limitations of colposcopy16,17 for detection of glandular dysplasia and cancer, and endocervical sampling or endocervical curettage because there is a greater likelihood of glandular involvement. We suggest that the future American Society for Colposcopy and Cervical Pathology guidelines should recommend, instead of just preferring, HPV DNA testing to be included with colposcopy, endocervical sampling, and endometrial sampling to assist in the proper management of AGC cytology in women.
1. Solomon D, Davey D, Kurman R, Moriarty A, O’Connor D, Prey M, et al. The 2001 Bethesda System: terminology for reporting results of cervical cytology. JAMA 2002;287:2114–9.
2. Wright TC Jr, Massad LS, Dunton CJ, Spitzer M, Wilkinson EJ, Solomon D. 2006 consensus guidelines for the management of women with abnormal cervical screening tests. J Low Genit Tract Dis 2007;11:201–22.
3. Westin MC, Derchain SF, Rabelo-Santos SH, Angelo-Andrade LA, Sarian LO, Oliveira E, et al. Atypical glandular cells and adenocarcinoma in situ according to the Bethesda 2001 classification: cytohistological correlation and clinical implications. Eur J Obstet Gynecol Reprod Biol 2008;139:79–85.
4. Diaz-Montes TP, Farinola MA, Zahurak ML, Bristow RE, Rosenthal DL. Clinical utility of atypical glandular cells (AGC) classification: cytohistologic comparison and relationship to HPV results. Gynecol Oncol 2007;104:366–71.
5. Kumar N, Bongiovanni M, Molliet MJ, Pelte MF, Egger JF, Pache JC. Reclassification and analysis of clinical significance of atypical glandular cells on ThinPrep using Bethesda 2001: Geneva experience. Swiss Med Wkly 2007;137:635–41.
6. Lai CR, Hsu CY, Tsay SH, Li AF. Clinical significance of atypical glandular cells by the 2001 Bethesda System in cytohistologic correlation. Acta Cytol 2008;52:563–7.
7. Irvin W, Evans SR, Andersen W, Jazaeri A, Taylor P, Stoler M, et al. The utility of HPV DNA triage in the management of cytological AGC. Am J Obstet Gynecol 2005;193:559–65.
8. Schnatz PF, Sharpless KE, O’Sullivan DM. Use of human papillomavirus testing in the management of atypical glandular cells. J Low Genit Tract Dis 2009;13:94–101.
9. Zhao C, Florea A, Onisko A, Austin RM. Histologic follow-up results in 662 patients with Pap test findings of atypical glandular cells: results from a large academic women’s hospital laboratory employing sensitive screening methods. Gynecol Oncol 2009;114:383–9.
10. Sharpless KE, O’Sullivan DM, Schnatz PF. The utility of human papillomavirus testing in the management of atypical glandular cells on cytology. J Low Genit Tract Dis 2009;13:72–8.
11. Schiffman M, Castle PE, Jeronimo J, Rodriguez AC, Wacholder S. Human papillomavirus and cervical cancer. Lancet 2007;370:890–907.
12. Cook LS, Weiss NS, Doherty JA, Chen C. Endometrial cancer. In: Schottenfeld D, Fraumeni JF Jr, editors. Cancer epidemiology and prevention. 3rd ed. New York (NY): Oxford; 2006. p. 1027–43.
13. Confortini M, Di Bonito L, Carozzi F, Ghiringhello B, Montanari G, Parisio F, et al. Interlaboratory reproducibility of atypical glandular cells of undetermined significance: a national survey. Cytopathology 2006;17:353–60.
14. Long JS. Nominal outcomes: multinomial logit and related models. In: Regression models for categorical and limited dependent variables. Thousand Oaks (CA): Sage Publications; 1997. p. 148–86.
15. Wang SS, Sherman ME, Hildesheim A, Lacey JV Jr, Devesa S. Cervical adenocarcinoma and squamous cell carcinoma incidence trends among white women and black women in the United States for 1976-2000. Cancer 2004;100:1035–44.
16. Guido R, Schiffman M, Solomon D, Burke L, ASCUS LSIL Triage Study (ALTS) Group. Postcolposcopy management strategies for women referred with low-grade squamous intraepithelial lesions or human papillomavirus DNA-positive atypical squamous cells of undetermined significance: a two-year prospective study. Am J Obstet Gynecol 2003;188:1401–5.
17. Gage JC, Hanson VW, Abbey K, Dippery S, Gardner S, Kubota J, et al. Number of cervical biopsies and sensitivity of colposcopy. Obstet Gynecol 2006;108:264–72.