In regions of the world in which cervical cytologic screening is practiced routinely, the incidence of cervical carcinoma reportedly has declined dramatically over the past 40 years. However, the proportion of adenocarcinoma relative to squamous cell carcinoma continues to increase.1, 2 Data derived from the Surveillance, Epidemiology, and End Results (SEER) program's Public-Use database (a population-based database derived from registries that record all invasive malignancies for approximately 10% of the U.S. population) indicate that the 5-year, average age-adjusted incidence rates for cervical adenocarcinoma increased by 29.1% over 24 years (from 1973–1996, using SEER* Stat 2.0).2 Although this SEER study was to our knowledge among the first to demonstrate an increase in invasive adenocarcinoma cases relative to the population at risk, not only relative to squamous cell carcinoma, similar population-based data now have been reported from Canada, the United Kingdom, and western Europe.3–7 The increase in the incidence of adenocarcinoma has led to speculation that cytologic screening methods identify squamous precursor lesions more effectively compared with adenocarcinoma precursor lesions.2 Given these concerns, new data that provide insight into the likelihood of detecting adenocarcinoma in situ (AIS) and/or other glandular precursor lesions using standard cytologic techniques are both valuable and timely.
In this issue of Cancer Cytopathology, the sensitivity of detection of AIS by standard cervical cytologic screening was assessed.8 Cases and controls used in the study were obtained primarily from two central laboratories (the Western Australia Centre for Pathology and Medical Research [PathCentre] and Western Diagnostic Pathology [WDP]), which review the majority of cytologic smears collected from throughout Western Australia. A total of 123 patients with a pathologic diagnosis made by cervical or excisional biopsy of either AIS or AIS and a possible or definite high-grade epithelial abnormality (HGEA) (high-grade squamous intraepithelial lesion [HSIL]) who also had had ≥ 1 cytologic smears evaluated at the PathCentre or WDP within 36 months of the diagnosis were identified using the Western Australian Cervical Cytology Registry. All but 13 (5.9%) of 222 cytological smears so identified, including 63 slides initially accessioned at other laboratories, underwent secondary review. Although all pathology reports were reviewed, confidentiality guidelines did not permit the examination of pathologic specimens to confirm the diagnosis. The control group was comprised of 100 cases with Grade 3 cervical intraepithelial neoplasia (CIN 3). AIS cases were analyzed separately from AIS and any HGEA. The sensitivity, defined as the percentage of smears with an initial diagnosis of possible or definite high-grade disease (either glandular or squamous [HGEA]), was calculated independently for each laboratory. Because Australia has established guidelines for referral for colposcopy that include a cervical smear diagnosis of CIN 1, the sensitivity of detection was calculated with and without CIN 1. At the outset, reviewing cytopathologists were given the information that all smears undergoing review were obtained from women with a diagnosis of AIS, AIS + HGEA, or CIN 3 (control cases).
For the purposes of this discussion, the WDP results are given in parentheses next to the PathCentre results. For the PathCentre (WDP), the sensitivity of detecting any HGEA was 47.6% (54.3%) for AIS alone and 65.6% (48.9%) for AIS +HSIL. This compared favorably with the CIN 3 controls (42.5%). For a high-grade glandular lesion, the sensitivity was 47.6% (50.6%) for AIS alone and 37.5% (21.6%) for AIS + HSIL. When CIN 1 lesions were included, the sensitivity of detection increased to 64.3% (94.6%) for AIS and 90.5% (94.1%) for AIS + HSIL. The percentage of women with a cytologic diagnosis of possible or definite HGEA within 36 months of the biopsy confirmation of AIS was 64.3% (91.9%) for AIS and 80.9% (76.5%) for AIS +HSIL and in the case of a cytologic diagnosis of an abnormal glandular lesion, these figures were 64.3% (91.9%) for AIS and 91.9% (54.2%) for AIS +HSIL. Both the sampling error (the percentage of smears lacking evidence of possible or definite HGEA) and the screening error (the percentage of smears with an initial negative smear and possible or definite HGEA found on review) were low. That fewer precursor slides were evaluated at the PathCentre compared with the WDP may have accounted for the differences in sensitivity for the detection of AIS, but not in that for AIS + HSIL.
These results are in keeping with other published data with respect to screening and are encouraging because the screening efficacy for AIS was found to be comparable to that for CIN 3. Detection of all forms of glandular neoplasia including AIS has a sensitivity of 50–72%, which is not significantly different from the sensitivity of a single repeat test ultimately detecting CIN 2 and CIN 3 in women with atypical squamous cells detected on a previous smear (67–85%).9
The authors readily acknowledge limitations to their study. To their knowledge, the prevalence of AIS in the population screened is unknown. At the outset, the cytopathologists performing slide review knew that all the slides had been obtained from women who eventually were diagnosed with AIS or CIN 3. Sensitivities calculated with the end results known have the potential for significant bias. Including a control group of cases without significant neoplasia, and a slide review without knowledge of the final diagnosis, would have strengthened this study. Formal review of the cervical biopsy specimens was not permitted, and the number of women who underwent a definitive conization or hysterectomy after cervical biopsy was not reported. Because AIS can resemble other malignant or nonmalignant processes, excisional biopsy extending deep into the endocervical canal is necessary to confirm the diagnosis and to exclude the presence of invasive carcinoma.10 Cases with atypical glandular cells not characterized as high-grade lesions also were not included.11 Although it has been argued that atypical glandular precursors to AIS do not exist,12 the majority of recent studies indicate that 17–33% of women with these findings have significant glandular and squamous cell lesions or underlying endometrial carcinoma.13–16 The authors comment that including such cases is not likely to enhance the sensitivity of screening. Although this may be true, including these cases should enhance the specificity and negative predictive value for AIS and certainly those for invasive adenocarcinoma. Accurately determining the absence of disease (specificity) is as important as determining sensitivity. This opinion is emphasized by the authors' previous study, which found possible or definite HGEA in two of three cases initially interpreted as demonstrating nonspecific minor glandular changes. The authors also previously reported that sensitivity was improved when these cases were included, although the study design also was limited by the nonblinded nature of the slide review.17 Finally, cases diagnosed with invasive adenocarcinoma were not included. This is unfortunate, because there are nearly no data available to address the relation between AIS on cervical cytology and underlying adenocarcinoma.
The 2001 Bethesda system classifies glandular cell abnormalities that are less severe than malignancy into atypical glandular cells not otherwise specified (ACG), atypical glandular cells favor dysplasia, and AIS. This change was made based on the findings of CIN in 9–54% and underlying AIS or invasive disease in 0–8% of cases with a diagnosis of ACG “favor dysplasia.”18 The 2001 Consensus Guideline recommendations for such cases include colposcopy in the clinical setting of ACG “favor dysplasia,” endometrial biopsy for any woman with atypical endometrial cells or who is age > 35 years, and excisional biopsy (conization or loop electrosurgical excision procedure [LEEP]) if the colposcopic evaluation is negative. This approach was recommended because of the lack of sensitivity of colposcopy in detecting AIS or early adenocarcinoma.9 Because the risk for underlying disease in women with these abnormalities is uncertain, it is important to include all glandular precursor lesions in studies evaluating the sensitivity, specificity, and predictive value of any method used for cervical screening.
Although to our knowledge there has been no significant reduction in the incidence of adenocarcinoma in Australia and other countries in which screening is commonplace, the authors remain cautiously optimistic that routine cytologic screening for AIS eventually will reduce the incidence rate of cervical adenocarcinoma. In support of this theory, the authors referenced another important SEER study that proposes that the latency period between AIS and invasive adenocarcinoma and squamous cell carcinoma is 13 years compared with 18 years, and that adenocarcinoma incidence rates are increasing (by 1.42% per year).19 Unfortunately, SEER data can only indicate the differences in the average age of the patient at the time of detection of in situ versus invasive disease, which is different from longitudinal data that demonstrate the natural history of the disease. Data in SEER regarding in situ disease are known to be problematic. There is substantial variability in the reported incidence rates for carcinoma in situ between registries, a finding that is more likely related to differences in the interpretation of reports and to changes in terminology rather than to actual differences in incidence rates. Because of these concerns, the national database no longer includes in situ cases.19 Data derived using a more recent version of the SEER database (SEERStat* 4.0) indicate that the increase in age-adjusted incidence rates including all adenocarcinoma subtypes and adenosquamous histology per 100,000 women-years has continued (from 1.32 in 1973–1977 to 1.71 in 1993–1997 [unpublished data]). Rates also are increasing in younger women, declining in older women, declining in white women compared with black women, and increasing for localized and regional disease compared with distant disease.20 These data indicate that current screening methods have enhanced the detection of earlier stage disease, especially in younger women.
The study by Schoolland et al. in the current issue of Cancer Cytopathology is of particular relevance in that it appears to be the only report specifically evaluating the sensitivity of cytologic screening for AIS. The results are reassuring in that the sensitivity for AIS was comparable to that for high-grade squamous lesions. The diagnostic error rates (4.8% [7.4%] and 4.9% [9.95%] for AIS and AIS +HSIL, respectively) emphasize the importance of the careful evaluation of ACS “favor dysplasia” as adopted by the Consensus Conference on the management of cervical cytologic abnormalities.9 Although it is uncertain that screening for AIS will eventually reduce the incidence rate of adenocarcinoma, these data and current population-based studies indicate that screening has improved the detection of adenocarcinoma at earlier stages, which substantially impacts the potential for cure.