Atypical glandular cells of undetermined significance in conventional cervical/vaginal smears and thin-layer preparations

A follow-up comparison study




Biopsy follow-up of a cervical/vaginal smear interpretation of atypical glandular cells of undetermined significance (AGUS) most often reveals either a benign reactive process or a squamous cervical intraepithelial neoplasia (CIN) rather than a glandular one. The ThinPrep® Papanicolaou test (TP) has been shown to increase diagnostic sensitivity for CIN. To the authors' knowledge there are few studies examining its effectiveness in diagnosing uterine glandular lesions, either endocervical or endometrial. The authors compared outcomes after AGUS interpreted in TP specimens and conventional smear preparations (CPs).


Follow-up was sought in all cases that were interpreted as AGUS during a 3-year period in which practice groups converted from CPs to TPs. A tissue diagnosis of adenocarcinoma in situ, CIN of Grade 2 or 3, or invasive carcinoma was considered to be a positive follow-up result. Either a benign biopsy finding with a subsequent benign cytologic result or two consecutive benign cytologic results were considered to be negative follow-up results.


AGUS was reported in 116 of 75,002 TPs (0.15%) and 151 of 79,322 CPs (0.19%). Follow-up information was available in 91 (78%) and 130 (86%) cases, respectively. The predictive value of AGUS for a positive follow-up result was 22% in the TP group and 15% in the CP group. An AGUS interpretation with a positive follow-up result was found in 0.027%of all cases using TPs and in 0.025% of all cases using CPs.


The use of TPs may increase the diagnostic specificity of AGUS for a high-grade precursor lesion or carcinoma. Cancer (Cancer Cytopathol) 2002;96:1–4. © 2002 American Cancer Society. DOI 10.1002/cncr.10312

In the Bethesda system for reporting cervical/vaginal cytologic diagnoses,1 atypical cells perceived as glandular that do not fulfill the diagnostic criteria for invasive adenocarcinoma are designated “atypical glandular cells of undetermined significance” (AGUS). This interpretation may be qualified further with regard to whether the cells are of endocervical or endometrial origin and, in the former case, whether a reactive or neoplastic process is favored. AGUS has been a problematic interpretation clinically because biopsy follow-up mostly has shown either a benign reactive process or a squamous precursor lesion (cervical intraepithelial neoplasia [CIN]) rather than a glandular one.2–16

The ThinPrep® Papanicolaou test (TP) (Cytyc Corporation, Boxborough, MA) is a slide preparation method that utilizes liquid fixation of the exfoliated cells. It has been shown to produce well preserved thin-layer slides devoid of the artifacts that may contribute to inaccuracies in the interpretation of conventional smears.17, 18 TPs have been shown to increase sensitivity for CIN,19 but to our knowledge there have been relatively few studies of its effectiveness in diagnosing uterine endocervical or endometrial glandular lesions.20–23 Thus, we decided to evaluate our laboratory's experience with AGUS in TP specimens.


The ThinPrep 2000 processor was introduced into the Brigham and Women's Hospital cytology laboratory in July 1997. The laboratory serves 97 different practice groups including both high-risk and low-risk populations. Many of these practices underwent a transition to TPs at different times and rates over the next 3 years, the interval chosen as our study period. To ensure as best we could comparable populations in the TP and conventional preparation (CP) groups, we included cases only from practice groups that had submitted > 100 cases during the 3-year study period of which > 10% were TPs; this included 37 of the 97 groups. During the 6 months after the study period, these 37 groups submitted 87% of samples as TPs.

All cases interpreted as AGUS during this 3-year period were identified and categorized according to TP or CP, and AGUS subtype (endometrial or endocervical, favor reactive, favor neoplastic, or not specified). Cases with additional interpretations such as atypical squamous cells of undetermined significance (ASCUS) and low-grade or high-grade squamous intraepithelial lesion (LSIL, HSIL) were excluded. If more than one specimen from the same patient was interpreted as AGUS, only the initial specimen was included. Records then were searched for follow-up cytologic and/or tissue sampling. A tissue diagnosis of adenocarcinoma in situ (AIS), CIN of Grade 2 or 3, or invasive carcinoma within 1 year of the AGUS interpretation was considered to be a positive follow-up result. Benign tissue diagnoses with subsequent (> 1 year) benign cytologic or biopsy follow-up were considered to be negative follow-up results. If no biopsy was performed, two consecutive benign cytologic results were considered to be a negative follow-up result.

AGUS frequencies as well as those for ASCUS, LSIL, and HSIL for the two groups were tabulated. The predictive values of the AGUS and endocervical AGUS subcategories for a positive follow-up (positive predictive value [PPV]) and for glandular neoplasia were computed. Possible differences in the PPVs between TPs and CPs were assessed by a two-sided (tailed) Fisher exact test. A P value < 0.05 was considered statistically significant.


During the 3-year study period, 75,002 TPs and 79,322 CPs were received from the designated practice groups. AGUS was reported in 116 of the TPs (0.15%) and 151 of the CPs (0.19%). The percentages of other abnormal interpretations during the 3–year period for TP and CP were: ASCUS, 6.0% and 5.1%, respectively; LSIL, 2.0% and 1.1%, respectively; and HSIL, 0.77% and 0.41%, respectively (Table 1). The frequency of each AGUS subtype interpreted for each preparation is shown in Table 2. Follow-up sampling results were available for 91 TPs (78%) and 130 CPs (86%). The difference in follow-up percentages may reflect the shorter follow-up interval in the TP group because these samples generally were collected later in the study period than the CP samples. There were 20 positive follow-up diagnoses in both the TP and CP categories, and 14 of these diagnoses in both categories were determined to be glandular neoplasia (Table 3). The PPV of AGUS for TPs was 22% and was 15% for CPs (P = 0.23); for glandular neoplasia, the PPVs were 15% and 11%, respectively (P = 0.31). The PPVs for TPs of endocervical AGUS “favor neoplastic,” “unspecified,” and “favor reactive” were 67%, 28%, and 10%, respectively; the corresponding PPVs for CPs were 77%, 23%, and 4%, respectively. In the 3-year study period, the percentage of all cases with both an AGUS interpretation and a positive follow-up result was 0.027% for TPs and 0.025% for CPs.

Table 1. Percentage of Cases in Each Diagnostic Category in Thin-Layer and Conventional Preparations over a 3-Year Period
 TP% (n = 75,002)CP% (n = 79,322)
  1. TP: thin-layer preparation; CP: conventional preparations; ASCUS: atypical squamous cells of undetermined significance; LSIL: low-grade squamous intraepithelial lesion; HSIL: high-grade squamous intraepithelial lesion; AGUS: atypical glandular cells of undetermined significance.

Table 2. Percentage of AGUS Subtypes Diagnosed in Thin-Layer and Conventional Preparations
AGUS subtypeTP (n = 116)CP (n = 151)
  1. AGUS: atypical glandular cells of undetermined significance; TP: thin-layer preparation; CP: conventional preparations.

Not specified2436
Table 3. Histologic and/or Cytologic Follow-Up of AGUS (Number of Cases in Each Follow-Up Category Compared with Original AGUS Subtype)
AGUS subtypeThin-layer preparations (91 cases)TotalConventional preparations (130 cases)Total
  1. AGUS: atypical glandular cells of undetermined significance; AIS: adenocarcinoma in situ; ACA Cx: adenocarcinoma of the cervix; ACA Em: adenocarcinoma of the endometrium; CIN: high-grade squamous cervical intraepithelial neoplasia; RCT: benign or reactive diagnoses.

Not specified3013152230443647


The results of this comparison study of AGUS in TPs and CPs demonstrated an 18% decrease in AGUS interpretations using TPs. There was a concomitant increase in the PPV of AGUS in TPs, although it did not reach the level of statistical significance. Although AGUS interpretations were less frequent in the TP group, cases interpreted as squamous lesions (ASCUS, LSIL, or HSIL) in TPs increased by 17%, 87%, and 88%, respectively (Table 1). Increases in the frequency of interpretations of squamous lesions using TPs may have reflected either a higher disease prevalence in the TP group, an increased sensitivity or decreased specificity of TP for squamous lesions, or all three possibilities. Because of our selection process, we assume, but cannot be certain, that disease prevalence in the CP and TP groups was similar. Prior studies of the ThinPrep 2000 processor, summarized by Austin et al.,19 have consistently shown an increased TP sensitivity for squamous precursor lesions compared with CP. Thus, although biopsy follow-up data for the reported interpretations of squamous lesions in the two groups was not obtained, we also assume that an increased TP sensitivity in the current study was largely responsible for the higher proportions of squamous abnormalities reported in the TPs. On the basis of these assumptions, the lower AGUS frequency in the TP group implies either an increase in the TP specificity for AGUS or a decrease in the TP sensitivity for lesions that generally are interpreted as AGUS in CPs. However, it is unlikely that the latter can explain completely the decreased frequency of AGUS in TPs, because the proportion of patients from the entire screened population who ultimately were found to have carcinoma or a high-grade precursor lesion after a diagnosis of AGUS actually was slightly higher in the TP group. Thus, although the higher PPVs for AGUS in TPs was not statistically different from that of the CPs, the observed slight increase in the PPV of TP, coupled with the decreased frequency of AGUS interpretations, is most compatible with an increase in AGUS specificity of TPs for a high-grade precursor lesion or carcinoma of either squamous or glandular type.

The PPVs for the endocervical AGUS subcategories demonstrated a marked difference between the “favor neoplasia” designation and either the “unspecified” or “favor reactive” designations in both types of preparations (67% vs. 28% vs. 10% in TPs and 77% vs. 23% vs.4% in CPs). These differences are similar to those reported in other studies of biopsy follow-up results after an AGUS interpretation in CPs,2–16 and in one similar study using TPs,20 and demonstrate that differences in PPVs among endocervical AGUS subcategories are similar in both TPs and CPs. Although the number of cases was small, we found that the AGUS “endometrial” category had a very low PPV in both groups.

Laboratory AGUS frequencies reported from other studies using CPs have varied from 0.1–0.7%.2–16 In general, those laboratories that report lower AGUS frequencies have higher PPVs for the detection of a high-grade precursor lesion or carcinoma and for lesions found to have glandular rather than squamous differentiation on follow-up.6, 9, 16 In contrast, laboratories with higher AGUS frequencies have shown a high preponderance of either benign processes or high-grade CIN on biopsy follow-up.3, 11, 14 In the current study, a similar trend was noted within a single laboratory when the frequency and follow-up of AGUS were compared for the two methods of slide preparation. This was true even though the AGUS frequency was relatively low in both groups (0.19% in CPs and 0.15% in TPs).

To our knowledge, few other reports to date have compared AGUS in TPs and CPs.21–23 Using historic CP controls and biopsy follow-up of AGUS in 25 TP cases and 39 CP cases, Ashfaq et al.21 reported a PPV for high-grade lesions or carcinoma (either squamous or glandular) of 52% for TPs and 33% for CPs. That report also assessed available prior cytologic results in patients with biopsy-confirmed adenocarcinoma or AIS and concluded that TP increased the sensitivity as well as the specificity for these lesions. Bai et al.,22 also using biopsy follow-up and historic CP controls, noted a 50% reduction in AGUS interpretations with TPs and a PPV of 40% for 35 TPs versus 26% for 72 CPs. A third study, using split-sample comparisons in 30 cases of cervical AIS, reported a lower sensitivity for TPs.23 However, seven of the false-negative TP results in that study were diagnosed as “endocervical atypia,”indicating, as the authors themselves stated, that the diagnostic criteria for AIS in TPs need refinement and that enhanced performance may be expected once they are learned. The results of the current study are similar to those of the two prior comparison studies using historic controls,21, 22 although our PPVs for AGUS were lower than those reported in either study for both TPs and CPs and were not significantly different. The lower PPVs that we observed may be due to our having included patients with only cytologic follow-up as well as those patients with biopsy follow-up, because lesions that are biopsied after a cytologic interpretation may be more likely to be neoplastic. Taken together, the current study data and those previously published suggest that the use of TPs increases AGUS specificity for a significant cervical lesion, thereby offering the possibility of the improved clinical management of this often troubling cytologic interpretation.