Concordance Between Biopsy and Resection Diagnoses of Uterine Cervical Adenocarcinoma According to the Updated World Health Organization 2020 Classification: A Multi-Institutional Study Elucidating Real-World Practice in Japan

The study was approved by the Ethics Committee (authorization number: RIN-2411) at Kumamoto University (Kumamoto, Japan). We retrieved 217 cases of endocervical adenocarcinoma diagnosed based on the 4th (2014)¹⁵  or 5th (2020)²  editions of the WHO classification with availability of glass slides for both biopsy and resection specimens (conization, n = 17; hysterectomy, n = 200) in the pathology database from 2014 to 2021 at 6 academic institutions in Japan. At all participating institutions, the diagnoses were confirmed by board-certified pathologists (Japanese Society of Pathology), who were also members of the Japanese Society of Gynecologic Pathology, specializing in gynecologic pathology and oncology. No standardized guideline or protocol for ancillary studies was available for this project, and each institution applied and performed ISH and/or IHC based on their own policy. Patients who received chemotherapy and/or radiotherapy before biopsy or had a diagnosis of mixed histology were excluded.

Original pathology reports for both biopsy and resection specimens and medical records were reviewed in all cases. Data regarding patient age, history of preoperative therapy, surgical procedure, biopsy and resection specimen diagnoses, ancillary studies, and pathologic stage classification were recorded. The original pathologic stage data, which followed the American Joint Committee on Cancer (AJCC)/Union for International Cancer Control (UICC) 8th edition, were converted to align with the AJCC version 9/UICC 2021 guidelines based on FIGO 2018 criteria. We converted the histopathologic diagnoses made prior to 2020 (following the 4th edition of the WHO classification) to the corresponding diagnostic categories defined in the 5th edition for the current analysis^2,6,15  (Figure 1).

In cases with discordant biopsy and resection specimen diagnoses, the glass sides were critically reviewed by 2 of the authors (F.K. and Y.M.), and misinterpretations were corrected before further analyses. To determine whether the algorithmic approach proposed by the international group contributed to improving the pathologic diagnostic concordance rate between biopsy specimens and resected specimens, we performed ISH and IHC using the recommended panel of antibodies¹⁴  with some modifications.

We first evaluated HPV-ISH (Inform HPV III Family 16 Probe, Ventana Medical Systems, Tucson, Arizona) and p16 IHC (E6H4, ready to use; Ventana Medical Systems), and then added estrogen receptor (ER) IHC (monoclonal, SP1, ready to use; Roche, Basel, Switzerland) for HPV-ISH-negative and/or p16 IHC-negative cases. GATA3 IHC (monoclonal, L50-823, ready to use; Biocare Medical, Pacheco, California) was performed in ER-negative cases.

HPV-ISH was performed as follows. First, 4-μm-thick sections from formalin-fixed, paraffin-embedded blocks were obtained. Samples were assessed by HPV DNA ISH with proprietary reagents on an automated system using a BenchMark ULTRA system (Ventana Medical Systems) according to the manufacturer’s instructions. The probe targets HPV 18, 31, 33, 35, 45, 52, 56, 58, and 66, which are the high-risk types. The presence of diffuse, punctate, or mixed-signal patterns in epithelial cell nuclei indicated positive reactivity.¹⁶ 

For IHC, 4-μm-thick sections from formalin-fixed, paraffin-embedded blocks were incubated with the primary antibodies. p16 and ER were stained using the BenchMark ULTRA system and GATA3 was stained using a BOND-III system (Leica Biosystems, Wetzlar, Germany). For p16, strong diffuse staining was recorded as positive, while heterogeneous, patchy, or no staining was interpreted as negative. The ER status was considered positive if >25% of tumor cell nuclei were positive, and the GATA3 status was considered positive if nuclear staining was noted in >5% of tumor cells.¹⁴ 

Concordance between the biopsy diagnosis and the resection specimen diagnosis was calculated using the κ coefficient for all cases as well as by institution. The κ coefficients were interpreted as follows: ≤0.00, no agreement; 0.01 to 0.20, no to slight agreement; 0.21 to 0.40, fair agreement; 0.41 to 0.60, moderate agreement; 0.61 to 0.80, substantial agreement; and 0.81 to 1.00, almost perfect agreement. The χ² contingency test was used to compare the diagnostic concordance between biopsy and resection specimens according to tumor category, the use of IHC as a diagnostic aid for the biopsy specimen, variations in the proportions of HPVA and HPVI, and the concordance rate of biopsy and resection specimen diagnoses among the participating institutions. To determine whether the 5th edition of the WHO criteria (Figure 2, A through D) could identify HPVA and HPVI in biopsy specimens, we calculated the sensitivity, specificity, positive predictive value, and negative predictive value of a biopsy diagnosis based on these criteria. All statistical analyses were conducted using GraphPad Prism 9 software, version 9.5.1 (GraphPad Software, San Diego, California). A P value of <.05 was considered statistically significant.

The patients’ mean age was 47.9 ± 0.8 years. There was no history of preoperative therapy in 212 patients. Fourteen patients received neoadjuvant chemotherapy, and 1 received concurrent chemoradiotherapy after establishment of a biopsy diagnosis. Conization or loop electrosurgical excision was performed in 40 patients and was followed by total hysterectomy in 23 patients. Total hysterectomy was performed as the primary surgery in 177 patients. The pathologic T stage was pT1a1 in 34 patients, pT1a2 in 16, pT1b1 in 57, pT1b2 in 51, pT1b3 in 12, pT2a1 in 11, pT2a2 in 3, pT2b in 31, and pT3b in 2. Lymph nodes were dissected and examined in 188 patients, and the pathologic N stage was pN0 in 148 patients, pN1 in 36, and pN2 in 4. Distant metastasis (pM1) was found in 4 patients (2 with ovarian metastases, 1 with omental dissemination, and 1 with both ovarian metastasis and omental and peritoneal dissemination).

The biopsy diagnosis matched the resection specimen diagnosis in 197 of 217 cases (concordance rate, 91%; κ = 0.75). Concordance was significantly higher for HPVA than for HPVI (95% versus 81%, P < .05). Table 1 shows the agreement rate between the biopsy and resection specimen diagnoses according to histologic subtype. In the routine clinical practice setting in the participating institutions from 2014 to 2020, biopsy had a sensitivity of 95%, specificity of 84%, positive predictive value of 95%, and negative predictive value of 84% for diagnosis of HPVA and 81%, 97%, 89%, and 96%, respectively, for HPVI (Table 2).

There was no significant difference in the proportions of HPVA and HPVI detected at the 6 institutions (range, 70%–100%; P = .20) or in the concordance rates (range, 89%–100%; P = .69). Although not statistically significant, the proportion of HPVA at one institution was close to 100% with a complete concordance rate (100%). All cases at this institution were diagnosed as HPVA with the exception of 1 case of high-grade adenocarcinoma with an unusual morphology (Figure 3). This deviation from the typical distribution of endocervical adenocarcinoma in Japan, which usually consists of 75% to 85% HPVA and 15% to 25% HPVI, presumably reflects variations in therapeutic policies between individual institutions. In Japan, gynecologists prefer to perform primary surgery for locally advanced cervical cancer (ie, FIGO stages IB or IIA). However, at this particular institution, concurrent chemoradiotherapy is preferred to primary radical surgery in such patients. Therefore, a significant number of patients with HPVI, who frequently have locally advanced disease at presentation, were excluded from this study.

Ancillary studies, including ISH and IHC, were performed in 89 (41%) of the 217 cases to establish the diagnosis in biopsy specimens. Various antibodies or markers were used for IHC (median, 3; range, 1–14), including p16, ER, PgR, HER2, vimentin, CEA, CK7, CK20, CDX2, PAX8, MUC1, MUC2, MUC5AC, MUC6, HIK1083 (anti-GlcNAc α1-R), carbonic anhydrase type-IX, claudin-18, CD10, PTEN, MSH6, PMS2, HNF-1β, napsin A, IMP3, AFP, glypican-3, GCDFP-15, GATA3, WT-1, p53, TTF-1, p40, p63, CK 5/6, INSM1, chromogranin A, synaptophysin, S-100 protein, and Ki-67. HPV-ISH was performed in 1 case. Ancillary studies were ordered more frequently in cases of HPVI than in cases of HPVA (76% [32 of 42] versus 33% [55 of 168]; P < .05).

The most common reason for performing IHC was differentiation between HPVA and HPVI, accounting for 60 of 89 (67%) of the IHC examinations performed. Twenty-three (38%) of these 60 cases were considered to be HPVA based on morphology, showing features that were usual-type (n = 19), signet-ring cell-type mucinous carcinoma (n = 3), or invasive stratified mucin-producing carcinoma (n = 1), and they were confirmed to be HPVA by positive staining for p16 IHC. However, 37 of 60 cases (62%) cases were considered morphologically definite or suspicious for HPVI, and ancillary studies were performed to confirm the diagnosis; additional studies for further classification were performed in a subset of these cases.

The second indication for IHC was determination of the primary tumor site, accounting for 20 of 89 (22%) of all IHC examinations performed. The most common scenario was distinguishing between HPVA showing a pattern of usual-type endocervical adenocarcinoma and endometroid endometrial carcinoma secondarily involving the uterine cervix. Other reasons for IHC included differentiation between nonkeratinizing SCC and poorly differentiated or high-grade endocervical adenocarcinoma with a predominant solid component (6 of 89 [7%] cases) and confirmation of a diagnosis of adenocarcinoma for cases that were morphologically equivocal because of a small amount of tissue sampled or crush artifact (3 of 89 [3%] cases).

There was no difference in the concordance rate for biopsy diagnosis and resection specimen diagnosis between the 128 cases diagnosed without ancillary studies and the 89 cases diagnosed with ancillary studies (91% versus 91%; P = .93).

The algorithmic approach proposed by the international group¹⁴  was applied to biopsy specimens in 19 of the 20 cases with a discordant diagnosis (Figure 4), and all 19 cases were successfully recategorized as HPVA or HPVI by HPV-ISH. However, p16 IHC was positive in 3 of 9 HPV-ISH-negative cases with intracytoplasmic mucin, which represents gastric-type HPVI (Figure 5, A through F). Further subclassification of HPVA and HPVI failed in 3 (16%) of the 19 cases, presumably because of spatial heterogeneity in IHC (Figure 6, A through F).

The distinction of 2 major categories of endocervical adenocarcinoma, namely HPVA and HPVI, is currently considered a critical issue in terms of etiology, early detection, clinical management, prediction of response to therapy and patient outcome, enrollment in clinical trials, and basic research. Therefore, a diagnosis of endocervical adenocarcinoma alone appears to be suboptimal, even as an interpretation of a biopsy specimen. Histologic distinction of the 2 categories proposed in the IECC and the updated WHO classification appears to be valid for resection specimens^7,8  Although p16 IHC or HPV detection is not essential for differentiation between HPVA and HPVI in the 5th edition of the WHO criteria, algorithmic application of ancillary studies may improve diagnostic concordance.²  One example of success using this approach is the diagnostic algorithm proposed by the international group that developed the IECC diagnostic criteria.¹⁴  Briefly, HPV-ISH separates HPVA from endometrial endometrioid carcinoma and HPVI. The ER status can then be used to separate endometrial endometrioid carcinoma from HPVI. For tumors with limited cytoplasmic mucin, GATA3 expression can then be used to differentiate HPVI, mesonephric-type from HPVI, clear cell type. For tumors with cytoplasmic mucin, an HPV-ISH–negative and ER-negative tumor is considered HPVI, gastric-type.

In this multi-institutional study, we found high concordance between biopsy and resection specimen diagnoses of endocervical adenocarcinoma according to the IECC and 5th edition of the WHO criteria in real-world practice in Japan. This finding indicates that application of the recently revised diagnostic criteria to biopsy specimens is feasible.

This study had some unique aspects. First, we skipped a central pathology review and instead adopted the matched resection specimen diagnosis as the gold standard. Second, the pathologists involved in the study were familiar with the histology of HPVI. The patients included in this study were enrolled from 2014 to 2021, which means that a proportion of them were diagnosed before the IECC and 5th edition of the WHO morphologic criteria were advocated. However, the pathologic diagnosis was supervised by gynecologic pathologists who had been familiar with the histology of HPVI since its first description in 2007; one of the authors (Y.M.) was a contributor to this description. Furthermore, another author (T.K.) contributed to establishment of the IECC and shared their diagnostic experience and knowledge with members of our study group. These unique features allowed us to capture a snapshot of the real-world practice of uterine cervical biopsy interpretation based on the IECC and 5th edition of the WHO criteria in Japan.

Our data show that the accuracy of diagnosis of endocervical adenocarcinoma based on a biopsy specimen is comparable to that based on a resection specimen when IHC ancillary studies are included, with a concordance rate of 91%. Our review of pathology reports showed that 59% of biopsy specimens were diagnosed by H&E staining alone and that IHC was performed in the remaining cases. There was no significant difference in the concordance rate between the group diagnosed based on morphology alone and the group diagnosed with the aid of ancillary studies. Moreover, although the accuracy of biopsy diagnosis was lower for HPVI than for HPVA (81% versus 95%), the proportion of cases diagnosed with the aid of ancillary studies was higher for HPVI (76% versus 33%).

There are 2 potential explanations for our unexpected finding that IHC did not improve diagnostic accuracy. The first is the lack of hierarchical application of IHC, such as the algorithmic approach proposed by the international group. The second is the general tendency to use ancillary studies for histologically challenging cases.

To examine the first possibility, we evaluated the impact of the algorithmic approach in selected cases of discordant diagnosis,¹⁴  considering that these cases were originally diagnosed under a different diagnostic policy and approach in various institutions. We found that HPV-ISH successfully categorized all 19 discordant cases as either HPVA or HPVI. Whether p16 IHC can serve as a surrogate for HPV-ISH is controversial. In the present study, p16 was positive in one-third of HPV-ISH-negative cases with abundant intracytoplasmic mucin. This false-positive staining rate is higher than the previously reported diffuse p16 positivity rate in HPVI, which ranges from 8% to 20%.^5,7,17  In the English-language literature, the sensitivity of p16 IHC for detection of HPVA ranges from 87% to 100%, with values of 91.3% and 96.0%, respectively, reported for HPV DNA-ISH and HPV RNA-ISH; this indicates the superiority of HPV-ISH.^6,7  Therefore, the use of HPV-ISH is recommended for differentiation of HPVA from HPVI, and the p16 IHC result should be interpreted with caution, particularly when the tumor shows mucinous histology.

The findings of this study also suggest a limitation of the algorithmic approach in small biopsy specimens. The limited sampling and the intratumoral heterogeneity of the morphology and IHC staining can be sources of confusion and misclassification of HPVI, and further subdivision might be impossible. Indeed, one case of HPVI of the clear cell type was confirmed to have been miscategorized as endometrioid carcinoma because of positive staining for ER. Biopsy diagnosis of endometrioid adenocarcinoma of the uterine cervix, which cannot be distinguished from endometrioid carcinoma of the endometrium by H&E staining, IHC, or HPV-ISH, is also challenging. Endometrioid carcinoma of the uterine cervix might be erroneously categorized as endometrioid carcinoma of the endometrium in biopsy specimens. In principle, subclassification requires examination of the resected specimen. Otherwise, use of IHC should include a broad panel of histology-specific markers such as HIK1083, carbonic anhydrase type IX, and claudin 18 for gastric-type carcinoma; HNF-1β and napsin A for clear cell carcinoma; TTF1 and calretinin for mesonephric carcinoma; and PTEN and ARID1A for endometrioid carcinoma. Secondary involvement of endometrial or extrauterine origin should also be a diagnostic concern in an HPV-negative case, and addition of WT1 may confirm a diagnosis of high-grade serous carcinoma of tubo-ovarian origin.

Regarding the second potential explanation for our unexpected finding that IHC did not improve diagnostic accuracy, ancillary studies are generally applied in challenging cases with rare histology, unusual or atypical morphology, or limited amounts of tissue with or without crush artifact. An algorithmic approach could help to avoid a diagnosis of “endocervical adenocarcinoma not otherwise specified” instead of assigning HPVA or HPVI, which would now be considered unsatisfactory in terms of optimal patient management.

Biopsy diagnosis of HPVI may be difficult because of a variety of morphologic variations.¹⁸  Therefore, strict application of the criteria for HPVA is considered a seminal point, and a subset of cases showing unusual morphology that does not fit with HPVA should provide candidates for ancillary studies. The algorithmic approach employing HPV-ISH and ER and GATA3 IHC proposed by the international group appears to be a reasonably effective tool for improvement of diagnostic accuracy¹⁴  but might not be cost effective or time efficient in diagnostic pathology laboratories; therefore, it should be combined with morphology. Information on imaging studies is also mandatory for distinguishing between endocervical adenocarcinoma and extension of endometrial endometrioid carcinoma.

In summary, we have shown good concordance between biopsy diagnosis and resected specimen diagnosis of endocervical adenocarcinoma in diagnostic pathology practice in the era of the IECC and 5th edition of the WHO criteria. This is the first study to validate use of these criteria in biopsy specimens. Our findings have the potential to impact multiple aspects of the care of patients with endocervical adenocarcinoma, given that a biopsy specimen is the only available tissue material for many patients with uterine cervical adenocarcinoma.

The study was aided by the Japanese Society of Gynecologic Pathology. We thank Edanz (https://jp.edanz.com/ac) for editing a draft of this manuscript.