- Email: [email protected]
Architectural overlap between benign endocervix and pattern-A endocervical adenocarcinoma: Are all pattern-A tumors invasive?
Accepted Manuscript Title: Architectural Overlap between Benign Endocervix and Pattern-A Endocervical Adenocarcinoma: Are All Pattern-A Tumors Invasive? Authors: Gregory Douglas, Brooke E. Howitt, John K. Schoolmeester, Lauren Schwartz, Zuzana Kos, Shahidul Islam, Bojana Djordjevic, Carlos Parra-Herran PII: DOI: Reference:
S0344-0338(16)30670-7 http://dx.doi.org/doi:10.1016/j.prp.2017.03.008 PRP 51772
To appear in: Received date:
16-11-2016
Please cite this article as: Gregory Douglas, Brooke E.Howitt, John K.Schoolmeester, Lauren Schwartz, Zuzana Kos, Shahidul Islam, Bojana Djordjevic, Carlos ParraHerran, Architectural Overlap between Benign Endocervix and Pattern-A Endocervical Adenocarcinoma: Are All Pattern-A Tumors Invasive?, Pathology - Research and Practice http://dx.doi.org/10.1016/j.prp.2017.03.008 This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.
1 Architectural Overlap between Benign Endocervix and Pattern-A Endocervical Adenocarcinoma: Are All Pattern-A Tumors Invasive?
AUTHORS: Gregory Douglas1, Brooke E Howitt2, John K Schoolmeester3, Lauren Schwartz4, Zuzana Kos1, Shahidul Islam1, Bojana Djordjevic5 and Carlos Parra-Herran5.
Institutional affiliations: 1 The Ottawa Hospital and University of Ottawa, Ottawa ON, Canada
2
Brigham and Women's Hospital and Harvard Medical School, Boston MA, United States
3
Mayo Clinic, Rochester MN, United States
4
Perelman School of Medicine, University of Pennsylvania, Philadelphia PA, United States
5
Sunnybrook Health Sciences Centre and University of Toronto, Toronto ON, Canada
CORRESPONDING AUTHOR: Carlos Parra-Herran MD
Assistant Professor of Laboratory Medicine and Pathobiology, University of Toronto
Anatomic Pathologist, Sunnybrook Health Sciences Centre
2075 Bayview Ave Room E4 27a, Toronto ON Canada. [email protected]
2 CONFLICT OF INTEREST DISCLOSURE: The authors have no conflict of interest to disclose. ABSTRACT
Studies on the pattern-based classification for invasive endocervical adenocarcinoma showed that tumors with nondestructive invasion (patternA) have a 0% rate of nodal metastases. Our understanding of pattern-A tumors and their distinction from in-situ adenocarcinoma requires further study. Thirteen sections diagnosed independently as pattern-A adenocarcinoma by three gynecologic pathologists, and 14 sections of benign endocervix were selected. Three additional pathologists (reviewers) evaluated a digital image from each section and classified it as pattern-A or benign based on architecture only. To blind the interpretation to cytologic features, nuclei and cytoplasm were obscured using morphometric software (Zen 2011, Carl Zeiss Microscopy, Germany). 13/27 cases (48%; 8 pattern-A, 5 benign) were correctly classified by all reviewers; 19/27 (70%; 10 pattern-A, 9 benign) were correctly classified by ≥2 reviewers. 3/13 pattern-A cases (23%) were interpreted as benign by ≥2 reviewers. Conversely, 5/14 benign cervices (36%) were misinterpreted as pattern-A by ≥2 reviewers. The number of glands per 20x field was higher in pattern-A cases with high reviewer agreement (p=0.004). An abnormal architecture is seen in many pattern-A adenocarcinomas in support of their invasive nature; some, however, have architecture that overlaps with that of benign endocervix thus may actually represent in-situ lesions. Likewise, normal cervix can be architecturally complex and mirror patterns that pathologists would classify as pattern-A if malignant cytologic features were present. Based on this overlap and the nil risk of nodal spread, an emphasis on the non-destructive, rather than the invasive, nature of pattern-A adenocarcinoma is recommended.
Keywords: Cervical adenocarcinoma; pattern based classification; cervix histology
3 INTRODUCTION Endocervical adenocarcinoma of the usual type is an increasingly prevalent disease related to high-risk Human Papillomavirus infection. With the objective to identify better prognostic factors for invasive endocervical adenocarcinoma, a novel classification system has been recently proposed1,2. The system divides invasive tumors in three categories according to the pattern of stromal invasion. The pattern correlates with the rate of lymph node metastases, being 0% in tumors with a non-destructive pattern of invasion (pattern-A), compared to 4% and 23% in tumors with focal (pattern-B) and diffuse (pattern-C) destructive invasion, respectively1,2. Thus, pattern assessment can separate patients amenable for a more conservative management from patients with a risk for nodal spread and need for standard treatment (radical hysterectomy and regional lymphadenectomy). Subsequent studies have showed modest to good interobserver reproducibility for the pattern-based classification3–5, as well as excellent correlation between pattern in partial sampling (loop electrocautery excision procedure) and pattern of the overall tumor in hysterectomy6.
The pattern-based classification applies to invasive tumors only, and adenocarcinoma in situ should be excluded. Nonetheless, the distinction between invasive and in situ adenocarcinoma is frequently challenging, and it has been reported that such distinction is impossible to achieve in as much as 20% of cases7,8. Traditionally, the diagnosis of stromal invasion by endocervical adenocarcinoma is based on two types of morphology:
1) Destructive forms of neoplastic infiltration in the form of confluent, irregular, angulated or fragmented glands, tumor cell clusters or individual cells, lymphatic vascular space invasion and reactive changes in the stroma (desmoplasia and inflammation). This spectrum of changes applies to pattern B and pattern C invasive adenocarcinomas.
4 2) Increased number of glands and glandular density exceeding that of the normal cervix. In the absence of destructive forms of stromal invasion as defined above, this morphology applies to pattern-A invasive adenocarcinoma.
The latter morphology is more likely to overlap with adenocarcinoma in situ. In fact, distinction of pattern-A invasive tumors from adenocarcinoma in situ suffers from significant interobserver variation 4.
In order to further investigate the well-known diagnostic challenge of differentiating between in situ and invasive carcinoma and its repercussions on the pattern-based classification, we aimed to better characterize the architecture of Pattern-A adenocarcinomas and compare it to that of the normal cervix (as a reliable surrogate of the architecture of adenocarcinoma in-situ). We hypothesize that, although endocervical adenocarcinoma can invade in a non-destructive manner, a proportion of pattern-A invasive tumors may represent in-situ neoplasia colonizing benign glands, which would explain their nil rates of nodal spread and adverse features. Therefore, we sought to: 1) assess the reproducibility of the distinction between the two by pathologists using architectural features only, and 2) compare benign endocervix and pattern-A adenocarcinoma using quantitative criteria.
5 MATERIALS AND METHODS This study was approved by the Ottawa Health Science Research Ethics Board and the Partners Human Research Committee. Case selection and full histopathologic review Consecutive cases of invasive endocervical adenocarcinomas diagnosed on cone biopsy or hysterectomy specimens were obtained from the archives of The Ottawa Hospital (cases accessioned between 2010 and 2014) and Brigham and Women’s Hospital (cases accessioned between 2005 and 2014). Only endocervical adenocarcinomas of usual (endocervical) type were selected; other types (clear cell, gastric, metastatic) were excluded. One gynecologic pathologist (CPH) reviewed all the histologic material available and selected two glass slides containing the largest amount of tumor. Selected histologic material was independently reviewed by three academic gynecologic pathologists (BH, JKS and LS). Each pathologist rendered a diagnosis of in-situ or invasive adenocarcinoma, classifying the latter as pattern-A, B or C; for this step, seminal references outlining criteria for pattern-based classification were provided1,2. Cases with 100% (3/3 pathologists) agreement in the diagnosis of pattern-A invasive adenocarcinoma were selected for further analysis. Consecutive hysterectomy specimens containing benign cervix were also retrieved from the files of The Ottawa Hospital (cases accessioned between 2013 and 2014). Material was reviewed by one gynecologic pathologist (CPH); representative cases with type-B tunnel clusters, diffuse laminar hyperplasia or normal glandular architecture were selected for comparative analysis to match the pattern-A adenocarcinoma cohort. Image analysis and architecture-only review
6 One high resolution image at 20x magnification was obtained from the most representative area of each histologic section selected. The number of glands in each image was manually counted. Glands were counted as round or elongated (including clefts). The round gland / elongated gland ratio was calculated. To blind the pathologist interpretation to the cytologic features, nuclei and cytoplasm were obscured using the morphometric software Zen 2011 (Carl Zeiss Microscopy, Germany). The system detected all nuclear and cytoplasmic areas in the image based on ranges of optical density, and highlighted them in green. This step allowed blocking the cellular details while retaining the glandular architecture (see Figures 2 and 3). Three additional gynecologic pathologists (ZK, SI and BD, referred to as reviewers) individually evaluated the digitally manipulated images. They were asked to render a diagnosis of either benign endocervix or pattern-A invasive adenocarcinoma based on the glandular architecture. Data analysis Original diagnoses and interpretations by reviewers were compared as percentages of concordance. Levels of agreement among the three reviewers were recorded as perfect (3/3 reviewers) or majority (2/3 reviewers). Comparison of gland quantity between adenocarcinoma and normal endocervix groups was performed using unpaired T test for comparison of means (GraphPad Software Inc, La Jolla, CA United States); a p value of less than 0.05 was considered statistically significant.
7 RESULTS
A total of 77 cases of invasive endocervical adenocarcinoma were evaluated. Of these, 8 (10.4%) were uniformly interpreted as pattern-A adenocarcinoma by all three initial pathologists. A total of thirteen 20x magnification images were obtained from these 8 cases: 5 had two corresponding images each, and 3 had one corresponding image each.
The comparison group was comprised of fourteen 20x magnification images obtained from 12 hysterectomies with a benign endocervix: 10 cases had one corresponding image each, and 2 cases had two corresponding images each. Images depicted unremarkable endocervical glands (n=7, 50%), tunnel clusters (n=5, 35.7%) or diffuse laminar endocervical hyperplasia (n=2, 14.3%).
Distinction between Pattern-A adenocarcinoma and benign endocervix based on architecture alone
Distribution of cases based on levels of reviewer agreement is depicted in Table 1 and Figure 1. Out of the total of 27 images, 13 (48%) were correctly classified by all three reviewers (perfect agreement). Nineteen out of 27 images (70%) were correctly classified by at least two reviewers (majority agreement). Representative images of cases with high agreement are depicted in Figure 2.
Out of 13 pattern-A adenocarcinoma images, 8 (61.5%) had perfect agreement between full histopathologic and architecture-only interpretations, whereas 10 (77%) had majority agreement. The remaining 3 pattern-A adenocarcinoma images (23%) were misinterpreted as benign cervix by at least two reviewers in the architecture-only evaluation (see Figure 3).
8 Out of 14 benign cervix images, only 5 (35.7%) had perfect agreement and 9 (64.3%) had majority agreement (among these, 4 depicted unremarkable cervix, 4 contained type-B tunnel clusters and one showed laminar endocervical hyperplasia, see Figure 2). The remaining 5 benign cervix images (36%) were misinterpreted as pattern-A adenocarcinoma by at least two reviewers after architecture-only evaluation. Among these, three depicted normal endocervical glandular architecture, one contained type-B tunnel clusters and one showed diffuse laminar endocervical hyperplasia (see Figure 3).
Quantitative differences between Pattern-A adenocarcinoma and benign endocervix
The number of glandular structures in each image was manually counted. Mean and standard deviation values and comparison between groups are depicted in Table 2. When comparing all cases of benign endocervix and pattern-A adenocarcinoma, all variables (total number of glands, number of round and elongated glands, and round-to-elongated gland ratio) were similar between groups with no statistical differences. However, when comparing only cases with high agreement between architecture-only reviewers (≥2/3) and original diagnosis, pattern-A adenocarcinomas had a significantly higher number of glands (p=0.004). The amount of round glands was also higher in pattern-A adenocarcinomas (p=0.007).
9 DISCUSSION
Endocervical adenocarcinoma is prevalent gynecologic malignancy9,10. A significant proportion of women affected by this condition are young and of reproductive age. In these patients, standard treatment modalities such as hysterectomy and regional lymph node dissection greatly affect fertility and childbearing and impose significant comorbidity. This underscores the need to better differentiate low risk lesions (including in situ adenocarcinoma) which are amenable to a more conservative management. The new pattern based classification is a promising useful tool to achieve this distinction 11.
Pattern-A tumors are characterized by well-demarcated glands which frequently form groups. Glands demonstrate a pushing, rather than a destructive pattern of stromal invasion. Nonetheless, many of the features assigned to this pattern are not easily distinguishable from benign growth and in-situ adenocarcinoma. Indeed, the distinction between in situ and pattern-A invasive adenocarcinoma among gynecologic pathologists is limited by significant interoberver variation4.
Using quantitative and interobserver reproducibility approaches, in this proof-of-concept study we present evidence that pattern-A endocervical adenocarcinomas are indeed non-destructive forms of stromal invasion. The high level of concordance in distinguishing pattern-A adenocarcinoma from normal cervical glands in our study supports this statement. Furthermore, we document that pattern-A adenocarcinomas with high interobserver agreement have a higher number of glandular structures compared to the normal cervix. This validates the use of increased glandular density as a criterion in the diagnosis of non-destructive invasive adenocarcinoma, as outlined in previous literature7,12.
10 Although diagnosis based on architectural appraisal was accurate in most instances, misinterpretation occurred in a minority of our cases. This underscores the overlap in glandular architecture between normal endocervix and pattern-A invasive endocervical adenocarcinoma, significant enough to lead to misclassification when only architectural criteria are used. In practice, attention to the cytologic features will indicate if a glandular population is benign or neoplastic. However, our findings raise the argument that some of the lesions interpreted as pattern-A invasive adenocarcinoma as per current definitions may actually represent in situ neoplasia (i.e. adenocarcinoma in-situ).
Our current understanding of the development, anatomy and histology of the normal endocervix is based on the initial models by C. F. Fluhmann, who described the endocervical mucosa as a complex system of mucosal infoldings13,14. Unlike other mucosal linings such as the endometrium, the basic structural unit of the endocervix is not a vertical tube but a series of folds forming grooves and clefts coursing in various directions, which appear after periods of epithelial proliferation driven by hormonal stimulation13,14. The haphazard distribution and orientation of the endocervical mucosal clefts and grooves imparts the highly heterogeneous architecture seen in routine bidimensional tissue preparations. It may also contribute to our limitation in distinguishing between abnormal proliferations confined to the epithelial compartment (in-situ) and those invading cervical stroma. As seen in our results, the architecture of benign cervices can be sufficiently complex to lead pathologists to interpret them as invasive carcinoma when blinded to the cytologic features. Of note, this pitfall was not restricted to tunnel clusters or diffuse laminar endocervical hyperplasia, but was also seen in cases with unremarkable endocervical glandular architecture.
Since the diagnostic agreement of pattern-A adenocarcinoma suffers from interobserver variation, we chose to include only cases with full consensus of Pattern-A diagnosis among three gynecologic pathologists. Although this step decreased the number of pattern-A adenocarcinomas
11 in the study group, it also decreased the chances of including other patterns (in situ carcinoma, pattern B and pattern C). It is important to note that, compared to destructively invasive forms of adenocarcinoma (patterns B and C), pattern-A tumors are less frequent representing 12 to 21% of all cases in published series1–3,6. When diagnosis is based not only on strict morphologic criteria but also consensus by multiple pathologists, the proportion of pattern-A cases appears to be lower: 8/77 cases (10.4%) in our study.
The diagnosis of stromal invasion in endocervical adenocarcinoma has important prognostic and therapeutic implications. Current therapeutic guidelines will recommend aggressive surgical treatment to invasive lesions stage IA2 or larger, including lymphadenectomy15. Conversely, in situ adenocarcinomas are suitable to conservative management including loop electrosurgical excision, cone or trachelectomy. Over-diagnosis of a non-destructive glandular proliferation as invasive may lead to excessive treatment and related complications. Our study underscores the difficulties in distinguishing invasive from non-invasive proliferation, and acknowledging this limitation in routine practice is advisable. The nondestructive nature of pattern-A adenocarcinoma appears to be more reproducible and consistent with its biologic behavior4,5.
In summary, an abnormal architecture distinct from benign patterns is seen in many pattern-A adenocarcinomas, which supports their denomination as invasive lesions. Pattern-A carcinomas with high interobserver agreement have significantly higher glandular densities compared to normal cervix, supporting the above statement. Nevertheless, a minority of cases diagnosed as pattern-A invasive adenocarcinoma display an architecture that overlaps with that of benign endocervical glands. In addition, benign endocervix can show be architecturally complex and mimic the layout attributed to pattern-A tumors. Based on this overlap and the nil risk of nodal spread, we recommend an emphasis on the non-
12 destructive, rather than the invasive, nature of these tumors. As new evidence on the pattern-based classification emerges, this proposed shift in nomenclature will support a more conservative management of patients with low-risk forms of cervical glandular neoplasia.
13 REFERENCES 1.
Diaz De Vivar A, Roma AA, Park KJ, et al. Invasive endocervical adenocarcinoma: proposal for a new pattern-based classification system with significant clinical implications: a multi-institutional study. Int. J. Gynecol. Pathol. 2013;32:592–601.
2.
Roma AA, Diaz De Vivar A, Park KJ, et al. Invasive endocervical adenocarcinoma: a new pattern-based classification system with important clinical significance. Am. J. Surg. Pathol. 2015;39:667–672.
3.
Paquette C, Jeffus SK, Quick CM, et al. Interobserver variability in the application of a proposed histologic subclassification of endocervical adenocarcinoma. Am. J. Surg. Pathol. 2015;39:93–100.
4.
Parra-Herran C, Taljaard M, Djordjevic B, et al. Pattern-based classification of invasive endocervical adenocarcinoma, depth of invasion measurement and distinction from adenocarcinoma in situ: interobserver variation among gynecologic pathologists. Mod. Pathol. 2016;29:879–892.
5.
Rutgers JKL, Roma AA, Park KJ, et al. Pattern classification of endocervical adenocarcinoma: reproducibility and review of criteria. Mod. Pathol. 2016;29:1083–1094.
6.
Djordjevic B, Parra-Herran C. Application of a Pattern-based Classification System for Invasive Endocervical Adenocarcinoma in Cervical Biopsy, Cone and Loop Electrosurgical Excision (LEEP) Material: Pattern on Cone and LEEP is Predictive of Pattern in the Overall Tumor. Int. J. Gynecol. Pathol. 2016;35:456–466.
7.
Zaino RJ. Symposium part I: adenocarcinoma in situ, glandular dysplasia, and early invasive adenocarcinoma of the uterine cervix. Int. J. Gynecol. Pathol. 2002;21:314–326.
8.
Ostör AG. Early invasive adenocarcinoma of the uterine cervix. Int. J. Gynecol. Pathol. 2000;19:29–38.
9.
Smith HO, Tiffany MF, Qualls CR, et al. The rising incidence of adenocarcinoma relative to squamous cell carcinoma of the uterine cervix in the United States--a 24-year population-based study. Gynecol. Oncol. 2000;78:97–105.
10. Webb JC, Key CR, Qualls CR, et al. Population-based study of microinvasive adenocarcinoma of the uterine cervix. Obstet. Gynecol. 2001;97:701–706. 11. Roma AA. Patterns of Invasion of Cervical Adenocarcinoma as Predicators of Outcome. Adv. Anat. Pathol. 2015;22:345–354. 12. McCluggage WG. Endocervical glandular lesions: controversial aspects and ancillary techniques. J. Clin. Pathol. 2003;56:164–173.
14 13. Fluhmann CF. The Cervix Uteri and its Diseases. First edition. Philadelphia: W.B. Saunders Company; 1961:12-100. 14. Fluhmann CF. The nature and development of the so-called glands of the cervix uteri. Am. J. Obstet. Gynecol. 1957;74:753–766,766–768. 15. Pecorelli S, Zigliani L, Odicino F. Revised FIGO staging for carcinoma of the cervix. Int. J. Gynaecol. Obstet. Off. Organ Int. Fed. Gynaecol. Obstet. 2009;105:107–108.
15 FIGURE LEGENDS
Figure 1. Distribution of cases according to the level of interobserver agreement among pathologists in based on architecture-only review. Most images in both pattern-A adenocarcinoma and benign groups were correctly classified by all (3/3) or most (2/3) reviewers.
Figure 2. Examples of benign endocervix with tunnel clusters (A), benign endocervix with superficial laminar hyperplasia (C) and pattern-A invasive endocervical adenocarcinoma (E). In these images there was perfect agreement between original and reviewer diagnosis (3/3 reviewers). Typical architectural features allowed for correct interpretation after masking the cytologic details (B, D and F).
16
Figure 3. Endocervical adenocarcinoma diagnosed as invasive pattern-A by 3 pathologists (A). Once cytologic details were obscured (B), all 3 reviewers classified it as benign endocervix. In contrast, a case of benign cervix with a florid glandular component (C) was misdiagnosed as pattern-A invasive adenocarcinoma by 2/3 reviewers when blinded to the cytologic features (D).
17
Table 1. Distribution of cases according to level of agreement among reviewers with the original diagnosis Index Dx CASE No. Reviewer 1 Reviewer 2 Reviewer 3
2
5
7
14
15
Normal endocervix 11 22 9 20 12
18
27
Concordance between index and reviewer diagnoses Discordance between index and reviewer diagnoses
3
24
1
4
6
8
Pattern-A adenocarcinoma 17 19 21 23 10 25
16
13
26
18 Table 2. Comparison of glandular density quantification between pattern-A endocervical adenocarcinoma and normal endocervix All cases
Cases with high agre
Benign endocervix (n=14)
Pattern-A invasive adenocarcinoma (n=13)
p value
Benign endocervix (n=9)
Total number of glands*
134.6 (73-229)
183.7 (39-317)
0.1
125.3 (73-195)
225.
Number of round glands*
110.6 (43-200)
159.3 (33-312)
0.09
102.2 (43-168)
198
Number of elongated glands or clefts*
24 (4-58)
24.1 (5-104)
0.99
23.1 (4-58)
27.
0.13
8.1 (1.4-30.5)
20.1
Round to elongated 7.3 (1.4-30.5) 16.2 (1.7-62.4) gland ratio* * Per 20x microscopic field. Mean (range) **Concordance with original diagnosis by ≥2/3 reviewers
Patter adenoca
S0344-0338(16)30670-7 http://dx.doi.org/doi:10.1016/j.prp.2017.03.008 PRP 51772
To appear in: Received date:
16-11-2016
Please cite this article as: Gregory Douglas, Brooke E.Howitt, John K.Schoolmeester, Lauren Schwartz, Zuzana Kos, Shahidul Islam, Bojana Djordjevic, Carlos ParraHerran, Architectural Overlap between Benign Endocervix and Pattern-A Endocervical Adenocarcinoma: Are All Pattern-A Tumors Invasive?, Pathology - Research and Practice http://dx.doi.org/10.1016/j.prp.2017.03.008 This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.
1 Architectural Overlap between Benign Endocervix and Pattern-A Endocervical Adenocarcinoma: Are All Pattern-A Tumors Invasive?
AUTHORS: Gregory Douglas1, Brooke E Howitt2, John K Schoolmeester3, Lauren Schwartz4, Zuzana Kos1, Shahidul Islam1, Bojana Djordjevic5 and Carlos Parra-Herran5.
Institutional affiliations: 1 The Ottawa Hospital and University of Ottawa, Ottawa ON, Canada
2
Brigham and Women's Hospital and Harvard Medical School, Boston MA, United States
3
Mayo Clinic, Rochester MN, United States
4
Perelman School of Medicine, University of Pennsylvania, Philadelphia PA, United States
5
Sunnybrook Health Sciences Centre and University of Toronto, Toronto ON, Canada
CORRESPONDING AUTHOR: Carlos Parra-Herran MD
Assistant Professor of Laboratory Medicine and Pathobiology, University of Toronto
Anatomic Pathologist, Sunnybrook Health Sciences Centre
2075 Bayview Ave Room E4 27a, Toronto ON Canada. [email protected]
2 CONFLICT OF INTEREST DISCLOSURE: The authors have no conflict of interest to disclose. ABSTRACT
Studies on the pattern-based classification for invasive endocervical adenocarcinoma showed that tumors with nondestructive invasion (patternA) have a 0% rate of nodal metastases. Our understanding of pattern-A tumors and their distinction from in-situ adenocarcinoma requires further study. Thirteen sections diagnosed independently as pattern-A adenocarcinoma by three gynecologic pathologists, and 14 sections of benign endocervix were selected. Three additional pathologists (reviewers) evaluated a digital image from each section and classified it as pattern-A or benign based on architecture only. To blind the interpretation to cytologic features, nuclei and cytoplasm were obscured using morphometric software (Zen 2011, Carl Zeiss Microscopy, Germany). 13/27 cases (48%; 8 pattern-A, 5 benign) were correctly classified by all reviewers; 19/27 (70%; 10 pattern-A, 9 benign) were correctly classified by ≥2 reviewers. 3/13 pattern-A cases (23%) were interpreted as benign by ≥2 reviewers. Conversely, 5/14 benign cervices (36%) were misinterpreted as pattern-A by ≥2 reviewers. The number of glands per 20x field was higher in pattern-A cases with high reviewer agreement (p=0.004). An abnormal architecture is seen in many pattern-A adenocarcinomas in support of their invasive nature; some, however, have architecture that overlaps with that of benign endocervix thus may actually represent in-situ lesions. Likewise, normal cervix can be architecturally complex and mirror patterns that pathologists would classify as pattern-A if malignant cytologic features were present. Based on this overlap and the nil risk of nodal spread, an emphasis on the non-destructive, rather than the invasive, nature of pattern-A adenocarcinoma is recommended.
Keywords: Cervical adenocarcinoma; pattern based classification; cervix histology
3 INTRODUCTION Endocervical adenocarcinoma of the usual type is an increasingly prevalent disease related to high-risk Human Papillomavirus infection. With the objective to identify better prognostic factors for invasive endocervical adenocarcinoma, a novel classification system has been recently proposed1,2. The system divides invasive tumors in three categories according to the pattern of stromal invasion. The pattern correlates with the rate of lymph node metastases, being 0% in tumors with a non-destructive pattern of invasion (pattern-A), compared to 4% and 23% in tumors with focal (pattern-B) and diffuse (pattern-C) destructive invasion, respectively1,2. Thus, pattern assessment can separate patients amenable for a more conservative management from patients with a risk for nodal spread and need for standard treatment (radical hysterectomy and regional lymphadenectomy). Subsequent studies have showed modest to good interobserver reproducibility for the pattern-based classification3–5, as well as excellent correlation between pattern in partial sampling (loop electrocautery excision procedure) and pattern of the overall tumor in hysterectomy6.
The pattern-based classification applies to invasive tumors only, and adenocarcinoma in situ should be excluded. Nonetheless, the distinction between invasive and in situ adenocarcinoma is frequently challenging, and it has been reported that such distinction is impossible to achieve in as much as 20% of cases7,8. Traditionally, the diagnosis of stromal invasion by endocervical adenocarcinoma is based on two types of morphology:
1) Destructive forms of neoplastic infiltration in the form of confluent, irregular, angulated or fragmented glands, tumor cell clusters or individual cells, lymphatic vascular space invasion and reactive changes in the stroma (desmoplasia and inflammation). This spectrum of changes applies to pattern B and pattern C invasive adenocarcinomas.
4 2) Increased number of glands and glandular density exceeding that of the normal cervix. In the absence of destructive forms of stromal invasion as defined above, this morphology applies to pattern-A invasive adenocarcinoma.
The latter morphology is more likely to overlap with adenocarcinoma in situ. In fact, distinction of pattern-A invasive tumors from adenocarcinoma in situ suffers from significant interobserver variation 4.
In order to further investigate the well-known diagnostic challenge of differentiating between in situ and invasive carcinoma and its repercussions on the pattern-based classification, we aimed to better characterize the architecture of Pattern-A adenocarcinomas and compare it to that of the normal cervix (as a reliable surrogate of the architecture of adenocarcinoma in-situ). We hypothesize that, although endocervical adenocarcinoma can invade in a non-destructive manner, a proportion of pattern-A invasive tumors may represent in-situ neoplasia colonizing benign glands, which would explain their nil rates of nodal spread and adverse features. Therefore, we sought to: 1) assess the reproducibility of the distinction between the two by pathologists using architectural features only, and 2) compare benign endocervix and pattern-A adenocarcinoma using quantitative criteria.
5 MATERIALS AND METHODS This study was approved by the Ottawa Health Science Research Ethics Board and the Partners Human Research Committee. Case selection and full histopathologic review Consecutive cases of invasive endocervical adenocarcinomas diagnosed on cone biopsy or hysterectomy specimens were obtained from the archives of The Ottawa Hospital (cases accessioned between 2010 and 2014) and Brigham and Women’s Hospital (cases accessioned between 2005 and 2014). Only endocervical adenocarcinomas of usual (endocervical) type were selected; other types (clear cell, gastric, metastatic) were excluded. One gynecologic pathologist (CPH) reviewed all the histologic material available and selected two glass slides containing the largest amount of tumor. Selected histologic material was independently reviewed by three academic gynecologic pathologists (BH, JKS and LS). Each pathologist rendered a diagnosis of in-situ or invasive adenocarcinoma, classifying the latter as pattern-A, B or C; for this step, seminal references outlining criteria for pattern-based classification were provided1,2. Cases with 100% (3/3 pathologists) agreement in the diagnosis of pattern-A invasive adenocarcinoma were selected for further analysis. Consecutive hysterectomy specimens containing benign cervix were also retrieved from the files of The Ottawa Hospital (cases accessioned between 2013 and 2014). Material was reviewed by one gynecologic pathologist (CPH); representative cases with type-B tunnel clusters, diffuse laminar hyperplasia or normal glandular architecture were selected for comparative analysis to match the pattern-A adenocarcinoma cohort. Image analysis and architecture-only review
6 One high resolution image at 20x magnification was obtained from the most representative area of each histologic section selected. The number of glands in each image was manually counted. Glands were counted as round or elongated (including clefts). The round gland / elongated gland ratio was calculated. To blind the pathologist interpretation to the cytologic features, nuclei and cytoplasm were obscured using the morphometric software Zen 2011 (Carl Zeiss Microscopy, Germany). The system detected all nuclear and cytoplasmic areas in the image based on ranges of optical density, and highlighted them in green. This step allowed blocking the cellular details while retaining the glandular architecture (see Figures 2 and 3). Three additional gynecologic pathologists (ZK, SI and BD, referred to as reviewers) individually evaluated the digitally manipulated images. They were asked to render a diagnosis of either benign endocervix or pattern-A invasive adenocarcinoma based on the glandular architecture. Data analysis Original diagnoses and interpretations by reviewers were compared as percentages of concordance. Levels of agreement among the three reviewers were recorded as perfect (3/3 reviewers) or majority (2/3 reviewers). Comparison of gland quantity between adenocarcinoma and normal endocervix groups was performed using unpaired T test for comparison of means (GraphPad Software Inc, La Jolla, CA United States); a p value of less than 0.05 was considered statistically significant.
7 RESULTS
A total of 77 cases of invasive endocervical adenocarcinoma were evaluated. Of these, 8 (10.4%) were uniformly interpreted as pattern-A adenocarcinoma by all three initial pathologists. A total of thirteen 20x magnification images were obtained from these 8 cases: 5 had two corresponding images each, and 3 had one corresponding image each.
The comparison group was comprised of fourteen 20x magnification images obtained from 12 hysterectomies with a benign endocervix: 10 cases had one corresponding image each, and 2 cases had two corresponding images each. Images depicted unremarkable endocervical glands (n=7, 50%), tunnel clusters (n=5, 35.7%) or diffuse laminar endocervical hyperplasia (n=2, 14.3%).
Distinction between Pattern-A adenocarcinoma and benign endocervix based on architecture alone
Distribution of cases based on levels of reviewer agreement is depicted in Table 1 and Figure 1. Out of the total of 27 images, 13 (48%) were correctly classified by all three reviewers (perfect agreement). Nineteen out of 27 images (70%) were correctly classified by at least two reviewers (majority agreement). Representative images of cases with high agreement are depicted in Figure 2.
Out of 13 pattern-A adenocarcinoma images, 8 (61.5%) had perfect agreement between full histopathologic and architecture-only interpretations, whereas 10 (77%) had majority agreement. The remaining 3 pattern-A adenocarcinoma images (23%) were misinterpreted as benign cervix by at least two reviewers in the architecture-only evaluation (see Figure 3).
8 Out of 14 benign cervix images, only 5 (35.7%) had perfect agreement and 9 (64.3%) had majority agreement (among these, 4 depicted unremarkable cervix, 4 contained type-B tunnel clusters and one showed laminar endocervical hyperplasia, see Figure 2). The remaining 5 benign cervix images (36%) were misinterpreted as pattern-A adenocarcinoma by at least two reviewers after architecture-only evaluation. Among these, three depicted normal endocervical glandular architecture, one contained type-B tunnel clusters and one showed diffuse laminar endocervical hyperplasia (see Figure 3).
Quantitative differences between Pattern-A adenocarcinoma and benign endocervix
The number of glandular structures in each image was manually counted. Mean and standard deviation values and comparison between groups are depicted in Table 2. When comparing all cases of benign endocervix and pattern-A adenocarcinoma, all variables (total number of glands, number of round and elongated glands, and round-to-elongated gland ratio) were similar between groups with no statistical differences. However, when comparing only cases with high agreement between architecture-only reviewers (≥2/3) and original diagnosis, pattern-A adenocarcinomas had a significantly higher number of glands (p=0.004). The amount of round glands was also higher in pattern-A adenocarcinomas (p=0.007).
9 DISCUSSION
Endocervical adenocarcinoma is prevalent gynecologic malignancy9,10. A significant proportion of women affected by this condition are young and of reproductive age. In these patients, standard treatment modalities such as hysterectomy and regional lymph node dissection greatly affect fertility and childbearing and impose significant comorbidity. This underscores the need to better differentiate low risk lesions (including in situ adenocarcinoma) which are amenable to a more conservative management. The new pattern based classification is a promising useful tool to achieve this distinction 11.
Pattern-A tumors are characterized by well-demarcated glands which frequently form groups. Glands demonstrate a pushing, rather than a destructive pattern of stromal invasion. Nonetheless, many of the features assigned to this pattern are not easily distinguishable from benign growth and in-situ adenocarcinoma. Indeed, the distinction between in situ and pattern-A invasive adenocarcinoma among gynecologic pathologists is limited by significant interoberver variation4.
Using quantitative and interobserver reproducibility approaches, in this proof-of-concept study we present evidence that pattern-A endocervical adenocarcinomas are indeed non-destructive forms of stromal invasion. The high level of concordance in distinguishing pattern-A adenocarcinoma from normal cervical glands in our study supports this statement. Furthermore, we document that pattern-A adenocarcinomas with high interobserver agreement have a higher number of glandular structures compared to the normal cervix. This validates the use of increased glandular density as a criterion in the diagnosis of non-destructive invasive adenocarcinoma, as outlined in previous literature7,12.
10 Although diagnosis based on architectural appraisal was accurate in most instances, misinterpretation occurred in a minority of our cases. This underscores the overlap in glandular architecture between normal endocervix and pattern-A invasive endocervical adenocarcinoma, significant enough to lead to misclassification when only architectural criteria are used. In practice, attention to the cytologic features will indicate if a glandular population is benign or neoplastic. However, our findings raise the argument that some of the lesions interpreted as pattern-A invasive adenocarcinoma as per current definitions may actually represent in situ neoplasia (i.e. adenocarcinoma in-situ).
Our current understanding of the development, anatomy and histology of the normal endocervix is based on the initial models by C. F. Fluhmann, who described the endocervical mucosa as a complex system of mucosal infoldings13,14. Unlike other mucosal linings such as the endometrium, the basic structural unit of the endocervix is not a vertical tube but a series of folds forming grooves and clefts coursing in various directions, which appear after periods of epithelial proliferation driven by hormonal stimulation13,14. The haphazard distribution and orientation of the endocervical mucosal clefts and grooves imparts the highly heterogeneous architecture seen in routine bidimensional tissue preparations. It may also contribute to our limitation in distinguishing between abnormal proliferations confined to the epithelial compartment (in-situ) and those invading cervical stroma. As seen in our results, the architecture of benign cervices can be sufficiently complex to lead pathologists to interpret them as invasive carcinoma when blinded to the cytologic features. Of note, this pitfall was not restricted to tunnel clusters or diffuse laminar endocervical hyperplasia, but was also seen in cases with unremarkable endocervical glandular architecture.
Since the diagnostic agreement of pattern-A adenocarcinoma suffers from interobserver variation, we chose to include only cases with full consensus of Pattern-A diagnosis among three gynecologic pathologists. Although this step decreased the number of pattern-A adenocarcinomas
11 in the study group, it also decreased the chances of including other patterns (in situ carcinoma, pattern B and pattern C). It is important to note that, compared to destructively invasive forms of adenocarcinoma (patterns B and C), pattern-A tumors are less frequent representing 12 to 21% of all cases in published series1–3,6. When diagnosis is based not only on strict morphologic criteria but also consensus by multiple pathologists, the proportion of pattern-A cases appears to be lower: 8/77 cases (10.4%) in our study.
The diagnosis of stromal invasion in endocervical adenocarcinoma has important prognostic and therapeutic implications. Current therapeutic guidelines will recommend aggressive surgical treatment to invasive lesions stage IA2 or larger, including lymphadenectomy15. Conversely, in situ adenocarcinomas are suitable to conservative management including loop electrosurgical excision, cone or trachelectomy. Over-diagnosis of a non-destructive glandular proliferation as invasive may lead to excessive treatment and related complications. Our study underscores the difficulties in distinguishing invasive from non-invasive proliferation, and acknowledging this limitation in routine practice is advisable. The nondestructive nature of pattern-A adenocarcinoma appears to be more reproducible and consistent with its biologic behavior4,5.
In summary, an abnormal architecture distinct from benign patterns is seen in many pattern-A adenocarcinomas, which supports their denomination as invasive lesions. Pattern-A carcinomas with high interobserver agreement have significantly higher glandular densities compared to normal cervix, supporting the above statement. Nevertheless, a minority of cases diagnosed as pattern-A invasive adenocarcinoma display an architecture that overlaps with that of benign endocervical glands. In addition, benign endocervix can show be architecturally complex and mimic the layout attributed to pattern-A tumors. Based on this overlap and the nil risk of nodal spread, we recommend an emphasis on the non-
12 destructive, rather than the invasive, nature of these tumors. As new evidence on the pattern-based classification emerges, this proposed shift in nomenclature will support a more conservative management of patients with low-risk forms of cervical glandular neoplasia.
13 REFERENCES 1.
Diaz De Vivar A, Roma AA, Park KJ, et al. Invasive endocervical adenocarcinoma: proposal for a new pattern-based classification system with significant clinical implications: a multi-institutional study. Int. J. Gynecol. Pathol. 2013;32:592–601.
2.
Roma AA, Diaz De Vivar A, Park KJ, et al. Invasive endocervical adenocarcinoma: a new pattern-based classification system with important clinical significance. Am. J. Surg. Pathol. 2015;39:667–672.
3.
Paquette C, Jeffus SK, Quick CM, et al. Interobserver variability in the application of a proposed histologic subclassification of endocervical adenocarcinoma. Am. J. Surg. Pathol. 2015;39:93–100.
4.
Parra-Herran C, Taljaard M, Djordjevic B, et al. Pattern-based classification of invasive endocervical adenocarcinoma, depth of invasion measurement and distinction from adenocarcinoma in situ: interobserver variation among gynecologic pathologists. Mod. Pathol. 2016;29:879–892.
5.
Rutgers JKL, Roma AA, Park KJ, et al. Pattern classification of endocervical adenocarcinoma: reproducibility and review of criteria. Mod. Pathol. 2016;29:1083–1094.
6.
Djordjevic B, Parra-Herran C. Application of a Pattern-based Classification System for Invasive Endocervical Adenocarcinoma in Cervical Biopsy, Cone and Loop Electrosurgical Excision (LEEP) Material: Pattern on Cone and LEEP is Predictive of Pattern in the Overall Tumor. Int. J. Gynecol. Pathol. 2016;35:456–466.
7.
Zaino RJ. Symposium part I: adenocarcinoma in situ, glandular dysplasia, and early invasive adenocarcinoma of the uterine cervix. Int. J. Gynecol. Pathol. 2002;21:314–326.
8.
Ostör AG. Early invasive adenocarcinoma of the uterine cervix. Int. J. Gynecol. Pathol. 2000;19:29–38.
9.
Smith HO, Tiffany MF, Qualls CR, et al. The rising incidence of adenocarcinoma relative to squamous cell carcinoma of the uterine cervix in the United States--a 24-year population-based study. Gynecol. Oncol. 2000;78:97–105.
10. Webb JC, Key CR, Qualls CR, et al. Population-based study of microinvasive adenocarcinoma of the uterine cervix. Obstet. Gynecol. 2001;97:701–706. 11. Roma AA. Patterns of Invasion of Cervical Adenocarcinoma as Predicators of Outcome. Adv. Anat. Pathol. 2015;22:345–354. 12. McCluggage WG. Endocervical glandular lesions: controversial aspects and ancillary techniques. J. Clin. Pathol. 2003;56:164–173.
14 13. Fluhmann CF. The Cervix Uteri and its Diseases. First edition. Philadelphia: W.B. Saunders Company; 1961:12-100. 14. Fluhmann CF. The nature and development of the so-called glands of the cervix uteri. Am. J. Obstet. Gynecol. 1957;74:753–766,766–768. 15. Pecorelli S, Zigliani L, Odicino F. Revised FIGO staging for carcinoma of the cervix. Int. J. Gynaecol. Obstet. Off. Organ Int. Fed. Gynaecol. Obstet. 2009;105:107–108.
15 FIGURE LEGENDS
Figure 1. Distribution of cases according to the level of interobserver agreement among pathologists in based on architecture-only review. Most images in both pattern-A adenocarcinoma and benign groups were correctly classified by all (3/3) or most (2/3) reviewers.
Figure 2. Examples of benign endocervix with tunnel clusters (A), benign endocervix with superficial laminar hyperplasia (C) and pattern-A invasive endocervical adenocarcinoma (E). In these images there was perfect agreement between original and reviewer diagnosis (3/3 reviewers). Typical architectural features allowed for correct interpretation after masking the cytologic details (B, D and F).
16
Figure 3. Endocervical adenocarcinoma diagnosed as invasive pattern-A by 3 pathologists (A). Once cytologic details were obscured (B), all 3 reviewers classified it as benign endocervix. In contrast, a case of benign cervix with a florid glandular component (C) was misdiagnosed as pattern-A invasive adenocarcinoma by 2/3 reviewers when blinded to the cytologic features (D).
17
Table 1. Distribution of cases according to level of agreement among reviewers with the original diagnosis Index Dx CASE No. Reviewer 1 Reviewer 2 Reviewer 3
2
5
7
14
15
Normal endocervix 11 22 9 20 12
18
27
Concordance between index and reviewer diagnoses Discordance between index and reviewer diagnoses
3
24
1
4
6
8
Pattern-A adenocarcinoma 17 19 21 23 10 25
16
13
26
18 Table 2. Comparison of glandular density quantification between pattern-A endocervical adenocarcinoma and normal endocervix All cases
Cases with high agre
Benign endocervix (n=14)
Pattern-A invasive adenocarcinoma (n=13)
p value
Benign endocervix (n=9)
Total number of glands*
134.6 (73-229)
183.7 (39-317)
0.1
125.3 (73-195)
225.
Number of round glands*
110.6 (43-200)
159.3 (33-312)
0.09
102.2 (43-168)
198
Number of elongated glands or clefts*
24 (4-58)
24.1 (5-104)
0.99
23.1 (4-58)
27.
0.13
8.1 (1.4-30.5)
20.1
Round to elongated 7.3 (1.4-30.5) 16.2 (1.7-62.4) gland ratio* * Per 20x microscopic field. Mean (range) **Concordance with original diagnosis by ≥2/3 reviewers
Patter adenoca