Interobserver Agreement Across Subspecialties for Diagnosis of Differentiated Vulvar Intraepithelial Neoplasia and Predictive Values of 20 Histologic Features

With institutional review board approval, the University of Michigan (Ann Arbor) surgical pathology database was searched for “differentiated vulvar intraepithelial neoplasia,” “lichen sclerosus,” and “lichen simplex chronicus” diagnosed on vulvar biopsies between 2008 and 2018. The final cohort included all cases diagnosed as dVIN (n = 65; 34 with prior, concurrent, or subsequent SCC; 31 without prior, concurrent, or subsequent SCC), 126 cases of LS, 112 cases of LSC, and 6 cases of LS with LSC. All patients with a diagnosis of dVIN carried a diagnosis of LS. IHC stains for p53 were performed by the signing pathologist for 20 cases of dVIN (31%); p53 immunostains were ordered for research purposes on 20 additional cases. Absence of definitive aberrant p53 expression by immunohistochemistry did not preclude a diagnosis of dVIN in the presence of convincing morphology and absence of HPV-associated disease. Twenty dVIN biopsies (31%) showed definitive aberrant p53 staining. Twenty-three biopsies (35%) lacking definitive aberrant p53 expression were negative for p16 and were from patients with a history of VIN with definitive aberrant p53 staining. The remaining 22 dVIN biopsies (34%) lacking aberrant p53 expression were negative for p16 and were from patients without a history of HPV-associated disease. Cases of LS, LSC, and LS with LSC were included if the patients had at least 2 years of follow-up without diagnosis of dVIN or SCC. Clinical data were recorded. Four pathologists (2 pathology-trained dermatopathologists, 1 gynecologic pathologist, and 1 then–general surgical pathology fellow) independently evaluated the presence or absence of a list of histopathologic features relevant to dVIN and provided a diagnosis for each case. The histopathologic features included epidermal thickening, elongated retes, anastomosing retes, parakeratosis, orthokeratosis, hypergranulosis, keratin pearls, homogenized dermal collagen, lichenoid inflammation, blurring of the basement membrane/dermoepidermal junction, nuclear enlargement, vesicular nuclei, prominent nucleoli, paradoxic maturation, dyskeratosis, intercellular bridges, basal layer disarray, basal pleomorphism, suprabasilar mitoses, and atypical mitotic figures.

Statistical analysis was performed using the R statistical package (version 3.6.3). Interobserver agreement between multiple pathologists for each of the histopathologic features was calculated by the Fleiss κ statistic using the irr package. The cutoff values of the κ statistic used to measure interobserver agreement were <0.00 poor, 0.00 to 0.20 slight, 0.21 to 0.40 fair, 0.41 to 0.60 moderate, 0.61 to 0.80 substantial, and 0.81 to 1.00 almost perfect. Logistic regression analysis was performed using histopathologic features with κ values greater than 0.40 as independent variables and the presence or absence of dVIN as the dependent variable, using the R statistical package. Multicollinearity among the dependent variables was ruled out by checking for variance inflation. Dominance analysis was performed using the dominance analysis (2.0.0) package, which was used to identify the relative predictive value of each histopathologic feature used in logistic regression.

Logistic regression analysis indicated that keratin pearls, basal pleomorphism, and basal layer disarray (Figure 1; Table 1) were independent variables for diagnosing dVIN (coefficients 1.95, 1.97, and 0.91, respectively; P < .001). The same 3 features also predicted progression to SCC (coefficients 1.96, 1.20, and 1.08, respectively; P < .001). These findings were supported by conditional dominance of these features over the other histopathologic features in the diagnosis of dVIN and in the prediction of progression to SCC (Table 1).

The κ values for the 20 histomorphologic features ranged from 0.06 to 0.67, corresponding to slight to substantial agreement (Table 2). The κ value for the diagnoses of dVIN, LS, and LSC among the 4 pathologists was 0.52 (moderate agreement; Table 2).

Agreement was at least moderate among all study pathologists regarding epidermal thickening, elongated retes, and gestalt diagnosis (Figure 2) and there was fair to substantial agreement about the presence of anastomosing retes. Agreement was poor to slight across all pathologists for blurring of the basement membrane and paradoxical maturation. Agreement between each pair of pathologists for each feature is shown in Table 3.

Basal pleomorphism, keratin pearls, and basal layer disarray provided higher predictive value for the diagnosis of dVIN in small models with fewer predictor variables. When model size was increased by adding additional predictors, basal pleomorphism, keratin pearls, and basal layer disarray continued to demonstrate a higher average additional contribution compared with the other predictor variables. Interestingly, basal pleomorphism, basal layer disarray, and keratin pearls showed significantly higher average additional contribution in models across all sizes (Supplemental Tables 1 and 2, see the supplemental digital content at https://meridian.allenpress.com/aplm in the December 2023 table of contents).

The 5th edition of the WHO Classification of Female Genital Tumours updated the preferred terminology of dVIN to “vulvar intraepithelial neoplasia, HPV-independent,” although dVIN is still an accepted term.¹⁷  Sequencing studies have shown that somatic mutations involving TP53 frequently occur in dVIN and HPV-independent SCC, whereas mutations in HRAS, PIK3CA, NOTCH1, and other genes have also been identified in a subset of HPV-independent squamous lesions.^12,14,17  Additional HPV-independent squamous lesions, including differentiated exophytic vulvar intraepithelial lesion and vulvar acanthosis with altered differentiation, have been described as verrucous squamous proliferations without cytologic atypia.^14,17,20,21  Differentiated vulvar intraepithelial neoplasia may be arduous to diagnose, because of its subtle yet diverse histomorphology and many morphologic mimics.^{5,7,19,22–27}  Based on survey data from 23 pathologists with demonstrated expertise in diagnosis of dVIN (for example, membership in the International Society for the Study of Vulvovaginal Disease or authorship of dVIN publications), only “basal layer atypia” met consensus as essential for diagnosis of dVIN, among at least 13 histopathologic features.²⁷  The IHC stain for p53 serves as a powerful ancillary tool but is imperfect and cannot replace histomorphologic evaluation. This is because the correlation between p53 expression and TP53 mutation is not 100%, and a subset of dVINs have nonaberrant expression of p53 by immunohistochemistry.^5,7,26,28,29  Furthermore, as previously mentioned, the p53 stain can be difficult to interpret because of overlap between benign and dysplastic areas, as well as overinterpretation of staining of variable intensity.^7,12 

Our data again emphasize the importance of basal cytologic atypia and disturbed maturation, as keratin pearls, basal layer disarray, and basal pleomorphism were the features most strongly associated with a diagnosis of dVIN and with the development of SCC. Our findings add to the body of literature identifying pertinent histopathologic features for the diagnosis of dVIN, underscore the notion that no single histopathologic feature is pathognomonic for dVIN, and emphasize the importance of a constellation of histopathologic features. Despite the diverse histomorphology of dVIN, our study shows that some features carry more weight than others and that dVIN as well as SCC should be strongly considered in the presence of basal pleomorphism, basal layer disarray, and/or keratin pearls. Although basal atypia is widely considered to be a critical feature of dVIN, Dasgupta et al¹⁹  noted that parakeratosis and cobblestone appearance are useful in cases without nuclear atypia appreciable at ×100 magnification. Interestingly, homogenized dermal collagen had a negative coefficient in our study, indicating that the presence of this finding might weigh against the diagnosis of dVIN. In our cohort, atypical mitotic figures were exceedingly rare, thus limiting their utility.

In subspecialty practices, dVIN may be diagnosed by dermatopathologists or gynecologic pathologists. However, because of dermatopathologists’ limited exposure to vulvar biopsies and gynecologic pathologists’ lesser degree of familiarity with inflammatory dermatoses, diagnosing dVIN remains challenging for pathologists in both subspecialties. Interobserver agreement is particularly suboptimal between gynecologic pathologists and general surgical pathologists or dermatopathologists.^18,19,27  The 4 pathologists in our study had fellowship training in general surgical (1), gynecologic (1), or dermatologic (2) pathology. Interestingly, despite the variable agreement about the presence or absence of many histopathologic features, there was a moderate to substantial agreement between all study pathologists with respect to the diagnoses. The relatively high agreement among the 4 pathologists in this study may be explained by continual self-learning and frequent intersubspecialty consultation during several years because of our shared academic interest in vulvar pathology. Dasgupta et al¹⁹  similarly found more substantial agreement between pathologists working in close collaboration at the same center than between pathologists working at different institutions.

Agreement between our study dermatopathologists was strongest for epidermal thickening, elongated retes, parakeratosis, keratin pearls, homogenized dermal collagen, lichenoid inflammation, nuclear enlargement, and basal pleomorphism. Many of these features are consistently relevant when assessing cutaneous neoplasia or inflammatory dermatoses. Interestingly, the gynecologic pathologist and the general surgical pathologist showed substantial agreement for many of these same features (but not often with dermatopathologists). The most widely agreed-upon features were epidermal thickening and elongated retes. The gynecologic pathologist and the general surgical pathologist substantially agreed on a number of cytologic features, including nuclear enlargement, vesicular nuclei, and basal pleomorphism, as well as basal layer disarray. The distribution of features with substantial agreement likely reflects the features that are emphasized in different subspecialty or general pathology training. The cross-subspecialty differences have also been demonstrated by the survey data reported by Reutter et al²⁷  that included gynecologic pathologists and dermatopathologists with an average 17 years of experience. The study found that dermatopathologists value thickened epithelium, fused and branched retes, dyskeratotic cells in the basal layer, and background lichenoid inflammation more than gynecologic pathologists.²⁷ 

As alluded to above, despite the differences in interobserver agreement for the histopathologic features, the interobserver agreements for the diagnosis of dVIN and its mimics are similar and good among the study pathologists. Furthermore, the deficiency in interobserver concordance can be improved with training. A recent study shows that studying the diagnostic guidelines for dVIN significantly improved diagnostic accuracy among general surgical and gynecologic pathologists.²² 

Strengths of this study include the examination of a large cohort of vulvar biopsies, including 65 cases of dVIN, 126 LS, 112 LSC, and 6 LS with LSC. All cases were reviewed by 2 fellowship-trained dermatopathologists, a fellowship-trained gynecologic pathologist, and a fellowship-trained general surgical pathologist. Weaknesses of this study include the absence of a true gold standard for diagnosis of dVIN. Additionally, TP53 mutational status and p53 IHC expression were not examined in this study. Both ancillary studies when combined with clinical information and histomorphologic evaluation improve the accurate diagnosis of dVIN. Although the diagnostic criteria for dVIN are somewhat subjective, the large proportion of cases with concurrent or subsequent SCC supports the diagnosis. Nevertheless, it is reassuring that the histopathologic features most associated with dVIN diagnosis were also strongly associated with the development of SCC, suggesting that many if not all of these cases did truly represent dVIN. Although the morphologic overlap between LS, LSC, and dVIN is a potential limitation, the included cases of LS and LSC had at least 2 years of benign follow-up, arguing against the possibility that the lesions represented significant dysplasia at the time of the included specimens.

Based on our data, keratin pearls, basal layer disarray, and basal pleomorphism are the histopathologic features most strongly associated with the diagnosis of dVIN and with development of SCC. Although pairwise comparisons show that interobserver agreement about individual features is variable, our study pathologists were able to agree on gestalt diagnoses on most cases. Our findings demonstrate that knowledge in both subspecialties is advantageous for diagnosing dVIN. Consultation of challenging cases with gynecologic and dermatologic pathology should be encouraged.