Improving endometrial carcinoma diagnostics

Abstract

Endometrial carcinoma is the most common gynecological malignancy and greatly affects global disease burden. Despite it being among the most common cancers it is relatively understudied and until recently poorly understood. Diagnostics of endometrial carcinoma have been underdeveloped for many years resulting in perpetual poor interobserver reproducibility in turn inhibiting correct diagnosis and complicating comparisons of clinical trials. In this thesis we used a three-tiered approach in trying to improve endometrial carcinoma diagnostics: strictly improving morphological criteria, adding immunohistochemistry (IHC)/biomarker panels, as well as genomics data in order to improve diagnostic and prognostic ability.

In study 1 we compared standard FIGO grading of endometrial carcinoma to a newly developed morphologic system we named cell-type independent (CTI). We compared interobserver reproducibility using both systems in 70 patient`s endometrial biopsies. We could not identify any significant improvement in the form of increased interobserver reproducibility nor increased accuracy when compared to final hysterectomy diagnosis. We conclude that the limitations of the biopsy setting hampered the CTI system's ability to outperform current FIGO grading system and that the assessment of nuclear grade, no matter what the definitions of a high nuclear grade are, remain subjective.

In study 2 12 separate subspecialized gynaecological pathologists reviewed pathology slides of 70 cases of endometrial biopsies. We compared the interobserver agreement when adding a biomarker panel (IHC for p53, ER and PGR and DNA ploidy analysis) to only standard morphology diagnosis with accompanying results of final hysterectomy diagnosis. The addition of a biomarker panel increased interobserver agreement from 75,8%, kappa = 0.52 to 84%, kappa= 0.68. Diagnostic agreement to final hysterectomy diagnosis also increased with use of the panel, going from 83.6% to 88.7% agreement with final diagnosis after incorporating the panel in biopsy diagnosis (p<0.05). The two biomarkers p53 IHC and DNA ploidy analysis showed significant effect on up-or downgrading of tumors. We conclude that selective use of a simple biomarker panel greatly aids in endometrial biopsy diagnostics.

Study 3 was a collaborative effort between Bern and Karolinska University Hospital we sought to implement a surrogate molecular marker in 604 patients with endometrial carcinoma in order to try to replicate the TCGA (The cancer genome atlas) molecular markers classifiers and to hopefully aid in endometrial carcinoma diagnostics and prognostics. The panel consisted of IHC for p53 (to define a p53mut group) and MSH2, MSH6, PMS2, MLH1 (to define a MMRd, mismatch repair defective, group) and Sanger sequencing of the POLE gene (to define a POLEmut) group. We found that this simple surrogate molecular marker, using readily available methods, could reproduce the findings of the TCGA and render four prognostically different molecular categories. We could however not prove an added prognostic benefit over current risk assessment prognostic tools, nor any benefit in adding a surrogate molecular marker to current prognostic markers. Moreover only one of the molecular groups (p53 mut) proved to be statistically significant marker for predicting Progression free survival (PFS) and overall survival (OS). From this we concluded that in the setting of non-selected group of patients consisting mostly of low grade tumors proving significant differences between each of the molecular groups couldn’t be done since progression events were so rare and the group as a whole behaved indolently. Nonetheless there are clear benefits with the addition of genomic data for other purposes such as aid in suitable patient selection for adjuvant treatment.

In study 4 we conducted sequencing of exons 9-14 of the POLE gene in a cohort of 604 patients with endometrial carcinoma in order to more deeply analyze mutation events and the phenotype of POLE mutated endometrial carcinomas. We found mostly previously described hotspot mutations but also a substantial number of mutations with unknown significance and characterized POLEmut tumors as occurring in mainly younger women with nulliparity and consisting predominantly of endometrioid endometrial carcinomas. This study shows that when POLE mutations are clearly defined as either hotspot mutations and/or POLE mutations with a known hypermutated phenotype this patient group of POLEmut had very good outcomes with only 1 recurrence in 38 patients. The Cox regression couldn’t however show any significant difference between POLEmut and non POLEmut tumors, however we believe again this was because we are dealing with a low risk EC population where both recurrence events and POLE mutations are rare, and thus we believe we were underpowered.