Identifying biomarkers of response to progestin therapy in conservatively managed endometrial cancer and atypical hyperplasia/endometrioid intraepithelial neoplasia (266)

Abstract

<b>Objectives:</b> The incidence of endometrial cancer (EC) and its precursor lesion, atypical hyperplasia/endometrioid intraepithelial neoplasia (AH/EIN), continues to increase among young women. In women desiring fertility, identifying clinical-pathologic markers predictive of response to progestin-based therapy (PT) is essential. Loss of AT-Rich Interactive Domain 1A (<i>ARID1A</i>) activates the PI3K/ AKT/mTOR signaling pathway leading to downregulation of progesterone receptor (PR) expression, resulting in primary progesterone resistance in EC cell lines. We propose that <i>ARID1A</i> deficient tumors, which will have relatively higher phosphorylated AKT (p-AKT) and lower PR expression than <i>ARID1A</i> proficient tumors, will have a lower response rate to PT. <b>Methods:</b> Patients with histologically proven AH/EIN or well-differentiated EC who were treated with PT at an academic center were retrospectively reviewed. Demographic and clinical data were abstracted from the medical record. For each patient, all endometrial biopsy specimens were evaluated, which included the initial diagnostic and any follow-up biopsy. A tissue microarray was constructed of all biopsy specimens, and immunohistochemical stains for <i>ARID1A</i>, <i>PTEN</i>, p-AKT, PR, estrogen receptor (ER), and mismatch repair protein expression were evaluated. Complete response (CR) to treatment was defined as resolution of EC or AH/EIN. Cox Proportional Hazards models were used to estimate the relative hazard (HR) of complete response to progestin by select patient characteristics. Kaplan-Meier survival curves were also estimated for patient characteristics of interest. Trends in gene expression over the follow-up time were assessed using linear regression fitted by generalized estimating equations. Analyses were conducted in R version 4.0.2. <b>Results:</b> A total of 42 women were treated with PT (12 EC and 30 EIN). The median age was 35 years, and the median BMI was 35.9 kg/m². The response rate was 81%. There were no clinical or pathological differences between responders and non-responders. In addition, time to response was not impacted by age, BMI, histology, or investigated immunohistochemical markers. There was no correlation between PR/ER (r_S = 0.29, p=0.09), <i>ARID1A</i>/ER (r_S = -0.26, p=0.13), and <i>ARID1A</i>/PR (r_S = 0.13, p=0.45) expression. On multivariable analysis, <i>ARID1A</i>, <i>PTEN</i>, ER, nor PR expression were associated with CR. However, age at diagnosis (HR: 0.84, p=0.02), obesity (HR: 0.15, p=0.02), and EC (HR: 0.17, p=0.01) were seen to have a reduced probability of response on multivariable analysis. Over time from diagnosis, ER and PR expression decreased, and <i>ARID1A</i> expression was lost in both responders and non-responders. Furthermore, the decrease over time for PR and <i>ARID1A</i> expression occurred at a faster rate for responders. <b>Conclusions:</b> In human endometrial samples taken at diagnosis, contrary to preclinical work, there were no potential biomarkers or clinical characteristics identified to be associated with improved response rate or time to respond.