Baseline Serum HE4 But Not Tissue HE4 Expression Predicts Response to the Levonorgestrel-Releasing Intrauterine System in Atypical Hyperplasia and Early Stage Endometrial Cancer

The levonorgestrel-releasing intrauterine system (LNG-IUS) is a conservative management option for atypical hyperplasia (AH) and low grade early stage endometrial cancer (EEC), but around 1 in 3 patients fail to respond to treatment. The aim of this study was to investigate if serum and/or tissue HE4 expression could predict response to LNG-IUS therapy. Patients with AH or presumed Stage I EEC had serum and endometrial samples taken at baseline and at 3-month intervals over 12 months post-insertion of LNG-IUS. 74 patients were recruited and baseline demographics recorded. Of 57 patients for whom response was histologically determinable, 39 (68%) were responders and 18 (32%) non-responders. Mean baseline serum HE4 was significantly lower in responders (62.1 ± 1.1 pM, 95% confidence interval (CI) 52.7-73.2), compared to non-responders (125.6 ± 1.3 pM, 95% CI 74.5-211.7, p = 0.014), including when considering age, BMI, menopausal status, smoking status, and histological grade as covariables (p = 0.005). Baseline tissue HE4 expression was not significantly different in responders compared to non-responders (p = 0.999). Responders showed a significant mean reduction (-9.8 ± 3.4%, 95% CI -16.7 to -2.8%, p = 0.008) in serum HE4 between baseline and 3 months (p = 0.008), whereas non-responders showed no significant change (p = 0.676). Neither responders nor non-responders showed a significant percentage change in serum HE4 from baseline beyond 3 months (p > 0.05). Change in serum HE4 between baseline and 3 and 6 months and tissue HE4 tissue expression between baseline and 3, 6, and 12 months was not significantly different in responders compared to non-responders (p > 0.05). This study suggests that baseline serum HE4, but not baseline tissue HE4 expression, is independently predictive of response to the LNG-IUS and could be used to guide management decisions.