Tumor Growth Rate Estimates Are Independently Predictive of Therapy Response and Survival in Recurrent High-Grade Serous Ovarian Cancer Patients

Abstract

This study aimed to assess the predictive value of tumor growth rate estimates based on serial cancer antigen-125 (CA-125) levels on therapy response and survival of patients with recurrent high-grade serous ovarian cancer (HGSOC). In total, 301 consecutive patients with advanced HGSOC (exploratory cohort: n = 155, treated at the Medical University of Vienna; external validation cohort: n = 146, from the Ovarian Cancer Therapy-Innovative Models Prolong Survival (OCTIPS) consortium) were enrolled. Tumor growth estimates were obtained using a validated two-phase equation model involving serial CA-125 levels, and their predictive value with respect to treatment response to the next chemotherapy and the prognostic value with respect to disease-specific survival and overall survival were assessed. Tumor growth estimates were an independent predictor for response to second-line chemotherapy and an independent prognostic factor for second-line chemotherapy use in both univariate and multivariable analyses, outperforming both the predictive (second line: p = 0.003, HR 5.19 [1.73-15.58] vs. p = 0.453, HR 1.95 [0.34-11.17]) and prognostic values (second line: p = 0.042, HR 1.53 [1.02-2.31] vs. p = 0.331, HR 1.39 [0.71-2.27]) of a therapy-free interval (TFI) < 6 months. Tumor growth estimates were a predictive factor for response to third- and fourth-line chemotherapy and a prognostic factor for third- and fourth-line chemotherapy use in the univariate analysis. The CA-125-derived tumor growth rate estimate may be a quantifiable and easily assessable surrogate to TFI in treatment decision making for patients with recurrent HGSOC.

Keywords: growth rate; ovarian cancer; platinum-resistant; recurrence; therapy response.

Figure 1

(a) The plotted two-phased mathematical equation $\mathrm{f}\left(\mathrm{t}\right)=\exp \left(-\mathrm{d}\times \mathrm{t}\right)+\exp \left(\mathrm{g}\times \mathrm{t}\right)-1$ was proposed by Stein et al. to reflect the sum of decay and subsequent growth of a recurrent solid tumor under therapy. An initial logarithmic decay may shrink the relative tumor size to subclinical dimensions. As soon as cell growth outnumbers decay, the tumor may re-grow to a clinical recurrence. The assumed tumor growth rate (g) constitutes the primary objective variable of the present study. (b) Plotted CA125-course of patient no. 164 throughout primary and recurrence therapy of a high-grade serous ovarian cancer high-grade serous epithelial ovarian cancer (HGSOC) patient undergoing platinum-based chemotherapy. The black lines demonstrate the respective growth rate estimates (g1 and g2) calculated applying the two-phased mathematical model, which were further evaluated for their predictive and prognostic values.

Figure 2

Flowchart depicting therapy courses and respective overall response rates (ORR) in comparison with platinum-based chemotherapy and bevacizumab application for both exploratory and validation cohort. Dashed boxes describe the survival status of patients who did not undergo a subsequent therapy line for any reason at a 5-year overall survival follow-up. The endpoint “death” equals patient death for any reason, including intercurrent disease. Numbers are given absolute and as percentages. n/a—data not available; ORR—overall response rate (complete and partial response).

Figure 3

The area under the curve (AUC) receiver operating characteristic analysis to assess the predictive value of the logarithm of the growth rate constant in patients with recurrent HGSOC with the endpoint of complete or partial response to 2nd-line chemotherapy. Youden’s J-statistics was performed to define −3.32 as an optimal cut-off.

Figure 4

Kaplan–Meier curves depicting overall survival at the timepoint of secondary cytotoxic chemotherapy broken down by tumor growth rate estimates at an optimal cut-off of ≤/>−3.32 in both exploratory and validation cohorts.