Progression-free survival by local investigator versus independent central review: Comparative analysis of the AGO-OVAR16 Trial

Highlights

• Pazopanib maintenance therapy extended PFS in patients with AEOC.

• HR estimates for PFS by investigator were consistent with those of central review.

• There was no evidence of investigator bias in estimates of disease progression.

Abstract

Background

Analysis of progression-free survival (PFS) as the primary endpoint in advanced epithelial ovarian, fallopian tube, and primary peritoneal cancer (AEOC) trials may be confounded by the difficulty of radiologic evaluation of disease progression and the potential for discrepancy between investigator and blinded independent central assessments. PFS as assessed by local investigator (INV) was the primary endpoint of AGO-OVAR16, a randomized, double-blind trial of pazopanib maintenance therapy in AEOC. To confirm the robustness of the primary analysis, PFS was also evaluated by blinded independent central review (BICR).

Methods

Patients with histologically confirmed AEOC (N = 940) were randomized 1:1 to receive pazopanib 800 mg/day or placebo for up to 24 months. Tumor response in the intent-to-treat population was evaluated by CT/MRI every 6 months and analyzed per RECIST 1.0.

Results

Pazopanib prolonged PFS versus placebo by INV (median 17.9 vs 12.3 months; hazard ratio [HR] = 0.766, 95% confidence interval [CI]: 0.643–0.911; P = 0.0021). Results for PFS by BICR were similar (median 15.4 vs 11.8 months; HR = 0.802, 95% CI: 0.678–0.949; P = 0.0084). Progression events were recorded later by INV in 23% of pazopanib-treated patients and 17% of placebo-treated patients. The overall concordance between INV and BICR assessments was 84% and 86% in the pazopanib and placebo arms, respectively.

Conclusions

By INV and BICR assessments, maintenance therapy with pazopanib in AEOC provided a significantly longer PFS than placebo. The good overall concordance between INV and BICR assessments, as well as HR and P value consistency, supports the reliability of investigator-assessed PFS as the primary endpoint in AGO-OVAR16.

Introduction

Ovarian cancer is among the most common cancers in women, and has the highest mortality among all gynecologic cancers [1]. First-line treatment includes debulking surgery followed by platinum-taxane–based chemotherapy [2]. Although the initial response rate is high, most patients experience recurrence and eventually die from the disease [3]. Recent studies have demonstrated that addition of the angiogenesis inhibitor bevacizumab to the chemotherapy regimen can delay recurrence of advanced ovarian cancer [4], [5]. Pazopanib, an oral angiogenesis inhibitor targeting VEGF receptors, platelet-derived grown factor receptors, and c-kit, has demonstrated promising activity in ovarian cancer [6], [7]. The phase 3, randomized, placebo-controlled trial AGO-OVAR16 (OVAR16; NCT00866697) evaluated pazopanib monotherapy as maintenance therapy in patients with advanced epithelial ovarian, fallopian tube, or primary peritoneal cancer (AEOC) whose disease had not progressed following first-line treatment [8].

The primary endpoint of OVAR16 was progression-free survival (PFS), and the primary analysis was based on radiologic progression per Response Evaluation Criteria in Solid Tumors version 1.0 (RECIST) as assessed by the local investigator (INV).

In ovarian cancer, PFS is considered to be an appropriate primary endpoint because disease progression is associated with recurrence of symptoms and relatively short survival, and further treatment is usually initiated, which brings additional toxicities [5], [9], [10]. However, assessment of PFS is sensitive to timing of evaluations and requires consistent interpretation of radiologic progression; variability in timing and interpretation between the control and treatment arms can introduce bias to the analysis [11]. In addition, in ovarian cancer the radiologic assessment of progression is particularly difficult because: 1) upfront standard surgery and chemotherapy are generally able to remove most, if not all, of the tumor, so most patients are disease-free at the end of their initial therapy; 2) surgical scarring is often difficult to distinguish from actual tumor lesions; and 3) in many patients progression may be characterized by diffuse peritoneal implantation and the prevalence of non-target lesions, such as ascites. To avoid investigator bias, regulatory agencies commonly require an additional analysis by blinded independent central review (BICR) to confirm the reliability of results obtained by INV evaluation of disease progression. Recent studies have suggested that independent review of all scans, which is associated with increased cost and complexity, may not be necessary. A meta-analysis of 27 phase 3 oncology trials revealed good concordance between INV and BICR assessments of PFS, and little evidence of investigator bias [12]. The United States Food and Drug Administration (US FDA) has discussed the use of an independent audit of random sample assessments rather than a complete review of all patients [13]. In this report we compare the primary endpoint of OVAR16 as assessed by INV and BICR, and explore the use of a subset of patient evaluations to confirm the primary analysis.