Meta-Analysis of 49 Roche Oncology Trials Comparing Blinded Independent Central Review (BICR) and Local Evaluation to Assess the Value of BICR

Progression-free survival assessment by local investigators versus blinded independent central review in randomized clinical trials in metastatic breast cancer: A systematic review and meta-analysis

INTRODUCTION

In randomized clinical trials (RCTs), blinded independent central review (BICR) is used to minimize heterogeneity and bias associated with radiological response evaluation by local investigators. However, BICR adds costs and complexity to the trial management. We assessed the discrepancy index between progression-free survival (PFS) assessment by local investigators and by BICR in RCTs conducted in patients with metastatic breast cancer (MBC).

METHODS

A systematic search of PubMed, Embase, Cochrane databases and conference proceedings (ASCO, SABCS, ESMO) was performed up to January 4, 2023 (PROSPERO: CRD42021229865). All RCTs published from 2000 to 2022, including MBC patients treated in first- or second-line, and reporting PFS assessed by local investigators and BICR were included. A discrepancy index between BICR-assessed and investigator-assessed HR was calculated for each trial and an overall combined DI was obtained using a fixed-effects model. The agreement between hazard ratios (HR) of PFS assessed by local investigators and BICR was measured using intraclass correlation coefficient (ICC).

RESULTS

We analyzed 24 studies including 13,168 patients. Among them, 19 (79%) were in first-line, 18 (75%) were phase III trials and 23 (96%) had PFS as primary endpoint. The overall combined discrepancy index was 0.97 (95%CI 0.85-1.10; ICC 0.831, p < 0.001) suggesting no statistically significant difference in PFS assessment between local investigators and BICR. This result was consistent across all analyzed subgroups.

CONCLUSIONS

The good concordance between local investigator and BICR assessments supports the reliability of local investigator-assessed PFS as primary endpoint for RCTs in MBC and questions the practical utility of implementing BICR in all RCTs.

Detection bias in open-label trials of anticancer drugs: a meta-epidemiological study

OBJECTIVES

In anticancer clinical trials, particularly open-label trials, central reviewers are recommended to evaluate progression-free survival (PFS) and objective response rate (ORR) to avoid detection bias of local investigators. However, it is not clear whether the bias has been adequately identified, or to what extent it consistently distorts the results. Therefore, the objective of this study was to evaluate the detection bias in oncological open-label trials by confirming whether local investigators overestimate the PFS and ORR compared with the findings of central reviewers.

DESIGN

Meta-epidemiological study.

DATA SOURCES

MEDLINE via PubMed from 1 January 2010 to 30 June 2021.

ELIGIBILITY CRITERIA FOR SELECTING STUDIES

Open-label, parallel-group superiority, randomised trials of anticancer drugs that adjudicated PFS or ORR by both central reviewers and local investigators.

REVIEW METHODS

We assessed the values for the same outcome (PFS and ORR) adjudicated by both central reviewers and local investigators. A random-effects model was used to estimate the ratio of HR (RHR) for PFS and the ratio of OR (ROR) for ORR between central reviewers and local investigators. An RHR lower than 1 and an ROR higher than 1 indicated an overestimation of the effect estimated by local investigators.

RESULTS

We retrieved 1197 records of oncological open-label trials after full-text screening. We identified 171 records (PFS: 149 records, ORR: 136 records) in which both central reviewers and local investigators were used, and included 114 records (PFS: 92 records, ORR: 74 records) for meta-analyses. While the RHR for PFS was 0.95 (95% CI 0.91 to 0.98), the ROR of ORR was 1.00 (95% CI 0.91 to 1.09). The results remained unchanged in the prespecified sensitivity analysis.

CONCLUSIONS

This meta-epidemiological study found that overestimation of local investigators has a small impact on evaluating PFS and ORR in oncological open-label trials. However, a limitation of this study is that it did not include data from all trials; hence, the results may not fully evaluate detection bias. The necessity of central reviewers in oncological open-label trials needs to be assessed by further studies that overcome this limitation.

TRIAL REGISTRATION NUMBER

CTR-UMIN000044623.

Optimizing disease progression assessment using blinded central independent review and comparing it with investigator assessment in the PRIMA/ENGOT-ov26/GOG-3012 trial: challenges and solutions

OBJECTIVE

Progression-free survival is an established clinically meaningful endpoint in ovarian cancer trials, but it may be susceptible to bias; therefore, blinded independent centralized radiological review is often included in trial designs. We compared blinded independent centralized review and investigator-assessed progressive disease performance in the PRIMA/ENGOT-ov26/GOG-3012 trial examining niraparib monotherapy.

METHODS

PRIMA/ENGOT-ov26/GOG-3012 was a randomized, double-blind phase 3 trial; patients with newly diagnosed stage III/IV ovarian cancer received niraparib or placebo. The primary endpoint was progression-free survival (per Response Evaluation Criteria in Solid Tumors [RECIST] v1.1), determined by two independent radiologists, an arbiter if required, and by blinded central clinician review. Discordance rates between blinded independent centralized review and investigator assessment of progressive disease and non-progressive disease were routinely assessed. To optimize disease assessment, a training intervention was developed for blinded independent centralized radiological reviewers, and RECIST refresher training was provided for investigators. Discordance rates were determined post-intervention.

RESULTS

There was a 39% discordance rate between blinded independent centralized review and investigator-assessed progressive disease/non-progressive disease in an initial patient subset (n=80); peritoneal carcinomatosis was the most common source of discordance. All reviewers underwent training, and as a result, changes were implemented, including removal of two original reviewers and identification of 10 best practices for reading imaging data. Post-hoc analysis indicated final discordance rates between blinded independent centralized review and investigator improved to 12% in the overall population. Median progression-free survival and hazard ratios were similar between blinded independent centralized review and investigators in the overall population and across subgroups.

CONCLUSION

PRIMA/ENGOT-ov26/GOG-3012 highlights the need to optimize blinded independent centralized review and investigator concordance using early, specialized, ovarian-cancer-specific radiology training to maximize validity of outcome data.