Meta-analytical methods for estimating outcomes from overall response rate in patients with relapsed/refractory diffuse large B-cell lymphoma

Ling Wang, Hung Lam, Yaping Shou, Aaron Galaznik _

PDF  |  HTML  |  Order a Reprint

Oncotarget. 2019; 10:3285-3293. https://doi.org/10.18632/oncotarget.26904

Metrics: HTML 33 views  |   ?  


Ling Wang1,3, Hung Lam2, Yaping Shou1,4 and Aaron Galaznik1,5

1 Millennium Pharmaceuticals, Inc., a wholly owned subsidiary of Takeda Pharmaceutical Company Limited, Cambridge, MA, USA

2 MMS Holdings Inc., Canton, MI, USA

3 Currently employed at Pfizer, Inc.

4 Currently employed at Trillium Therapeutics Inc.

5 Currently employed at SHYFT Analytics, a Medidata Company

Correspondence to:

Aaron Galaznik,email: aaron.galaznik.wg04@wharton.upenn.edu

Keywords: diffuse large B-cell lymphoma; meta-analyses; outcomes; overall response rate; durable response rate

Received: August 24, 2018     Accepted: April 14, 2019     Published: May 14, 2019


Relapsed/refractory diffuse large B-cell lymphoma (DLBCL) is highly heterogeneous and current trials are investigating new approaches to improve outcomes. Limited data on response endpoints can confound estimation of a treatment effect when designing studies of novel agents in this setting, which can hinder study sample size calculations, especially if a net estimate is required for a ‘physician’s choice’ comparator arm. Here we estimate complete response rate (CRR), overall response rate (ORR), and extrapolate durable response rates (DRR; CR/partial response lasting ≥16 weeks) for such a comparator arm from published ORRs in DLBCL.

CRR, ORR, and DRR (if reported) were obtained from published clinical trials for approved single-agent therapies in patients with relapsed/refractory aggressive non-Hodgkin lymphoma after ≥2 prior therapies. Meta-analyses were performed to estimate CRR, ORR, and DRR based on ORR data reported from these studies.

Published data from studies of eight monotherapies were included. Meta-analyses using fixed and random effects models showed a pooled estimate for a CRR of 12% (95% confidence interval [CI]: 9−15) and 11% (95% CI: 8−15), respectively, an ORR of 30% (95% CI: 25−35) and 30% (95% CI: 24−36), respectively, and a DRR of 14% (95% CI: 11−18; same for fixed and random effects models). Bayesian meta-analysis estimated a pooled DRR of 14% (95% credible interval: 11−19).

CRR estimates for a physician’s choice comparator arm in patients with relapsed/refractory DLBCL were 11−12%; DRR estimates were 14% regardless of methodology. Lack of consistency in reported data and choice of endpoints can be addressed using meta-analytic approaches.

Creative Commons License All site content, except where otherwise noted, is licensed under a Creative Commons Attribution 3.0 License.
PII: 26904