Comparison of binary efficacy endpoints in 11 North Central Cancer Treatment Group phase II metastatic breast cancer clinical trials
Abstract Abstract #6147 Background: Phase II metastatic breast cancer (MBC) clinical trials evaluating efficacy of cancer treatments are often designed using a binary primary endpoint (i.e., each evaluable patient [pt] is classified as a “success” or “failure”). In the era of novel agents in cancer...
Saved in:
Published in | Cancer research (Chicago, Ill.) Vol. 69; no. 2_Supplement; p. 6147 |
---|---|
Main Authors | , , , , , , |
Format | Journal Article |
Language | English |
Published |
15.01.2009
|
Online Access | Get full text |
Cover
Loading…
Summary: | Abstract
Abstract #6147
Background: Phase II metastatic breast cancer (MBC) clinical trials evaluating efficacy of cancer treatments are often designed using a binary primary endpoint (i.e., each evaluable patient [pt] is classified as a “success” or “failure”). In the era of novel agents in cancer research, endpoints such as 6-month progression-free survival [PFS6] for measuring efficacy of cytostatic agents are more commonly being used. This meta-analysis was undertaken to compare two binary classifications of PFS6 and to compare these binary endpoints with other efficacy endpoints in the phase II setting.
Material and Methods: All closed North Central Cancer Treatment Group (NCCTG) phase II MBC clinical trials using Response Evaluation Criteria in Solid Tumors (RECIST) with at least 1 year of follow-up since last pt accrued were selected. All eligible pts initiating treatment were included. Two binary classifications of PFS6 were computed for each trial. Success for PFS6-1 is defined as on study treatment 6 months from registration without documentation of disease progression. Success for PFS6-2 does not require a pt to be on study treatment at 6 months. Also computed for each trial are Kaplan-Meier (KM) estimates of PFS6 (PFS6-KM) and 1-year overall survival (OS1-KM). Trial-level endpoints were summarized using descriptive statistics and compared using weighted (by trial sample sizes) Pearson correlations. Lastly, the concordance rate of PFS6-1 and PFS6-2 status with OS status at 1 year at the pt level was computed across all pts (pts censored for OS prior to one year were excluded [n=10]).
Results: 11 trials met inclusion criteria. All trials required measurable disease and had a single arm. 485 evaluable pts were accrued (median 48 pts per trial [range 19-77]). Median PFS6-1 was 27% (range 10-44%) and median PFS6-2 was 34% (range 10-73%). The median trial-level difference between PFS6-1 and PFS6-2 was 5% (range 0-43%). The correlation between PFS6-1 and PFS6-2 was 0.81 (p<0.01). Among the endpoints, PFS6-2 and PFS6-KM had the highest correlation (>0.99, p<0.01) due to only 2 pts being censored for PFS prior to 6 months. Among the PFS endpoints, OS1-KM was most highly correlated with PFS6-1 (0.79, p<0.01) with the correlations with PFS6-2 and PFS6-KM not being statistically different from zero (both 0.59 with p>0.05). However, overall patient-level concordance between PFS6 status and OS status at 1 year was higher using PFS6-2 (68%) than PFS6-1 (59%).
Discussion: Differences were observed between the two binary classifications of PFS6. PFS6 with (as compared to without) the requirement that a pt be on study treatment at 6 months appears to have higher correlation with OS at 1 year at the trial level but lower concordance with OS status at 1 year at the pt level. Selection of the historical control should take into consideration the definition of PFS6 being used.
Citation Information: Cancer Res 2009;69(2 Suppl):Abstract nr 6147. |
---|---|
ISSN: | 0008-5472 1538-7445 |
DOI: | 10.1158/0008-5472.SABCS-6147 |