RECONSTRUCTION OF INDIVIDUAL PATIENT DATA BASED ON PUBLISHED KAPLAN-MEIER CURVES- CASE OF REGORAFENIB FOR COLORECTAL CANCER
Author(s)
Ali AA, Adunlin G, Xiao H, Diaby V
Florida A & M University, Tallahassee, FL, USA
Presentation Documents
OBJECTIVES: To conduct pharmacoeconomic analyses, both cost and effectiveness data are required. Randomized controlled trials (RCTs) are often used as a source of efficacy data. As RCTs are often of short duration, efficacy data need to be extrapolated beyond the trial follow-up period to be fit for use in model-based pharmacoeconomic evaluations. However, RCTs usually report effectiveness data in terms of Kaplan-Meier (KM) estimates. As a result, researchers need to reconstruct individual patient data (IPD) from published KM curves of trials’ treatment arms to estimate their long-term effects. This study aims to reconstruct the survival benefits of regorafenib monotherapy for previously treated metastatic colorectal cancer (CORRECT): an international, multicentre, randomized, placebo-controlled, phase 3 trial. METHODS: An algorithm developed by Guyot and colleagues was adopted to reconstruct IPD using R statistical package, based on the overall survival KM curves of the CORRECT trial. The reconstruction of IPD included the following steps: 1) extraction of coordinates from published KM curves, 2) data accuracy check, 3) creation of a second dataset, and application of the algorithm. The results of the original trial were compared to the reconstructed data using graphical and quantitative methods, for validation purposes. RESULTS: Based on the IPD reconstruction, 162 and 88 events occurred in the regorafenib and placebo groups respectively. The median overall survival time in the regorafenib arm was 6.5 months (95% CI 5.83, 8.43) which is about the same as the original trial (6.4 months). In the placebo group, the median overall survival for the reconstruction data was 5.09 months (95% CI 4.30, 6.81) compared to the trial median survival of 5.0 months. CONCLUSIONS: The results of this study can be utilized to estimate transition probabilities for model-based pharmacoeconomic evaluations in the absence of individual patient data (IPD).
Conference/Value in Health Info
2015-05, ISPOR 2015, Philadelphia, PA, USA
Value in Health, Vol. 18, No. 3 (May 2015)
Code
PRM52
Topic
Methodological & Statistical Research
Topic Subcategory
Modeling and simulation
Disease
Oncology