Analysis of Data from Open-Label Extension (OLE) Study

In the previous article, the open-label extension (OLE) study was discussed. The OLE study is usually designed as a separate study from the RCT (the parent study) with its own study protocol and a separate electronic data capture (EDC) system even though all participants in the OLE are rollovers from the parent study. 

When analyzing the data from the OLE study, the data from the parent study often needs to be considered or combined for the analysis. The data can be analyzed with three different baselines:

• Baseline at the beginning of the OLE study - the data from the OLE study will be analyzed separately from the parent study
• Baseline at the beginning of the parent study (at the randomization of the parent study (RCT)) – the data from the OLE study and the parent study are combined and delayed start analysis can be performed to look at the delayed effect or never catch up effect
• Baseline at the first dose of the active drug – to look at the long-term trajectory of safety and efficacy variables for the experimental drug. For participants who were in the active arm of the parent study, the baseline would be at the randomization; for participants who were in the placebo arm of the parent study, the baseline would be at the beginning of the OLE study

If the efficacy outcome is a continuous variable, the delayed start analysis can be performed with the combined data from the parent and the OLE studies. See previous discussions: 

• Demonstrating the disease-modifying effect through delayed start study design or delayed start analyses
• Randomized withdrawal design and delayed start design in rare disease clinical trials
• Randomized Start Design (RSD)

Here are some articles discussing the application of the delayed start analysis in this setting. One of the delayed start analysis approaches is to perform the non-inferiority test to see if the treatment difference observed at the end of the RCT is preserved at the end of the OLE. A non-inferiority margin is pre-defined, the Mixed Models for Repeated Measures (MMRM) method is used to analyze the combined data from the parent study (RCT) and the OLE study, and the results from MMRM analysis are compared to the non-inferiority margin.  

• Lynch et al (2022) Efficacy of Omaveloxolone in Friedreich’s Ataxia: Delayed-Start Analysis of the MOXIe Extension
• Liu-Seifert et al (2014) A Novel Approach to Delayed-Start Analyses for Demonstrating Disease-Modifying Effects in Alzheimer’s Disease
• Liu-Seifert (2015) Delayed-start analysis: Mild Alzheimer’s disease patients in solanezumab trials, 3.5 years 
• Vershuur (2019) Randomized Delayed-Start Trial of Levodopa in Parkinson's Disease - the protocol/SAP
• Cohen et al (2019) An open-label extension study todemonstrate long-term safety and efficacyof ABP 501 in patients with rheumatoidarthritis

Two potential outcomes from the delayed start analyses are meaningful: 

Never catch up:

the placebo group (or delayed start group) will never catch up with the experimental treatment group after switching to the experimental treatment in the OLE study - suggesting the importance of the early treatment with the experimental drug and potential disease-modifying effect. For example, Chapman et al (2015) performed the delayed start analysis using the data from the double-blind trial and the subsequent open-label extension study. The results depicted below indicated the 'never catch up' scenario where the patients in the placebo group were never able to catch up with the AIPI (an enzyme augmentation treatment) group in terms of lung density change from baseline. 

Placebo group catch-up after treatment switching:

The placebo group (or delayed start group) catch up with the experimental treatment group after switching to the experimental treatment in the OLE study - emphasizing the treatment effects observed in the RCT. For example, Rosich et al (2022) performed the delayed start analysis using the data from a double-blind trial and its OLE study for the drug galcanezumab in patients with chronic migraine. After switching to galcanezumab doses at the start of OLE study (at month 3), the previous placebo group experienced a rapid mean reduction of 6.8 migraine headache days within the first month, catching up with the previous double-blind galcanezumab groups by month 4, and then maintaining that reduction over time.

If the efficacy outcome is overall survival (time to death), there is usually an insufficient number of death events from the randomized, controlled parent study for running the log-rank test or Cox regression. However, combining the data from the RCT and the OLE study, there may be enough death events for performing meaningful survival analyses. When performing the survival analysis using the data from the RCT and the OLE study, the start of the OLE study can be considered as treatment crossover or treatment switching - a situation often occurs in oncology clinical trials. Different approaches have been proposed to handle the treatment crossover or treatment switching due to the transition from the RCT to OLE study. These approaches were summarized in EMA's documet "Question and answer on adjustment for cross-over in estimating effects in oncology trials" and in a presentation by Norbert Hollaender (2014) "Methods to estimate survival time after treatment switching in oncology – overview and practical considerations".

Compared to the traditional intention-to-treat approach, it is better to perform the overall survival analyses using more sophisticated methods (such as the rank preserving structural failure time (RPSFT) method and The Inverse Probability of Censoring Weighting (IPCW) to adjust for treatment crossover or treatment switching due to the transition from the RCT to OLE study.