Real-World Evidence: Successful Cases Using Synthetic Control Methods

Single-arm studies using real-world data as external controls (synthetic control methods)

For single-arm clinical trials, it is possible to use an external control with real-world data. This is especially helpful in trials for rare disease where recruitment is difficult due to the scarcity of cases. Using real-world data as a control is also helpful for trials of life-threatening diseases that lack effective treatment. 

External controls can be divided into two main groups. The first type, historical controls, are based on real-world data obtained prior the trial. The second type, parallel controls, use data from the same period as the single-arm trial. 

Using an external control does not fully follow a random allocation method, so there are some limitations on this model. These limitations can be overcome by ensuring the data collected meet applicability standards for real-world data. This can be achieved by giving priority to parallel external controls and by using appropriate statistical analysis methods such as preference scoring matching. The effect of confounding factors and model assumptions can be assessed using sensitivity analysis and other quantitative methods to control bias. 

Synthetic Control Arm Trials

HLT has conducted several studies using synthetic control methods in different therapeutic areas in recent years.

FRESCO-Hybrid

In a phase III trial (FRESCO), fruquintinib demonstrated acceptable safety and better efficacy than placebo. Based on these results, fruquintinib was successfully launched as a third-line treatment for metastatic colorectal cancer (mCRC). 

Several new drugs were launched after fruquintinib. However, there are no head-to-head randomized trials comparing third-line treatments for mCRC. In our study, we used real-world data as an external active control arm to compare the effectiveness of third-line treatments for mCRC. The external control arm was comprised of authorized data collected between July 2014 and February 2020 from 10 hospitals in China. The study used the same inclusion/exclusion criteria as the FRESCO study to identify other tyrosine kinase inhibitors as the control group. 

Mahalanobis distance matching with propensity score caliper was used to adjust for the unbalanced confounding factors. After matching, Kaplan Meier scores and log-rank tests was applied to evaluate progression-free survival (the primary endpoint) between the two groups. 

This research was presented at the 2020 CSCO Conference. The FRESCO hybrid study provided timely evidence on the comparative effectiveness of third-line treatments for mCRC. Summary statistics, equilibrium tests of key baseline confounding covariates, and median progression-free survival results were reported.