Investigators Propose Different Measures for Biosimilar Equivalence

May 12, 2020
Deana Ferreri, PhD

Investigators use the HERITAGE trial findings to argue for a more precise way of measuring clinical equivalence for biosimilars.

Choosing appropriate clinical end points and analysis procedures is essential to both establish clinical equivalence of biosimilars and convince physicians and patients that the agents are worthy of use. In a recent study, investigators use the HERITAGE study to argue for a more accurate system for measuring biosimilar equivalence.

There has been much discussion of moving away from hazard ratios (HRs), which measure risk of disease progression or death events between regimens or drugs, toward an analysis known as restricted mean survival time (RMST), which directly compares the number of these events over specified time intervals.

The authors of the study propose that RMST would provide more meaningful and interpretable information than HR.

The authors noted that studies of anticancer drugs typically use a time-to-event indicator, such as progression-free survival (PFS) or overall survival (OS), as the primary measure of efficacy. However, studies assessing clinical equivalence of biosimilars to those reference products instead use the overall response rate (ORR) at a specific time point, an HR-based measure.

The study details the authors’ analysis of reconstructed OS and PFS data from the HERITAGE trial using RMST.

The HERITAGE trial assessed equivalence of a trastuzumab biosimilar plus taxane compared with the reference product plus taxane in 458 patients with metastatic breast cancer. The primary end point was 24-week ORR, and disease events and deaths were recorded through 48 weeks. The authors reanalyzed the data on disease progression events and deaths as PFS and OS using RMST.

Reanalysis of HERITAGE Trial

In the original study, there were nonsignificant differences in disease progression events and deaths between groups. However, the CIs, according to the authors of the new study, were quite wide, leading them to doubt that there were no true differences between groups. “For PFS, the hazard from the biosimilar group might be 28% higher than that from the reference product,” they wrote. “Consequently, even if the biosimilar product is approved by the regulatory agencies, clinicians and patients may not be convinced to adopt the biosimilar as an alternative to the reference product.”

Reanalyzing using RMST, the time to disease progression or death was 38.8 weeks in the biosimilar group compared with 37.5 weeks for the reference product, a difference of 1.3 (95% CI, −1.3 to 3.8) weeks. The RMST difference for OS between the biosimilar and reference product was 0.7 (95% CI, −0.7 to 2.1) weeks. For the biosimilar, these findings demonstrate maximum differences in PFS and OS of just 10.1% and 5%, respectively, both reasonably-sized CIs, according to the authors.

“Unlike the results from the hazard ratio, these time-scaled small differences in OS and PFS…provide a clinically meaningful way to evaluate whether the biosimilar is clinically equivalent to trastuzumab over a 48-week time window,” the authors wrote.

The authors asserted that expressing the results as the number of weeks before a progression event in each group is more easily interpretable than describing the percent difference in events as a function of time using HR.

The Problem With HRs

The authors explained that the use of HR requires an assumption that each group being compared has a constant rate of events over time. However, in oncology biosimilar equivalence trials, the rate is not necessarily constant; the number of disease progression events or deaths during the study period is often limited. RMST, on the other hand, requires no assumption of a constant rate.

Furthermore, the precision of HRs, the authors said, depends on the number of events rather than the duration of the study, meaning that sufficiently precise HRs to determine clinical equivalence in many cases will require an unreasonably large sample size, and a longer study does not necessarily mean more precision.

Advantages of RMST

RMST requires no assumption of proportional hazards (constant rates of events). With a time-to-event measure such as RMST, a change in the rate of events over time will not violate the underlying assumptions of the model.

The authors concluded that RMST-based methods “provide a clinically interpretable between-group difference with feasible study size and study duration compared with the HR-based approach.”

Finally, they suggest the use of RMST-based methods for postmarket surveillance, because “for most current biosimilar trials, one may argue that the patients’ exposure times might be too short to assess PFS and OS and long-term safety.”


Uno H, Schrag D, Kim DH, et al. Assessing clinical equivalence in oncology biosimilar trials with time-to-event outcomes. JNCI Cancer Spectr. 2019;3(4):pkz058. doi:10.1093/jncics/pkz058

Related Content:

News | Oncology | Development