Hazard ratios are used to report outcomes from clinical trials. In oncology, a hazard ratio (HR) is commonly used to estimate the treatment effect for survival endpoints such as overall survival (OS) and progression free survival (PFS). Both endpoints, in particular OS, are regarded as the gold standard for demonstrating clinical benefit in oncology trials and is the preferred criterion for Heath Technology assessment (HTA) agencies.
What is a hazard ratio?
A HR provides an estimated ratio of patients experiencing the event of interest over a period of time in the experimental group versus the control group of the study. Taking OS as an example, a HR for OS is calculated for each week by dividing the rate of patients dying in the experimental arm by the rate of patients dying in the control arm. A HR equal to 1 means equal efficacy of both control and experimental interventions. If the experimental intervention is better than the control, the HR is <1; if it is worse than the control then the HR is >1. As such, declaring that a certain product reduces the risk of dying by 45% is based on a HR of 0.55 and is a common way of reporting survival benefit.
Why use a hazard ratio?
The advantage of the HR over other ways of estimating the magnitude and directions of survival outcomes is that a HR summarises the treatment effect over the entire length of the trial whereas other measures (i.e. median survival and time point analysis) only compare survival time at one point. For this reason, superiority and non-inferiority trial designs are usually based on the HR and not on one-point-at-a-time measures.
Proportional hazards and non-proportional hazards
Due to the significance of the HR in communicating survival benefit, it is essential to note certain aspects that may impact the interpretation of HR results. Commonly, the HR is estimated from a Cox proportional hazards (PH) model. If a proportional hazard assumption is made (i.e., a HR constant over time), it implies that the relationship between the HRs in the control and experimental group is consistent over time. If a PH assumption is not met (i.e., a HR that is not constant over time or non-proportional hazard: non-PH), the HR may not be representative of the effect of the new treatment during the entire period. In this instance, the HR could lead to inaccurate interpretations of the treatment effect.
There are regression models and alternative measures that accommodate non-PH to estimate survival, but it is integral to test for non-PH first before applying them. However, the use of alternative measures for non-PH and PH-assumption testing varies widely across HTA agencies around the world, with more attention given by HTA agencies in cost-effectiveness driven markets.1
How do HTA agencies use hazard ratios?
HTA agencies make recommendations on the best use of new medicines and their willingness to pay for those medicines. Understanding the magnitude of difference in outcomes, such as OS, between a new treatment and the standard of care is therefore critical. Used correctly, a HR can be a powerful tool for doing so and so make an important contribution to the evidence base for a new product.
Source:
- Monnickendam, G, et al. “Measuring survival benefit in health technology assessment in the presence of Nonproportional hazards.” Value in Health 22.4 (2019): 431-438.