PRM21 - Time Ratios Or Hazard Ratios: Accelerating Toward A New Approach?

Abstract

Syntheses of time-to-event data often rely on published hazard ratios (HR) therefore assuming proportional hazards (PH). Methods now exist to reconstruct individual patient data (IPD) from published Kaplan-Meier (KM) plots. This enables the PH assumption to be formally tested, and a time ratio (TR) to be estimated as an alternative measure of relative treatment effect. While TRs do not require the PH assumption, these are rarely used in evidence synthesis. We compared TRs with HRs to demonstrate their ease of interpretation and transparency. Relative treatment effect measures HRs and TRs were compared in terms of: scale, underlying assumptions, interpretation, availability in published literature and derivation. HRs act on the log-hazard scale representing the ratio of hazards for treatment vs. control. TRs act on the log-failure time scale, representing the ratio of failure times for treatment vs. control. A HR<1 represents a decrease in the event hazard whilst a TR<1 represents an acceleration in time-to-event, with treatment. The inverse of the TR, referred to as the acceleration factor (AF), therefore represents the same direction of benefit as for HRs. For TR=2, AF=0.5, the time-to-event for treatment is two times that for control. HRs are commonly reported to summarise the average relative treatment effect. TRs are rarely reported and hence less familiar to clinicians, but can be validated against any ratio of published survival times. Relative treatment effects expressed as TRs are not commonly utilised in evidence synthesis due to unfamiliarity around interpretation and lack of availability in the published literature. When converted to AFs, TRs have a similar interpretation to HRs and can be estimated from IPD and published KM data. Therefore, evidence synthesis should no longer rely on published HRs as the measure of relative treatment effect when the PH assumption is violated.