Quantification of Treatment Effect Modification on Both an Additive and Multiplicative Scale

Area of Science:

• Biostatistics
• Clinical Epidemiology
• Cardiovascular Surgery

Background:

• Survival analyses commonly use the Cox model (multiplicative scale) to assess treatment effects.
• Clinicians and researchers are increasingly interested in evaluating the absolute benefits of treatments.
• Older patients may show reduced relative treatment effects but potentially similar or greater absolute benefits due to higher baseline event rates.

Purpose of the Study:

• To compare the assessment of treatment effect and effect modification on both multiplicative and additive scales.
• To investigate how age influences treatment effects in coronary surgery using different statistical models.
• To determine the most appropriate scale for evaluating treatment benefits, particularly in older populations.

Main Methods:

• Propensity score adjustment was used for analysis.
• Treatment effect was assessed using a multiplicative hazard model (Cox model).
• Treatment effect modification was assessed using an additive hazard model.

Main Results:

• The multiplicative model indicated a lower relative hazard reduction in older patients (HR for interaction/year = 1.03).
• The additive model showed a similar absolute hazard reduction with increasing age (Delta for interaction/year = 0.10).
• Number needed to treat was similar for younger (< = 60) and older (>70) patients at follow-up end.

Conclusions:

• A lower relative treatment effect in older patients can represent a similar absolute treatment effect due to baseline hazard differences.
• Absolute risk reduction can be calculated from multiplicative survival models.
• The study advocates for increased use of the absolute scale, particularly additive hazard models, for assessing treatment effects and modifications.