Impact of Treatment Effect Heterogeneity on the Estimation of Individualized Treatment Rules for Count Outcomes

Abstract

There is growing interest in tailoring treatment decisions to individual patient characteristics, but few studies have examined the implementation and performance of individualized treatment rules (ITRs) for count data. Our objective was to compare ITR methods in randomized trials with count outcomes and explore the impact of sample size and distribution of heterogeneity of treatment effect (HTE) on the validity of treatment recommendations. We conducted a simulation study where patients were randomized to receive one of two treatments and created five responder strata to reflect different HTE scenarios. Various ITR methods were used to estimate treatment effects and were evaluated in terms of value function and accuracy. We also conducted a case study involving patients with multiple sclerosis. All ITR methods performed better under favorable conditions such as larger sample size, greater treatment heterogeneity, or fewer neutral patients (also known as equivalent treatments effects), but they were outperformed by fixed treatment strategies with smaller sample sizes or limited HTE. However, larger sample sizes can compensate ITRs for smaller HTEs and high HTEs can compensate ITRs for limited data. In the case study, we identified HTE and developed a tree-based ITR that outperformed fixed treatment recommendations. In conclusion, ITR performance can be influenced by sample size and the distribution of HTE, as well as their interactions. Simulation scenarios, informed by clinical insights, can help us determine if HTE estimation is feasible and, if so, identify the most effective ITR.