Individual treatment effect prediction for amyotrophic lateral sclerosis patients

Abstract

A treatment for a complicated disease might be helpful for some but not all patients, which makes predicting the treatment effect for new patients important yet challenging. Here we develop a method for predicting the treatment effect based on patient characteristics and use it for predicting the effect of the only drug (Riluzole) approved for treating amyotrophic lateral sclerosis. Our proposed method of model-based random forests detects similarities in the treatment effect among patients and on this basis computes personalised models for new patients. The entire procedure focuses on a base model, which usually contains the treatment indicator as a single covariate and takes the survival time or a health or treatment success measurement as primary outcome. This base model is used both to grow the model-based trees within the forest, in which the patient characteristics that interact with the treatment are split variables, and to compute the personalised models, in which the similarity measurements enter as weights. We applied the personalised models using data from several clinical trials for amyotrophic lateral sclerosis from the Pooled Resource Open-Access Clinical Trials database. Our results indicate that some amyotrophic lateral sclerosis patients benefit more from the drug Riluzole than others. Our method allows gradually shifting from stratified medicine to personalised medicine and can also be used in assessing the treatment effect for other diseases studied in a clinical trial.

Survival forests under test: Impact of the proportional hazards assumption on prognostic and predictive forests for amyotrophic lateral sclerosis survival

“…In this case,θ is never explicitly computed; instead, the non-parametric maximum-likelihood estimator for the unconditional survivor function (Ŝ N , for example, Kaplan-Meier or Breslow) is used: a NP (t) θ = cloglog(1 −Ŝ N (t)), where cloglog is the complementary log-log function. In a parametric setting, Weibull (W) models with basis functions a W (t) = (1, log(t)) were studied (Seibold et al 2018). The corresponding log-cumulative hazard function a(t) ϑ = ϑ 1 + ϑ 2 log(t) with ϑ = (ϑ 1 , ϑ 2 ) features one intercept parameter ϑ 1 and an accelerator ϑ 2 > 0.…”

“…Because not all procedures are able to deal with missing values in prognostic or predictive variables, a complete case analysis was performed. A more detailed description of the final data set of N = 2711 observations and 18 patient characteristics is available elsewhere (Seibold et al 2018). To estimate the performance of the different procedures on the data set, we generated 100 random splits of the data into learning and validation samples in a 3 : 1 proportion, keeping the proportions of treated patients and the proportion of patients with right-censored overall survival time in all learning and validation samples the same as in the initial data set.…”

“…models describing the differential treatment effect of Riluzole as a function of patient characteristics, are important for a better understanding of the mechanisms of Riluzole interaction with the nervous system. To date, differential treatment effects of Riluzole have been reported in traditional subgroup analyses (Fang et al 2018), statistical learning approaches for "automated" subgroup analysis (Seibold et al 2016), and estimation of individualized treatment effects (Seibold et al 2018).…”

“…We compared several variants of survival forests in this class to established survival forests (RF-S, CForest, Ranger) and to one recent proposal (explicitly targeting the non-proportional hazards situation using L 1 splitting, Moradian et al 2017) in an artificial prognostic model setting and then investigated the empirical performance for ALS survival prognosis. In a second step, we compared predictive survival forest models based on Weibull models ("distributional survival forests", DSF, Seibold et al 2018) with a less restrictive novel variant of prognostic and predictive survival forests introduced herein ("transformation survival forests", TSF), both with respect to predictive ALS models and based on simulations.…”

Exploratory identification of predictive biomarkers in randomized trials with normal endpoints

“…Every forest consists of 100 trees and each tree is grown on a subsample that includes 63.2 % of all observations in the data set. This fraction corresponds to the inclusion probability of an observation in a bootstrap sample drawn with replacement and matches the one used in Seibold et al Ten biomarkers were considered as candidates for each split. The significance level for the M‐fluctuation test within the splitting process is set to α=0.05, but test results are not adjusted for multiple testing in order not to be too conservative and to limit too much the size of the trees in the random forest.…”