Causal inference methods to study gastric tube use in amyotrophic lateral sclerosis

ABSTRACT

Objective: To estimate effects of gastric tube (G-tube) on survival and quality of life (QOL) in people with amyotrophic lateral sclerosis (ALS) correcting for confounding by indication inherent in nonrandomized observational data.

Methods: To complement a recent causal inference analysis, which concluded that G-tube placement increases the hazard of death, permanent assisted ventilation, or tracheostomy by 28%, we fit causal inference models on a different sample of 481 patients with ALS enrolled in a recent clinical trial of ceftriaxone. Forward selection identified predictors of G-tube placement. Effects of G-tube on survival and QOL were estimated using structural nested models and marginal structural models, accounting for predictors of G-tube treatment.

Results: Forced vital capacity and the total score and bulbar subscale of the revised ALS Functional Rating Scale best predicted G-tube placement. Correcting for these confounders, G-tube placement decreased survival time by 46% (p < 0.001) and had no effect on QOL (p = 0.078). Sensitivity survival analyses varied in significance, but none revealed a survival benefit.

Conclusions: In the absence of randomization, causal inference methods are necessary to correct for time-varying confounding. G-tube placement may have a negative effect on survival with no QOL-related benefit for people with ALS. A randomized controlled trial is warranted to further evaluate the efficacy of this widely used intervention.

Neurology 2017