A recently published paper Ben et al. (2023) An R tutorial is provided for conducting economic evaluations (usually cost-effectiveness analyses) using data from clinical trials and analyzing them with R. This article begins by outlining the key issues researchers face when conducting these economic assessments:
- Missing values. Missing data are common in clinical trials due to dropouts, limited follow-up, or nonresponse. What are some ways to solve this problem? The author wrote: “Naive methods such as mean imputation of missing values and last observation carried forward are discouraged because they do not take into account the uncertainty in imputed observations. More robust methods for handling missing and/or censored data The methods are multiple imputation (MI), inverse probability weighting (IPW), likelihood-based models and Bayesian models. Among them, MI is the most commonly used method, when missing data and observation data (such as missing at random, MI is an effective method when MAR) is relevant.” The relevant R package for MI is mouse.
- Data is biased. The cost profile is generally right-skewed, with most observations around the median, but there are also a number of cost outliers. The authors cite a scoping review (El-Alili et al. 2022) and states that appropriate methods for handling skew cost data include: “nonparametric bootstrapping, generalized linear models (GLMs), hurdle models, and Bayesian models with gamma distributions.”
- Related costs and impacts. Sometimes, treatment effectiveness may be related to cost (positively or negatively). Methods for dealing with correlated costs and impacts include seemingly unrelated regression (SUR), pairwise bootstrapping costs and impacts, and Bayesian binary models.
- Baseline Imbalance of Trial Characteristics. Even if individuals are randomized within a trial, the randomization may be imperfect and trial characteristics may be unbalanced. Some methods to account for these differences include: regression-based adjustment, propensity score adjustment, and matching.
Here is some sample code for implementing each of these 4 methods.
Missing values. The relevant R package for MI is mouse.
Address skewed data and associated costs through bootstrapping and seemingly unrelated regression (SUR) methods.The author uses start up Features provided start up R package. The boot function is used to resample the data, and for each bootstrap sample, the following function is used to fit the SUR model System adaptation Function. [The authors note that rather than using SUR, a linear mixed model (LMM) could be fit instead using the lme4 or nlme R packages].
The relevant statistics of interest can then be extracted as follows:
Additional instructions on how to create cost-effectiveness planes and cost-effectiveness acceptability curves are also provided.You can read the full article here.