Scientists used CEM and ARMAX to predict pregabalin response in patients with diabetic peripheral neuropathy

Primary tabs

Scientists used CEM and ARMAX to predict pregabalin response in patients with diabetic peripheral neuropathy

A recent Integrating RCT and Observational Study revealed that the combination of cluster analyses, coarsened exact matching (CEM) and autoregressive moving average models (ARMAX) allows strong predictive capabilities with respect to pain scores. Every patient has a distinctive response towards a medical treatment. This finding sets a need for more patient-specific medical care for better results.

The evaluation of better medical care can be defined with the help of various analytical tools. These tools facilitate evaluation by finding and combining different treatment responses studies, like observational studies & randomised controlled trials (RCTs). The study involved the evaluation of pregabalin effect among patients with painful diabetic peripheral neuropathy (pDPN) from different treatment responses studies by integrating these studies into a single platform.
One largest observational study (3159 patients) and three RCTs (398 patients) were selected for evaluation. To determine a set of patients among observational study to which RCTs group patients could link, a hierarchical cluster analysis was conducted. The matching of patients was done through a CEM technique. A dataset of matched patients was made after matching was done. The ARMAXs were developed to evaluate pain scores weekly among matched patients received pregabalin by using the maximum likelihood method. The verification of ARMAX models was done by using t-tests between observed and predicted pain scores. For verification patients from an observational Study who had no match with RCT, patients were used.
A total of six clusters with nine variables were evaluated during cluster analysis. These variables were age, gender, body mass index, depression history, pDPN duration, pregabalin monotherapy, prior gabapentin use, baseline pain score, and baseline sleep interference. The CEM technique evaluated a total of 1528 patients from the matched dataset. A reduction of bias of covariates among five of six clusters was seen after adding RCT patients. The ARMAX models performance were well executed. The t-test exhibited no difference between predicted and observed scores among non-matched patients.
Integrating RCT and Observational Study data facilitated the effectual use of Observational Study data to evaluate patient responses by using CEM.


BMC Medical Research Methodology

Link to the source:

Original title of article:

Integrating data from randomized controlled trials and observational studies to predict the response to pregabalin in patients with painful diabetic peripheral neuropathy


Joe Alexander; et al.


Exploratory, Pregabalin, Diabetic peripheral neuropathy, Pain, Gabapentinoid, Hierarchical cluster analysis, Coarsened exact matching (CEM) technique, Autoregressive moving average models (ARMAXs), t tests
Log in or register to post comments