The recent developments in patient specific medical care has generated a lot of variations in medical treatments. Analytical tools establish a link between RCTs and observational studies. This cluster analysis provides the data from different studies which were collected and integrated to examine patient responses to pregabalin in patients with diabetic neuropathy.

More patient-specific medical care is demanded as further is discovered regarding variations in patient responses to medical practices. Analytical tools allow insights by combining treatment responses from various types of studies, such as observational studies and randomised controlled trials (RCTs). This study aimed to combine such types of data into a single predictive program to anticipate the response to Pregabalin among subjects with painful diabetic peripheral neuropathy (pDPN).

One largest observational study of Pregabalin comprising 3159 German patients and three pivotal RCTs of Pregabalin including 398 North American patients were selected for the analysis. A hierarchical cluster analysis was implemented to distinguish patient clusters in the Observational Study to which RCT subjects could be matched by using the coarsened exact matching (CEM) method, thereby performing a matched dataset.   Autoregressive moving average models (ARMAXs) were developed to measure weekly pain scores for Pregabalin-treated individuals within each cluster in the matched dataset applying the maximum likelihood approach. Lastly, ARMAX models validation was done by utilising Observational Study subjects who had not matched with RCT subjects, using t-tests between predicted and observed pain scores.

Cluster analysis capitulated six clusters with the subsequent clustering variables: age, body mass index, pDPN duration, gender, baseline sleep interference, depression history, baseline pain score, Pregabalin monotherapy, and prior Gabapentin use. CEM obtained 1528 individual patients during the matched dataset. The decline in global imbalance scores for the clusters after incorporating the RCT patients showed that the method decreased the bias of covariates in five of the six clusters. ARMAX models of pain score executed well. Further, t-tests did not reveal differences within predicted and observed pain scores in the 1955 subjects who had not matched with RCT patients.

The combination of CEM, ARMAX modelling, and cluster analyses allowed strong predictive abilities with respect to pain scores. Combining Observational Study and RCT data employing CEM allowed efficient utility of Observational Study data to predict patient responses.