Abstract
Background
Transcutaneous electrical nerve stimulation (TENS) is used as an adjunctive treatment for diabetic peripheral neuropathy (DPN), but long-term clinical trajectories vary among patients. The biological characteristics associated with these different trajectories remain incompletely characterized. This study aimed to identify longitudinal clinical trajectory patterns and examine their associations with vascular, inflammatory, and neurophysiological measures.
Methods
Prospective 24-month data from 685 patients with DPN receiving standardized TENS therapy were analyzed using an integrated framework comprising latent class growth analysis, Elastic Net feature selection, Bayesian network analysis, model-based counterfactual simulation, longitudinal mixed-effects modeling, and time-lagged analyses.
Results
Three longitudinal trajectory classes were identified: Favorable Improvement (29.9%), Moderate Improvement (42.0%), and Persistent High Burden (28.0%). In subsequent binary analyses, the first two classes were combined as the improved group and the Persistent High Burden class as the non-responder group. ABI, IL-6, and SNCV were among the variables most consistently associated with trajectory classification across complementary analyses. Bayesian network analysis identified conditional dependencies among vascular, inflammatory, neurophysiological, and trajectory-related variables. Model-based counterfactual simulations showed progressively lower estimated probabilities of Persistent High Burden under hypothetical modifications of ABI alone, ABI plus IL-6, and ABI plus IL-6 plus SNCV; the estimated probability changed from 42.7% at the observational baseline to 34.2%, 21.8%, and 11.5%, respectively. Longitudinal mixed-effects and time-lagged analyses showed broadly consistent temporal associations among ABI, IL-6, SNCV, and GCS. Sensitivity analyses, including an SNCV-excluded composite outcome, yielded similar trajectory classifications.
Conclusions
Distinct longitudinal clinical trajectories were observed among patients with DPN receiving TENS. ABI, IL-6, and SNCV were consistently associated with trajectory classification and longitudinal GCS measures. The network and counterfactual analyses provide a model-based description of relationships among these domains but do not establish causal effects or clinical benefits from modifying individual factors. These findings may help inform future studies investigating multidomain determinants of long-term DPN trajectories.