**1. Introduction**

The prevention and early detection of diabetes mellitus (DM) in high-risk subjects such as those with metabolic syndrome have become important issues in preventive medicine and public health [1,2]. As for type 2 diabetic patients, the outcomes of microvascular and macrovascular complications can be grievous. Diabetic peripheral neuropathy (DPN) is not only one of the most common complications

> 41

of chronic diabetes, but also a leading cause for disability due to foot amputation and ulceration, fall-related injury, and gait disturbance. [3,4]. Not only is the quality of life much lower among DPN patients but the mortality rate has also been shown to be higher than for DM alone [5]. To avoid developing a severe condition, blood glucose control is the most important issue for type 2 diabetic patients [6–8]. However, this is not easy to achieve, which is the reason many researchers have tried to develop some non-invasive methods and instruments [9] for type 2 diabetic patients to avoid the increased risk of developing atherosclerosis and autonomic nervous dysfunction [10–13].

Heart rate variability (HRV) using electrocardiography (ECG) is an assessment method of autonomic function and baroreflex sensitivity (BRS) [14]. The low-/high-frequency power ratio (LHR) is the autonomic function index for frequency domain analysis and it is considered to balance the reflection between sympathetic and parasympathetic activity changes [14–16]. However, the conventionally used frequency domain parameter is not always adequate for this purpose because of non-stationarity and the nonlinear physiological signals adopted [17–19]. Therefore, several new parameters based on nonlinear dynamics theory were recently applied to HRV studies on autonomic function and BRS [20–22]. Among these parameters, a small-scale multiscale entropy index (MEISS) was adopted to evaluate the autonomic nervous activities and BRS based on a nonlinear approach, using only the R–R interval (RRI) datasets [20]. In recent years, a new percussion entropy index (PEI) using synchronized ECG and photoplethysmography (PPG) signals was used to assess the BRS complexity in healthy elderly and diabetic subjects related to autonomic function changes [21,22]. Moreover, in a recent study [23] on a modified PEI, the PPG signals were measured and the peak-to-peak interval (PPI) series were calculated at the fingertip as an indicator of DPN in the aged and diabetic patients. However, real-time processing was not possible for the PPI-based index presented in [23]. On the other hand, parameters including the large-scale multiscale entropy index (MEILS) [20], in addition to pulse wave velocity (PWVmean) [24–27], were reported for atherosclerosis detection in type 2 diabetes mellitus.

From the viewpoint of data analysis, we propose that some of the above parameters (i.e., LHR, MEISS, MEILS, PWVmean, and PEI) could present highly significant di fferences between diabetic patients who underwent a synchronized ECG and PPG signals testing at baseline during the early stages of disease; these patients were then followed for a further six years, with and without DPN. Furthermore, this study was designed to apply synchronized ECG and PPG signals from a non-invasive instrument in predicting the development of peripheral neuropathy in type 2 diabetic patients. It is worth mentioning that PEI [21,22] has recently been introduced to assess the complexity of BRS, while the significance of smaller PEI values concerning the identification of subjects with type 2 diabetes who are more prone to developing diabetic neuropathy is unknown. This study aimed at testing the hypothesis that a smaller PEI value at the baseline measurement can provide valid information that may help identify type 2 diabetic patients at a greater risk of future DPN.

The rest of this paper is organized as follows. Section 2 describes the study population and the baseline examinations and protocol of measuring of synchronized ECG and PPG signals. A follow-up procedure and DPN status checking were addressed, followed by an explanation of the statistical analysis methods. As presented in Section 3, results from the five indices—LHR, MEISS, MEILS, PWVmean, and PEI—were first computed for diabetic subjects with peripheral neuropathy within six years (i.e., Group 3) for comparison with the healthy elderly subjects (i.e., Group 1) and diabetic patients without peripheral neuropathy (i.e., Group 2). Subsequently, three new diabetic subgroups (i.e., Groups A–C) with di fferent periods of PEI values were identified for follow-up and Cox proportional Hazards model as well as goodness-of-fit test for relative risks analysis. Finally, a Cox regression analysis of risk factors for the incidence of DPN within six years of follow-up in diabetic patients was verified. In Sections 4 and 5, the discussion of the results and conclusions from the present study are summarized, with suggestions for future work.

#### **2. Materials and Methods**

#### *2.1. Study Design and Study Population*

#### 2.1.1. The Inclusion and Exclusion Criteria Were as Follows

Between July 2010 to March 2013, 128 subjects were enrolled for this study. All diabetic patients were recruited from the diabetes outpatient clinic of the Hualien Hospital (Hualien City, Taiwan), while healthy controls were recruited from a physical check-up program at the same hospital. All of the age-controlled healthy subjects had no personal or family history of cardiovascular diseases. Of the 128 volunteers, 6 were excluded due to a history of coronary heart disease, heart failure, ischemic stroke, peripheral arterial disease, chronic atrial fibrillation, or permanent pacemaker implantation.
