Integrating heart rate variability and electrochemical skin conductance as severity screening of cardiovascular autonomic neuropathy in type 2 diabetes

Background: Clinical studies show that either heart rate variability (HRV) or electrochemical skin conductance (ESC) alone can serve as a simple and objective method for screening cardiovascular autonomic neuropathy (CAN). We tested the hypothesis that combining these two quantitative approaches can provide a better estimate of CAN severity in patients with type 2 diabetes (T2DM) who had already suffered from CAN in outpatient clinics. Methods: Each patient received a complete battery of cardiovascular autonomic reex tests (CARTs), with ESC measured by SUDOSCAN, time domain measured by standard deviation of all normal RR intervals (SDNN) and frequency domain of HRV (low frequency [LF], high frequency [HF], and LF/HF ratio), and peripheral blood studies for vascular risk factors. Severity of CAN was measured by CAN score. Results: The 90 T2DM patients included 50 males and 40 females. Those with more severe CAN had a higher CAN score value (P<0.0001) and lower values in feet ESC (P=0.023) and SDNN (P<0.0001). Stepwise linear regression analysis also showed that feet ESC and SDNN value (P<0.0001 and P<0.0001) were signicantly associated with CAN score, respectively. Conclusions: Based on our results, a combination of electrophysiologic biomarkers (SDNN and feet ESC) as a test battery can improve the diagnostic accuracy and reinforce the accuracy in estimating CAN severity and can serve as a time-effective screening service in outpatient clinics.


Background
Diabetic cardiovascular autonomic neuropathy (CAN), the impairment of the autonomic balance of the cardiovascular (CV) system in the setting of diabetes mellitus (DM), strongly in uences various CV diseases, causes detrimental effects on the quality of life, and leads to severe morbidity and mortality in patients with type 1 and type 2 diabetes mellitus (T1DM and T2DM) [1,2]. The autonomic nervous system is one of the major homeostatic regulatory systems of the body. Sympathovagal balance failure with a sympathetic dominance is the main trigger of lethal arrhythmias and sudden death.
The prevalence of CAN is variable and depends on the de nition and criteria used for diagnosis. A joint consensus statement by the American Diabetes Association (ADA) and American Academy of Neurology (AAN) has recommended that a battery of cardiovascular autonomic re ex tests (CARTs) should be performed to assess CAN [3].
Reduced heart rate variability (HRV) is an independent adverse prognostic factor used to detect CAN. HRV represents one of the most promising markers. A joint European Society of Cardiology and North American Society of Pacing and Electrophysiology task force de ned and established standards of measurement, physiological interpretation, and clinical use of HRV, and the most widely used methods can be grouped under the time domain and frequency domain methods [4,5]. Established clinical data based on numerous studies consider HRV have been proven useful in detecting CAN in diabetes patients [6,7]. Recently, portable devices that measure HRV provide cost-bene t along with the simplicity of the measurement and increased compliance for both research and clinical studies [8].
Sudomotor dysfunction, a length-dependent pattern of diseases in thin, non-myelinated sympathetic Cbers, has been observed in both prediabetic and diabetic patients. The consensus statement of the ADA, AAN, and Latin American Diabetes Association includes assessing the sudomotor function's role in the early diagnosis of CAN in patients with diabetes [9,10]. Sudomotor function can be assessed using SUDOSCAN ™ (Impeto Medical, Paris, France), which measures electrochemical skin conductance (ESC), which is a technological advance that calculates sweat function through the quantization of chronoamperometry measures of the hands and feet. It was recently developed to allow the measurement of diabetic small-ber neuropathy and autonomic dysfunction [11].
Although HRV and ESC are validated and can serve as screening tools for CAN in patients with diabetes [5,[12][13][14][15][16], there is paucity of information that focuses on its role in estimating CAN severity in patients with diabetes who was already diagnosed with CAN in outpatient clinics, owing to the possible bene ts of exploring the role of the CV autonomic function on subsequent CV events and the consequent development of therapeutic strategies to reduce the prevalence of CV events.
In this study, we evaluated the feasibility of a time-effective CV autonomic screening service at the outpatient clinic, including the time and frequency domains of HRV as well as the ESC. First, we evaluated which parameters in both HRV and ESC can serve as diagnostic biomarkers in the presence of CAN. Second, we combined the biomarkers in a test battery for CAN severity screening. We tested the hypothesis that CAN score and electrophysiologic parameters (time and frequency domains of HRV and ESC) have a strong association. If our hypothesis is true, when combined as a test battery, these two quantitative approaches can provide a better estimate of CAN severity and improve examinations.

Study population
A total of 90 patients (≥20 years old) with T2DM who visited the outpatient diabetic clinic at Kaohsiung Chang Gung Memorial Hospital in Taiwan were included. Exclusion criteria included those who (1) suffered from moderate-to-severe heart failure (New York Heart Association class III and IV) and (2) had any type of arrhythmia that prevents the analysis of HRV, or pacemaker implantation due to any cause. This study was approved by the Ethics Committee of Chang Gung Memorial Hospital Institutional Review Board (201800388B0C501 and 201901363B0).

Baseline clinical and laboratory measurements
All patients underwent complete neurological and physical examinations upon enrollment and at their subsequent follow-ups at the outpatient clinic. A detailed medical history regarding prior use of medications was obtained from the patients and their families through standardized questions.
Demographic data, including age, sex, duration of diabetes (years), body mass index (BMI), systolic and diastolic blood pressure (SBP and DSP), waist circumference (WC) during autonomic function testing, underlying disease (hypertension, coronary artery disease [CAD], ischemic stroke, and diabetic retinopathy [DR]), and laboratory parameters, were obtained at baseline.

Biochemical analysis
Blood samples were obtained by antecubital vein puncture in a fasting, nonsedative state between 09:00 and 10:00 AM in the control and study groups to exclude the possible in uence of circadian variations. All blood samples were collected in Vacutainer SST tubes (BD, Franklin Lakes, NJ) and centrifuged at 3000 rpm for 10 min, and serum samples were collected and stored at −80°C in multiple aliquots prior to biochemical measurement.
Serum levels of triglycerides, total cholesterol, high-density lipoprotein cholesterol (HDL-C), low-density lipoprotein cholesterol (LDL-C), blood sugar, HBA1c, and high-sensitivity C-reactive protein (hs-CRP) were analyzed by the hospital's central laboratory. Estimated glomerular ltration rate (eGFR) in each patient was calculated using an equation for Chinese subjects, as previously described [17]. Urine albumin/creatinine ratio (UACR) is less than 30 mg/day; therefore, UACR between 30 and 300 mg/day is called microalbuminuria, and UACR above 300 mg/day is considered macroalbuminuria [18].
Assessment of electrochemical skin conductance and SUDOSCAN risk scores ESC were assessed with patients placing their hands and feet on electrode plates for 3 min [19]. Measurement was based on an electrochemical reaction between electrodes and chloride ions, after stimulation of small bers innervating the sweat glands with a low-voltage current (<4 V). An ESC measurement for the hands and feet was generated from the derivative current associated with the applied voltage. Lower ESC values indicated dysfunction of the sweat glands.
SUDOSCAN risk scores were calculated automatically from ESC values, BMI, and age using an algorithm included in the device software. Scores were presented as percentages. Higher SUDOSCAN risk scores have been related to increased risk of cardiac autonomic abnormalities. Both ESC values and SUDOSCAN risk scores were displayed numerically with graphs on the device monitor [12].

Assessment and scoring of cardiovascular autonomic functions
CARTs are considered as gold-standard measures of autonomic function in patients with diabetes [3].
Parameters that were computed using Ewing's methods, including heart rate responses to deep breathing (E:I ratio), standing (30:15 ratio), and Valsalva maneuver and blood pressure responses to standing [20], were often used by diabetologists. These autonomic parameters were also obtained, and CAN was de ned as the presence of at least two abnormal test results [3].
The severity of CAN was quantitated by summation of points obtained from each of the four tests, where each test was given a point of 0 or 1 if it yielded normal or abnormal values, respectively. We termed this severity score as CAN score. Therefore, CAN score provided a score from 0 to 4 points in this study. In this study, CAN is de ned as no CAN, early CAN, de nite CAN, and severe CAN if no, one, two, and more than two abnormal autonomic function testings.

Parameters of heart rate variability (HRV)
Using the 5-min resting electrocardiogram (ECG) recording, the standard deviation of all normal RR intervals (SDNN) was calculated as the time domain parameter of HRV. In addition, power spectral density analysis of HRV was also done to obtain the frequency domain parameters. Three main spectral components were distinguished in a spectrum calculated from the 5-min recording, i.e., high frequency (HF, 0.15-0.4 Hz), low frequency (LF, 0.04-0.15 Hz), and very low frequency (VLF, 0-0.04 Hz). The components of LF and HF were computed both in absolute values of power (ms 2 ) and in normalized unit (n.u.). The LF/HF ratio, regarded as an index of sympathovagal balance, was also calculated [16]. The aforementioned computing process was done by Kubios HRV Standard version 3.2 (Kubios Oy, Finland)
Categorical variables were compared using chi-square or Fisher's exact tests. Continuous variables that were not normally distributed by Kolmogorov-Smirnov test were logarithmically transformed to improve normality and compared. Five separate statistical analyses were performed. First, patients were strati ed into four groups according to the severity of CAN. Second, receiver operating characteristic (ROC) curves were generated for parameters of ESC and time and frequency of HRV and CARTs in the presence of CAN.
The area under the ROC curve (AUC) for the presence of CAN in addition to the sensitivity, speci city, and Youden's index of each parameter was also calculated. Fourth, correlation analysis was used to evaluate the relationship between the CAN score and the variables that included age, diabetes duration, BMI, and parameters of ESC and time and frequency of HRV. Finally, stepwise models of multiple linear regression analysis were used to evaluate the in uence of independent variables on the mean CART score. All statistical analyses were conducted using the SAS software package, version 9.1 (2002, SAS Statistical Institute, Cary, North Carolina).

Results
General characteristics of patients with diabetes strati ed by severity of cardiovascular autonomic neuropathy The 90 patients with diabetes included 50 males and 40 females. Patient characteristics and baseline underlying diseases at assessment are presented in Table 1.
Laboratory and autonomic function testings of patients with type 2 diabetes strati ed by the severity of cardiovascular autonomic neuropathy Those with more severe CAN had higher UACR (P=0.012). Those with more severe CAN had higher CART score value (P<0.0001) and lower values in parasympathetic parameters, including the Valsalva ratio (VR), E:I ratio, and 30:15 ratio (P<0.0001, P<0.0001, and P<0.0001, respectively), and lower values in feet ESC (µS) (P=0.023), BP change related to standing (P<0.0001), and SDNN (<0.0001). Those with more severe CAN had a higher prevalence of orthostatic hypotension (P<0.0001) ( Table 2).
Diagnostic accuracy for parameters of autonomic functions in predicting cardiovascular autonomic neuropathy, using ROC curve analysis

Effects of the variables on cardiovascular autonomic re ex tests
The effects of the variables on CAN score in patients with T2DM according to correlation analysis are listed in Table 4. Based on the correlation analysis, only signi cant variables were enrolled into stepwise linear regression models. Results from the model analysis (Table 5) revealed that only age, SDNN, and feet ESC value were signi cantly associated with CAN score and were enrolled into stepwise linear regression models. The formula is as follows: CAN score=5.936 − 0.026 × (SDNN) − 0.045 × (age) − 0.019 × (feet ESC).

Discussion
Clinical data show that clinical manifestations of CAN appear in T2DM, and the estimated 5-year mortality is approximately 50% [21]. Thus, evaluating the severity of CAN regression or progression is important for risk strati cation and subsequent management. To our knowledge, efforts on those patients who had already suffered from CAN remain unsatisfactory.
Major ndings of our study Our study produced three major ndings. First, SDNN, total power of HF and LF, and feet ESC had su cient diagnostic accuracy in the presence of CAN in ROC analysis though only either HRV or ESC as a screening tool could be unsatisfactory. However, combined SDNN and feet ESC can increase the diagnostic accuracy of CAN (sensitivity, speci city, and AUC). Second, SDNN, total power of HF and LF, and feet ESC had been shown to bear a signi cant negative correlation with CAN score, indicating that they are surrogate markers of CAN severity. Finally, our study results showed that combining SDNN, total power of HF and LF, and feet ESC can improve assessment in a multiple linear regression model (Rsquared value is highest in model 3).

Risk factors associated with the severity of CAN
The pathophysiological mechanism of CAN development is multifactorial, and several studies reported the important role of CV risk factors, such as diabetes duration, degree of glycemic control, SBP, triglyceride levels, and BMI, in the development of CAN [22][23][24][25]. Further, there is enough evidence that CAN may precede DM [24]. A framework for screening and early multifactorial interventions (hyperglycemia, hypertension, dyslipidemia, and microalbuminuria) is the best prospect for preventing or halting CAN and its devastating sequelae and is the gold standard of diabetic care in T2DM [26] and T1DM [27]. Diabetes duration is an important risk factor for CAN. CAN is a length-dependent pattern of disease, and parasympathetic activity can be damaged in the early phase of CAN with sympathetic predominance. As the diseases progress, sympathetic denervation occurs in the late stage of CAN [28]. Our study also showed that all parasympathetic parameters (e.g., VR, E:I ratio, 30:15 ratio, and SDNN) were signi cantly lower in proportion to the severity of CAN.

Feet ESC and SDNN as electrophysiologic markers for evaluation of CAN severity
CARTs are the gold standard for the diagnosis of CAN. Nevertheless, there are some limitations to perform the whole test battery in clinical practice. It takes around 30 min to complete all the tests, and thus screening a large number of patients with diabetes may be impossible in some busy clinics. In addition, the tests sometimes may not be done due to patients' limitation, such as patients' underlying conditions (e.g., dementia or parkinsonism) and those who are unable to do deep breathing or Valsalva maneuver, or patients who are unable to stand. Therefore, a simpli ed effective method may be needed in busy outpatient clinics so that most patients can be screened or frequently followed up.
A clinical study showed that ESC reduction is proportional to the skin nerve ber density [29] and can be used as a noninvasive measurement and is correlated with the gold standard measurement for the severity of small ber neuropathy (e.g., skin biopsy with quantization of intra-epidermal nerve ber density). Furthermore, ESC can be a clinically meaningful tool to measure severity and follow-up for progression and regression [30]. Although several studies have demonstrated that SUDOSCAN risk score is a good screening test for CAN [12,31], the diagnostic accuracy of CAN in our study is not established.
The available information was calculated automatically from ESC values, BMI, and age using an algorithm included in the device software as we did not know the detailed algorithm of SUDOSCAN risk scores. However, several CV risk factors, such as diabetes duration, degree of glycemic control, SBP, triglyceride levels, and diseases of central or peripheral autonomic dysfunctions, could contribute to the severity of CAN, but may not be considered in the device software. Further, different ethnic groups also can affect the normal values of ESC that seemed to lower in Chinese populations [32]. However, other studies and our study showed that feet ESC is signi cantly correlated with the severity of CAN [12], which is compatible with the length-dependent pattern of diseases.
Sympathetic and parasympathetic stimuli directly in uence heart rate and are responsible for the physiologic variation in HRV. HRV can be evaluated in the time and frequency domains [33]. Short-term HRV may be another time-saving tool that can be used in busy clinics. Only a 5-min resting recording of ECG is needed for computation and the requirement of patient's cooperation is decreased.
There are several different approaches of HRV in the time and frequency domains. Among these methods, the time domain parameter, SDNN, is the most intuitional with the simplest computation.
Furthermore, the parameter is essential for almost all portable HRV devices. Since variance is mathematically equal to the total power of spectral analysis, SDNN re ects all the cyclic components responsible for the variability at the time of recording.
Although parameters in the frequency domain may provide information about parasympathetic and sympathetic modulations or balance in addition to cardiovagal function [34], it bears some pitfalls in the interpretation of results. Application of the such technique is critically dependent on the understanding of the underlying physiology, the mathematical analyses used, and the many confounders and possible technical artifacts. The measurement of VLF, LF, and HF power components is usually made in absolute values of power (power density, ms 2 ). LF and HF may also be measured in normalized units (n.u.), which represent the relative value of each power component in proportion to the total power minus the VLF component. The representation of LF (n.u.) and HF (n.u.) emphasizes the control and balance between the sympathetic and parasympathetic systems and tends to minimize the effect of the changes on the values of LF and HF total power components. Nevertheless, normalized units should always be quoted with absolute values of the LF and HF power in order to describe completely the distribution of power in spectral components. Our study showed that the values of LF and HF total power decreased in proportion to the severity to CAN, and both LF and HF total power were negatively correlated with the CAN score. It implies that both parasympathetic and sympathetic impairments were more severe as the severity of CAN progressed. In view of our current data in HRV, SDNN has the highest correlation coe cient with the CAN score among these HRV parameters. We recommend using SDNN as an indicator for the severity of CAN.
Finally, our study con rmed that this combination of electrophysiologic biomarkers (SDNN and feet ESC) as a simpli ed test battery is a feasible time-effective CV autonomic screening service for outpatient clinics and can be completed in less than 10 min. It also reduces clinic visits and provides objective and quantitative measures of CAN in those patients who already had CAN. Similar to the idea of a one-stop service for microvascular screening [35], our test battery can serve as a one-stop CAN screening service. If the test battery shows CAN progression, they will resort to the more sophisticated and speci c, but ultimately more time-consuming complete autonomic function testings (e.g., CARTs).

Study Limitations
This study is a cross-sectional prospective study. Although we observed close relationship between SDNN and feet ESC and the severity of CAN in this observational study, the role of the test battery together with the combination of the two electrophysiologic biomarkers (SDNN and feet ESC) in a longitudinal study to evaluate the regression or progression of CAN is mandatory. Second, our study only included a Chinese population with type 2 diabetes. We did not included those patients who had pre-diabetes, non-diabetes or normal control.

Conclusion
Based on our results, a combination of electrophysiologic biomarkers (SDNN and feet ESC) as a test battery can improve the diagnostic accuracy and reinforce the accuracy by estimating the severity of CAN and can serve as a time-effective screening service in outpatient clinics.
Abbreviations HRV, heart rate variability; ESC, electrochemical skin conductance; CAN, cardiovascular autonomic neuropathy; SDNN, standard deviation of all normal RR intervals; CARTs, cardiovascular autonomic re ex tests; LF, low frequency; HF, high frequency; CV, cardiovascular; DM, diabetes mellitus; T1DM and T2DM, type 1 and type 2 diabetes mellitus; ADA, American Diabetes Association; AAN, American Academy of Neurology; CARTs, cardiovascular autonomic re ex tests; BMI, body mass index; SBP and DSP, systolic and diastolic blood pressure; WC, waist circumference; CAD, coronary artery disease; HDL-C, high-density lipoprotein cholesterol; LDL-C, low-density lipoprotein cholesterol; hs-CRP, high-sensitivity C-reactive protein; eGFR, Estimated glomerular ltration rate; UACR, albumin/creatinine ratio; ECG, electrocardiogram; SDs, standard deviations; IQR, interquartile ranges; AUC, area under the ROC curve; HbA1c, glycohemoglobin; VR, Valsalva ratio Declarations Authors' contributions YRL participated in the design of the study and drafted the manuscript. WCC, CCH, NWT, BCC, and JFC participated in the sequence alignment and clinical evaluation of patients. CCH performed the statistical analysis. CHL and CCH conceived the idea for the study and participated in its design and coordination and helped draft the manuscript. All authors read and approved the nal manuscript.