Cardiac autonomic neuropathy (CAN) may lead to life threatening arrhythmia due to denervation of both the parasympathetic and sympathetic branches of autonomic nervous system innervating the heart. CAN is a frequently under diagnosed complication of diabetes, because a patient can have asymptomatic CAN for several years before it is clinically apparent. However, detection of CAN at the early or subclinical stage leads to more effective treatment outcomes. Cardiac autonomic reflex tests (CART) (i.e. Ewing test battery) are normally used for the detection and staging of CAN. These tests have limitations with the necessity of active participation of the patients for test maneuvers, as a majority of patients will not be able to complete all five tests required due to comorbidities such as frailty, obesity or cardiorespiratory disease. CAN affects both heart rate (measured by RR interval dynamics) and ventricular repolarization function (i.e. QT interval dynamics) of the heart, which can be efficiently analyzed from surface ECG. Therefore, ECG based diagnosis techniques of CAN analysis are becoming popular as they can reduce the limitations of CARTs used traditionally for CAN detection and it complements CART results. In this study, the performance of an ECG based QTV feature derived using a model free approach, which can quantify the QTV component not affected directly by the heart rate (HR) variation, is compared with some other measures of QTV and HRV in subclinical CAN detection in diabetes. Short-term ECGs (i.e. 5 min long) of 60 diabetic subjects without CAN and 50 diabetic subjects detected with early level of CAN determined by CART were analyzed. The proposed measure for quantifying the QTV component independent of HR denoted as QTV~RR stands out to be more discriminatory than other existing variability measures of QTV and HRV in subclinical detection of CAN.