Diagnosis of diabetic autonomic neuropathy: a multivariate approach

Aim/hypothesis In the diagnosis of diabetic autonomic neuropathy (DAN) various autonomic tests are used. We took a novel statistical approach to find a combination of autonomic tests that best separates normal controls from patients with DAN. Methods Twenty‐four patients with Type‐1 diabetes mellitus considered as having mild to moderate DAN as well as 10 normal, nondiabetic control subjects were analysed, searching for a test or a combination of tests that would optimally discriminate Type‐1 diabetes mellitus from controls. Variations of heart rate during deep breathing (ΔR6) and during a Valsalva manoeuvre (VR), the number of reactive sweat glands on the foot (testing sympathetic sudomotor function), and the response of human pancreatic polypeptide to hypoglycaemia [ln(ΔhPP + 1)] were evaluated. Results Respective values for respective sensitivity and specificity values were: ΔR6, 96 and 70%; VR, 96 and 60%; sweat gland function, 71 and 90%; and ln(ΔhPP + 1), 71 and 90%. In a multivariate analysis approach a single discriminant function separating patients with Type‐1 diabetes mellitus from nondiabetic controls was generated {Logit P = 288·5 –[14·7 ΔR6]–[26·6 ln(ΔhPP + 1)]}. This function allowed complete separation of patients with Type‐1 diabetes mellitus from normal controls. Conclusion/interpretation We conclude that the combined determination of ΔR6 and of ln (ΔhPP + 1) optimally separates subjects with parasympathetic impairment from normal subjects. In addition, this combination of tests may serve as a sensitive method for the assessment of DAN.