We applied multiscale entropy (MSE) to assess variation in crest time (CT), a parameter in arterial waveform analysis, in diagnosing patients with diabetes. Data on digital volume pulse were obtained from 93 individuals in three groups [Healthy young (Group 1, 20 < age ≤ 40, n = 30), healthy upper-middle-aged (Group 2, age > 40, n = 30), and diabetic (Group 3, n = 33) subjects]. Crest time, normalized crest time, crest time ratio (CTR), small- and large-scale MSE on CT [MSESS(CT) and MSELS(CT), respectively] were computed and correlated with anthropometric (i.e., body weight/height, waist circumference), hemodynamic (i.e., blood pressure), and biochemical parameters (i.e., serum triglyceride, high-density lipoprotein, fasting blood sugar, and glycosylated hemoglobin). The results demonstrated higher variability in CT in healthy subjects (Groups 1 and 2) compared with that in diabetic patients (Group 3) as reflected in significantly elevated MSESS(CT) and MSELS(CT) in the former (p < 0.003 and p < 0.001, respectively). MSELS(CT) also showed significant association with waist circumference and fasting blood sugar (i.e., two diagnostic criteria of metabolic syndrome) as well as glycosylated hemoglobin concentration. In conclusion, using MSE analysis for assessing CT variation successfully distinguished diabetic patients from healthy subjects. MSESS(CT) and MSELS(CT) therefore may serve as noninvasive tools for identifying subjects with diabetes and those at risk.