BACKGROUND The diagnosis of canine diabetic ketoacidosis (DKA) usually is based on measurement of urinary acetoacetate (ketonuria). In humans, this test is less sensitive and specific than blood 3-beta-hydroxybutyrate (ketonemia) evaluation. HYPOTHESIS Ketonemia measurement using a portable meter is more accurate than ketonuria determination with a dipstick to diagnose canine DKA. ANIMALS Seventy-two client-owned diabetic dogs with ketonemia, ketonuria, or both. METHODS Prospective observational study. Based on blood bicarbonate concentration and anion gap, dogs were divided into 2 groups: patients with DKA (n= 25); patients with diabetic ketosis (n= 47). Sensitivity, specificity, and positive and negative likelihood ratio (LR) at different cut-off points were determined for both ketonemia and ketonuria. Receiver operating characteristic (ROC) analysis was used to assess the accuracy of each diagnostic test to diagnose DKA. RESULTS With regard to ketonemia, cut-off values of 2.3 and 4.3 mmol/L revealed 100% sensitivity and 100% specificity, respectively, whereas cut-off values of 2.8 and 3.5 mmol/L showed a -LR of 0.05 and a + LR of 13.16, respectively. With regard to ketonuria, a cut-off value of 1+ revealed 92% sensitivity, 40% specificity, and -LR of 0.20, whereas a cut-off value of 3+ revealed 44% sensitivity, 94% specificity, and +LR of 6.89. The areas under the ROC curves for the ketonemia and ketonuria tests were significantly different (0.97 and 0.81, respectively, P= .003). CONCLUSIONS AND CLINICAL IMPORTANCE Measurement of ketonemia is accurate and more effective than measurement of ketonuria to diagnose canine DKA.