We evaluated the accuracy of the 10 μg desmopressin test in differentiating Cushing disease (CD) from non-neoplastic hypercortisolism (NNH) and ectopic ACTH syndrome (EAS). A systematic review of studies on diagnostic test accuracy in patients with CD, NNH, or EAS subjected to the desmopressin test obtained from LILACS, PubMed, EMBASE, and CENTRAL databases was performed. Two reviewers independently selected the studies, assessed the risk of bias, and extracted the data. Hierarchical and bivariate models on Stata software were used for meta-analytical summaries. The certainty of evidence was measured using the GRADE (Grading of Recommendations Assessment, Development, and Evaluation Working Group) approach. In total, 14 studies were included: 3 studies on differentiated CD versus NNH and 11 studies on differentiated CD versus EAS. Considering ΔACTH in 8 studies involving 429 patients, the pooled sensitivity for distinguishing CD from EAS was 0.85 (95% confidence interval [CI]: 0.80–0.89, I2 = 17.6%) and specificity was 0.64 (95% CI: 0.49–0.76, I2 = 9.46%). Regarding Δcortisol in 6 studies involving 233 participants, the sensitivity for distinguishing CD from EAS was 0.81 (95% CI: 0.74–0.87, I2 = 7.98%) and specificity was 0.80 (95% CI: 0.61–0.91, I2 = 12.89%). The sensitivity and specificity of the combination of ΔACTH > 35% and Δcortisol > 20% in 5 studies involving 511 participants were 0.88 (95% CI: 0.79–0.93, I2 = 35%) and 0.74 (95% CI: 0.55–0.87, I2 = 27%), respectively. The pooled sensitivity for distinguishing CD from NNH in 3 studies involving 170 participants was 0.88 (95% CI: 0.79–0.93) and the specificity was 0.94 (95% CI: 0.86–0.97). Based on the desmopressin test for differentiating CD from EAS, considering ΔACTH, Δcortisol, or both percent increments, 15%, 19%, or 20% of patients with CD, respectively, would be incorrectly classified as having EAS. For CD versus NNH, 11% of patients with CD would be falsely diagnosed as having NNH, whereas 7% of patients with NNH would be falsely diagnosed as having CD. However, in all hierarchical plots, the prediction intervals were considerably wider than the confidence intervals. This indicates low confidence in the estimated accuracy, and the true accuracy is likely to be different.