Introduction According to guidelines, a morning serum cortisol level <83 nmol/L is diagnostic for central adrenal insufficiency (CAI), a value >414 nmol/L excludes CAI, while values between 83 and 414 nmol/L require stimulation tests. However, there are no currently reliable data on morning serum cortisol for prediction of cortisol response to insulin tolerance test (ITT). Objective Using the receiver operating characteristic curve analysis, the purpose of this study was to detect the morning serum cortisol cut-off with a specificity (SP) or a sensitivity (SE) above 95% that identify those patients who should not be tested with ITT. Methods We included 141 adult patients (83 males) aged 42.7+/-12.3 (mean +/- SD) years old. Based on serum cortisol response to ITT, patients have been divided in two groups: subjects with CAI (peak serum cortisol <500 nmol/L; 65 patients) and subjects with preserved adrenocortical function (peak cortisol >500 nmol/L; 76 patients). Results The best morning cortisol cut-off, in terms of SE (87.7%) and SP (46.1%), was ≤323.3 nmol/L. The cut-off of morning serum cortisol concentration that best predicted a deficient response to ITT was ≤126.4 nmol/L (SE 13.8%, SP 98.7%). The cut-off of morning serum cortisol concentration that best predicted a normal response to ITT was >444.7 nmol/L (SE 96.9%, SP 14.5%). Conclusions This is the first study that identifies a morning serum cortisol cut-off that best predict the response to ITT in order to simplify the diagnostic process in patients with suspected CAI. A new diagnostic flow chart for CAI is proposed.