The pituitary origin of ACTH secretion in ACTH-dependent hypercortisolism can be difficult to assess, as magnetic resonance imaging (MRI) frequently fails to identify ACTH-secreting microadenomas or, on the contrary, may give false positive images of microadenomas. The choice of therapeutic option for patients with such normal MRI findings is controversial. Some groups propose routinely pituitary surgery, whereas others consider that neurosurgical exploration may be less successful and more harmful, and therefore prefer other types of management. The aim of this study was to compare surgical outcomes between patients with Cushing's disease (CD) and normal vs. positive pituitary MRI findings. Fifty-four patients (44 women and 10 men) with CD, operated on after 1996 in two centers (Kremlin-Bicêtre and Bordeaux) and followed postoperatively during a mean period of 19.9 +/- 22.7 months (range, 1-89 months), were enrolled in this retrospective study. Twenty-eight patients had normal pituitary MRI findings, and the pituitary origin of ACTH was established by bilateral petrosal sinus sampling in all of these cases. Twenty-six patients had positive MRI findings clearly showing a microadenoma. The two groups were not significantly different in terms of the sex ratio, age, frequency of hypertension, or diabetes, basal 24-h urinary free cortisol levels and follow-up. All of the patients were operated on by two experienced neurosurgeons using the same surgical protocol. Selective adenomectomy was performed when a tumor was identified, and subtotal hypophysectomy was performed when the lesion was uncertain or when no tumor was found during surgical exploration. Respectively, 50% and 84% of patients with normal and positive MRI results underwent adenomectomy (P < 0.05). A pituitary adenoma (confirmed by pathological examination) was found at surgery in 53% and 88% of patients in the normal and positive MRI groups, respectively (P < 0.05). The early surgical success rate (combining patients with corticotropic deficiency and patients with eucortisolism) was similar in the normal and positive MRI groups (78% and 88%, respectively; P = 0.85). The recurrence rate was lower in the normal MRI group, but the difference did not reach statistical significance (9% vs. 30%; P = 0.07). The final remission rate at the last visit was similar in the normal and positive MRI groups (72% and 61%, respectively; P = 0.29). Postoperative complications were also similar: 10 patients (36%) with normal MRI and five patients (20%) with positive MRI had at least one postoperative complication (surgical and/or pituitary deficiency; P = 0.12). Thus, the outcome of pituitary surgery in CD appears to be similar regardless of whether pituitary MRI shows a microadenoma. We recommend neurosurgical pituitary exploration as the first-line treatment of CD, provided that the pituitary origin of ACTH secretion is confirmed by bilateral petrosal sinus sampling in patients with normal pituitary MRI findings.