Unsupervised clustering analysis reveals distinct postoperative cortisol trajectories following pituitary adenoma resection in Cushing's disease.

purposeCushing’s disease (CD), caused by ACTH-secreting pituitary adenomas, results in hypercortisolism. Transsphenoidal surgery (TSS) is the first-line treatment. Early postoperative cortisol trends are frequently monitored, though standard interpretation is lacking. We sought to determine whether early postoperative cortisol trajectories are associated with biochemical remission at 1 year using unsupervised machine learning.

methodsWe retrospectively reviewed 94 adult patients with CD who underwent TSS between 2011 and 2021 at two academic centers. Serum cortisol was measured every 6 h, and the resulting time series was clustered with the goal of predicting the primary outcome of biochemical remission at 1-year follow up.

resultsOf 94 patients, 77 (82%) achieved biochemical remission. There were no significant differences in demographic or tumor characteristics between remission and non-remission groups. The nadir cortisol value for the remission group was lower than the non-remission group (4.8 [5.5] versus 19.3 [13.7]) though this difference was also not statistically significant (p = 0.1022). Clustering analysis identified three optimal clusters for remission subgroup. Statistically significant differences in cortisol trends were observed from 6 h to 30 h (p < = 0.0002). At 6 h, average cortisol values by cluster ranged from 25.51 to 61.17 mcg/dL. Two of the three clusters demonstrated an initial decrease in cortisol values, while one cluster did not immediately decrease. No significant differences in clinical, tumor, or treatment variables were observed between clusters.

conclusionPostoperative cortisol trajectories show distinct early patterns but are not independently predictive of 1-year biochemical remission following TSS for CD.