Pituitary Carcinoma Revisited: A Systematic Review and Individual Patient Data Analysis of Clinicopathological Features, Treatment, and Survival

Background: Pituitary carcinoma (PC) is a rare and aggressive pituitary neuroendocrine tumor (PitNET) defined by non-contiguous craniospinal or systemic metastatic spread. Evidence guiding treatment is limited to case reports and case series. This systematic review aimed to characterize the clinicopathological features, treatment patterns, and prognostic determinants of PC through pooled individual patient data (IPD) analysis. Methods: A systematic search of PubMed/MEDLINE and Embase (January 1990 to March 2026) was conducted in accordance with PRISMA 2020 guidelines (PROSPERO: CRD420261402665). Individual patient data were extracted for 56 pre-specified variables from the identified literature. Two additional institutional cases were analyzed separately as illustrative clinical examples and are not part of the pooled statistical cohort. Survival was estimated using the Kaplan–Meier method with log-rank tests. Univariate and multivariable Cox proportional hazards regression and binary logistic regression (outcome: long-survivor status, defined as OS > 24 months) were performed on the pooled literature cohort. Results: A total of 111 patients from 93 papers were included (51 males (45.9%); median age at PC diagnosis 54 years (IQR 40–61)), with corticotroph lineage predominating (47.7%). Median OS from PC diagnosis was 19 months (IQR 9.5–46); one-year OS was 67.7% and two-year OS 52.0%. Multimodal treatment was associated with significantly longer OS (median 45 vs. 12 months; log-rank p = 0.004). In multivariable Cox regression (n = 88; C-index 0.725), any objective treatment response (HR 0.245, 95% CI 0.101–0.593, p = 0.002), functional tumor status (HR 2.041, 95% CI 1.017–4.095, p = 0.045), and age at PC diagnosis (HR 1.018/year, p = 0.049) were independently associated with OS. Logistic regression identified multimodal treatment (OR 5.29, p = 0.009), any treatment response (OR 5.47, p = 0.006), and early metastasis (OR 0.18, p = 0.024) as independent predictors of long-survivor status. Ki-67 increased significantly from initial to PC-stage diagnosis (median 7% vs. 12%; Wilcoxon p < 0.001), which is, to our knowledge, the first demonstration of this longitudinal rise in a pooled pituitary carcinoma cohort. Conclusions: PC carries a poor prognosis, with a median OS of 19 months. Multimodal therapy and achievement of objective response are the strongest determinants of prolonged survival. Prospective registries with standardized molecular profiling are needed to advance precision management of this ultra-rare malignancy.