Laleh Foroutani, MD (she/her/hers)
Postdoctoral Research Fellow
University of California, San Francisco
San Francisco, California, United States
Andrew Gonzalez, MD
General Surgery Resident
University of California, San Francisco
San Francisco, California, United States
Sophia Hernandez, MD
General Surgery Resident
University of California, San Francisco
San Francisco, California, United States
.jpg "Jane Wang, MD (she/her/hers) photo")
Jane Wang, MD (she/her/hers)
Resident
University of California, San Francisco
San Francisco, California, United States
Thomas M. Li, MD
General Surgery Resident
University of California, San Francisco
San Francisco, California, United States
Lucia M. Calthorpe, MD
General Surgery Resident
University of California, San Francisco
San Francisco, California, United States
Kenzo Hirose, MD
Professor of Surgery
University of California, San Francisco
San Francisco, California, United States
Eric K. Nakakura, MD
Professor of Surgery
University of California, San Francisco
San Francisco, California, United States
Carlos U. Corvera, MD
Professor of Surgery
University of California, San Francisco
San Franisco, California, United States
Adnan Alseidi, MD
Professor of Surgery
University of California, San Francisco
San Francisco, California, United States
M. Haroon A. Choudry, MD
Associate Professor
University of Pittsburgh Medical Center
Pittsburgh, Pennsylvania, United States
Mohamed A. Adam, MD
Assistant Professor of Surgery
University of California, San Francisco
San Francisco, California, United States
Laleh Foroutani, MD (she/her/hers)
Postdoctoral Research Fellow
University of California, San Francisco
San Francisco, California, United States
Malignant peritoneal mesothelioma (MPM) is a rare, aggressive malignancy characterized by diffuse peritoneal spread, making conventional TNM staging frameworks impractical. Current NCCN guidelines acknowledge the lack of an accepted staging system for MPM but suggest a staging model based on the Peritoneal Cancer Index (PCI). This study developed an advanced data-driven algorithm integrating PCI with additional clinicopathologic variables to improve prognostication and staging.
Methods:
Patients with MPM who underwent cytoreductive surgery with hyperthermic intraperitoneal chemotherapy at a single center (2001-2022) were included. Recursive feature elimination was used to objectively select prognostic variables. Recursive partitioning with conditional inference survival trees was employed to develop a hierarchical staging model that divided patients into subgroups with similar overall survival (OS), while maximizing OS differences between groups. The newly derived staging groups were then compared with the existing system proposed by Yan et al.: Stage I [PCI ≤10/N0/M0]; Stage II [PCI 11–30/N0/M0]; Stage III [PCI >30 or N1 or M1].
Results: Of the 172 patients with MPM included in the analysis, 63% were women. Median age was 60 years; median PCI was 17. Epithelioid histology accounted for 86% of cases. After adjustment, higher mitotic rate, biphasic/sarcomatoid histology, higher PCI, and symptom progression were independently associated with worse OS (all p< 0.05). Survival algorithm objectively produced four distinct prognostic stages with divergent OS (p< 0.01): Stage I: [low mitotic rate (≤3)/PCI ≤10]; Stage II: [low mitotic rate/PCI >10]; Stage III: [high mitotic rate ( >3)/epithelioid subtype]; Stage IV: [high mitotic rate/biphasic/sarcomatoid subtype]. Median OS by the tree-based model was not reached for Stage I; 42 months for Stage II; 22 months for Stage III; 5 months for Stage IV. In comparison, the traditional system showed median OS of 95 months for Stage I; 29 months for Stage II; and 42 months for Stage III; confirming superior risk stratification and better survival separation with the developed model (Figure 1).
Conclusions: The current PCI-based staging framework for MPM appears to be suboptimal in discriminating risk of OS among disease stage groups. Our results suggest that the current TNM staging system could be improved with the newly identified clinicopathologic variables to achieve more accurate prognostic stratification and possibly treatment selection.