Performance of the Palliative Prognostic Index for survival prediction in patients with and without cancer: a systematic review and meta-analysis

Background The Palliative Prognostic Index (PPI) is well studied in cancer; however, its performance in non-oncologic populations has not been systematically synthesised or formally compared with oncologic populations. We aimed to assess PPI performance in patients with oncologic and non-oncologic diseases for short-term (≤42 days) and long-term (≥90 days) survival prediction. Methods We included clinical studies assessing PPI accuracy identified through searches of PubMed, Embase, LILACS and Web of Science from inception to April 2026. Risk of bias was assessed using the Prediction model Risk Of Bias Assessment Tool (PROBAST). Meta-analyses, hierarchical summary receiver operating characteristic analyses and bivariate random-effects meta-regression were performed. Results Fifty-nine studies comprising 31 510 patients were included: 50 oncologic studies (28 862 patients) and 9 non-oncologic or mixed-population studies (2648 patients). Oncologic studies were included as the comparator group. For short-term prognosis, accuracy was 65% (58% to 73%) in non-oncologic and 73% (72% to 74%) in oncologic studies. Comparative analyses showed no significant differences (relative sensitivity: 0.94, 95% credible interval (CrI) 0.65 to 1.23; relative specificity: 0.93, 95% CrI 0.73 to 1.08). For long-term prognosis, accuracy was 70% (60% to 80%) in non-oncologic and 61% (47% to 75%) in oncologic studies. No significant differences were found (relative sensitivity: 1.98, 95% CrI 0.80 to 4.77; relative specificity: 0.86, 95% CrI 0.53 to 1.12). Analyses showed high heterogeneity, and evidence from exclusively non-oncologic populations was limited. Conclusions The PPI demonstrated reasonable accuracy for survival prediction, with no significant differences between oncologic and non-oncologic study populations. These findings suggest that PPI performance may be similar across these populations, although evidence in non-oncologic diseases remains limited. PROSPERO registration number CRD42022383369.