Abstract
Introduction: Sunitinib is a standard second-line treatment in advanced gastrointestinal stromal tumours (GIST). We aimed to search for predictive factors for grade 3 and 4 toxicity, progression-free survival (PFS) and overall survival (OS) in a GIST reference center patient population, outside clinical trials.Methods: A retrospective analysis was performed of patients treated in two European Comprehensive Cancer Centers between January 2005 and December 2015. Demographic and clinical features, tumour characteristics and biological parameters were investigated. Logistic regression models were used to find factors associated with grade 3 and 4 toxicity. To identify predictive factors for PFS and OS, variables that were statistically significant in univariate analysis were used in the multivariate Cox proportional hazards model.Results: Ninety-one patients were included in this analysis. Age >60 years (HR 5.0, p = .006) and body weight ≤70 kg (HR 4.7, p = .009) were predictive factors for grade 3 and 4 toxicity. When divided into two categories, non-haematological grade 3 and 4 toxicity was predicted by age >60 years (HR 3.8, p = .012) and body weight ≤70 kg (HR 3.3, p = .025) whereas haematological toxicity had no significantly associated predictive factors. The median PFS and OS with sunitinib were 8.8 months and 27.5 months, respectively. The use of imatinib less than six months compared to 6-12 months (HR 0.2, p = .013) and to >12 months (HR 0.3, p = .016) and liver and/or peritoneal metastases (HR 0.1, p < .001, HR 0.2, p = .003 and HR 0.2, p = .004) compared to locally advanced disease only were predictive for longer PFS. High neutrophil (HR 3.1, p = 0.04) and platelet count (HR 2.4, p = .046) predicted a shorter OS. Flexible sunitinib dosing was associated with superior OS (p = .021).Conclusion: In advanced GIST patients treated with sunitinib, older and low-weight patients are at risk for grade 3 and 4 toxicity. Clinical (prior imatinib use and metastases), biological (neutrophil and platelet count) and treatment characteristics independently predict PFS and OS.