A risk-based individualized follow-up after complete surgery as an effective procedure to reduce the relapse (R) impact in GIST patients (pts).
10552 Background: FU care poses a burden on both pts and health system. FU aims to precociously identify recurrences, metastases or treatment-related adverse events so to undertake the appropriate therapy as soon as possible. Guidelines (NCCN, ESMO) admit lack of knowledge on optimal surveillance, but suggest FU based on experts’ opinion and risk stratification. Moreover, tumor burden (TB) is a well known negative prognostic factor (Van Glabbeke 2005). Therefore, low-TB at R might affect final outcome. To identify the impact, if any, of regular FU, we examined our prospectively collected database looking for Rs in which the early detection affected both clinical management and, likely, the outcome. Methods: 140 pts were stratified (AFIP classification). High risk (HR) pts had complete history + physical examination (H&P) and CT every 3 mos for 2 years, 4 mos in 3rd year, 6 mos in the 4th and 5th yrs, yearly thereafter; intermediate (IR), low (LR) and very low risk (VLR) pts had H&P + CT every 4 mos for 2 years, every 6 mos up to the 5th year, then yearly. Rs were divided in: low-TB + completely resectable R (Group 1) and high-TB + disseminated R (Group 2). The number of CT needed to detect (NND) one R and the incidence of early (<6 mos) R was calculated. Overall survival (OS) was estimated by Kaplan-Meier method. Results: In 73 male and 67 female, median age 63 (23-82), risk stratification was: HR 73, IR 31, LR 28, VLR 8. After a median FU of 63 mos we observed 58 Rs: 25 (43%) and 33 (57%) in group 2 and 1, respectively. Relapsed pts genotype was KIT Ex11 72%; Ex9 10%; Ex13 3%; wild-type 15%. Median time to R was 16 mos (1-120). 21 pts underwent surgical resection of their R. 16 pts remained free from progression for a median of 90 mos. We detected 15 (26%) early Rs. NND was 17 and 32 CT for HR and non-HR pts, respectively. Group 1 and 2 median estimated OS was 112 and 87 mos (p=.05), respectively. Conclusions: Though retrospective, this series shows that FU may detect low-TB Rs. In principle, this might affect pts final outcome and, therefore, justify this costly effort. Since it is difficult to foresee a prospective randomized trial, a confirmation of these data in a different series might increase their reliability.