Clinical predictors of prolonged clinical benefit (PCB) from pemetrexed (P) therapy in metastatic non-small cell lung cancer (mNSCLC).

Abstract

e18056 Background: P is an effective agent in mNSCLC, however many patients (pts) get minimal or no benefit from P while a subset has PCB. To date there are no established models to guide P therapy. Methods: From 9/2004 to 12/2008, 135 pts with mNSCLC were treated with P at a single institution. Pts were retrospectively analyzed for clinical predictors associated with PCB from P using chi-square and Fisher’s exact tests, and logistic regression models. Variables assessed included age, gender, race, smoking history, ECOG performance status (PS), stage at initial diagnosis, histology, number (n) and sites of metastatic disease, n of treatment lines, interval from metastasis (M) to initiation of P, timing of P in terms of lines of therapy (1 through 5), n of P cycles, toxicity (tox) and creatinine clearance pre- and post P therapy. Results: Median age was 63 years (34-84). P was given for a median of 3 cycles (1-66) with the majority of pts receiving it as 2nd line therapy (59%) and as a single agent (85%). There were no statistically significant differences between the different lines of treatment with respect to patient characteristics, toxicity, objective response, progression-free survival (PFS) or overall survival (OS). Overall 28% of pts achieved partial response, and median PFS and OS from the time of initiation of P were 2.9 and 9.0 months (mos), respectively. PCB was defined as PFS ≥12 mos whereas limited CB (LCB) was defined as PFS ≤6 mos. 14% of pts had PCB; 59% had LCB. In multivariable analysis of PCB vs LCB, that initially considered all the factors described above, ≤ 2 sites of M (p=.005), time from M to P therapy ≥ 12 m (p=.02) and PS ≤1 (p=.05) were independent predictors of PCB. Using these factors prognostic groups were formed based on the n of favorable features (FF) present; pts with ≤2 sites of M, ≥12 mos interval from M to P, and PS≤1 had a 69% PCB rate compared to 41% for pts with 2 FF and 2% for pts with 0-1 FF. Conclusions: No clinical difference was seen based on line of therapy. Limited sites of M, prolonged time to initiation of therapy, and PS were predictive of PCB from P and could be used to define 3 distinct prognostic groups. These findings however will need to be validated in an independent cohort of pts.