Progression-Free Survival Ratio As End Point for Phase II Trials in Advanced Solid Tumors TO THE EDITOR: Von Hoff et al 1 have investigated the potential of molecular profiling for the selection of treatments for patients with advanced solid tumors. These authors used the progression-free sur- vival (PFS) ratio as the end point of their phase II trial. The PFS ratio (also known as the time-to-progression ratio or growth modulation index) is defined for individual patients as the ratio of their PFS on the last line of therapy (in this case, a therapy selected through molecular profiling) to their PFS on the most recent previous line of therapy. 2 An attractive feature of this end point is that patients serve as their own control, which in general increases statistical sensitivity insofar as it elim- inates the between-patient variability. The premise that underscores the hypothesis testing framework for this design is that PFS tends to become shorter in successive lines of therapy. Therefore, a PFS ratio in excess of 1.0 (or, more conservatively, in excess of 1.3 to eliminate chance fluctuations) should be indicative of activity of the last line of therapy. However, the PFS ratio is a useful end point for phase II trials only under the assumption that there is a strong correlation between PFS 1 (the PFS on first-line therapy) and PFS 2 (the PFS on second-line therapy); patients with a good prognosis tend to have long PFS on all lines of therapy, whereas patients with a poor prognosis tend to have short PFS on all lines of therapy. As described by Mick et al, 3 the magnitude of correlation between the paired failure times is a key feature of the design. Through extensive simulations, these authors 3 demonstrated that reasonable power for the trial was only attainable given moderate to strong correlation between the paired failure times. We investigated within-patient variability by looking at the cor- relation between PFS 1 and PFS 2 using data from a randomized trial that compared the sequence folinic acid, fluorouracil, and irinotecan followed by folinic acid, fluorouracil, and oxaliplatin (FOLFIRI- FOLFOX) with the reverse sequence, FOLFOX-FOLFIRI, in patients with advanced colorectal cancer. 4 The advantage of this large trial is that PFS 1 and PFS 2 were measured prospectively using the same tech- niques, thereby avoiding problems associated with a retrospective, potentially unreliable, biased assessment of PFS 1 . 5 Figure 1 shows that in this large trial, there was no correlation to speak of between PFS 1 and PFS 2 (r = 0.17), which is in sharp contrast with the results of the pediatric study discussed by Mick et al, 3 in which the correlation was r = 0.56. The lack of correlation in the colorectal cancer trial was also seen in each randomized group (FOLFIRI/FOLFOX sequence, r = 0.22; FOLFIRI/FOLFOX sequence, r = 0.12). In this trial, 4 patients received effective therapies both in first-line and second-line treatments, and clearly under these circumstances the PFS ratio would not seem to be an attractive end point. Other investi- gators have successfully used the PFS ratio to document the activity of oxaliplatin in advanced colorectal cancer after failure of fluorouracil- based chemotherapy, 6,7 and the benefit of increasing the dose of ima- tinib (to 800 mg) after failure of a lower dose (400 mg) in patients with GI stromal tumors. 8,9 Whether and under what circumstances the PFS ratio will hold promise for future phase II trials in solid tumors are open questions. But quite apart from the issue of correlation between successive lines of therapy, the choice of PFS as an end point in each line of therapy is also debatable. Repeated measurements of tumor size over time (known as longitudinal data) are typically collected in trials that test new treatments for advanced disease, and from a statistical point of view, a model that uses all tumor size measurements for each patient may be preferable to a model that uses PFS, given that the latter design makes less efficient use of these data. Recent proposals to use changes in tumor size in the design of phase II trials could easily be extended to the situation of two successive lines of therapy, and the ratio of tumor growth rates could advantageously replace the PFS ratio while still making use of the within-patient comparison. 10,11 Von Hoff et al 1 have paved the way to innovative treatment approaches, although they acknowledge that randomized trials are now required to confirm their early findings. 5 We believe that better end points that are based either on tumor measurements and/or biomarkers will also be needed for such innovative approaches to make a convincing demonstration of their value in the clinical setting. Marc Buyse and Emmanuel Quinaux International Drug Development Institute, Louvain-la-Neuve, Belgium Alain Hendlisz and Vassilis Golfinopoulos Institut Jules Bordet, Brussels, Belgium Christophe Tournigand Ho ˆ pital Saint-Antoine, Paris, France Rosemarie Mick University of Pennsylvania, Philadelphia, PA 0 PFS 2 (months) PFS 1 (months) 20 15 10 5 5 10 15 20 25 Fig 1. Correlation between progression-free survival on first-line therapy (PFS 1 ) and on second-line therapy (PFS 2 ) in advanced colorectal cancer. Yellow circles represent the folinic acid, fluorouracil, and irinotecan followed by folinic acid, fluorouracil, and oxaliplatin (FOLFIRI-FOLFOX) sequence; blue circles represent the FOLFOX-FOLFIRI sequence. JOURNAL OF CLINICAL ONCOLOGY C O R R E S P O N D E N C E VOLUME 29  NUMBER 15  MAY 20 2011 © 2011 by American Society of Clinical Oncology e451 Journal of Clinical Oncology, Vol 29, No 15 (May 20), 2011: pp e451-e452 Downloaded from jco.ascopubs.org on March 9, 2016. For personal use only. No other uses without permission. Copyright © 2011 American Society of Clinical Oncology. All rights reserved.