The availability of numerous tyrosine kinase inhibitors that have been approved by the US Food and Drug Administration for both firstand second-line management of advanced and metastatic renal cell carcinoma has empowered oncologists to ask a variety of questions that have immediate impact on patient care. These questions include: Which drug is best? Which drug should be used first? What is the proper sequencing of these novel agents? What is the first-line efficacy and subsequent efficacy after development of progressive disease? Are combinations of agents better than monotherapy? What are the relative safety profiles of each agent? Which agent is easier to use? Such questions are reminiscent of those posed historically when differing chemotherapeutic agents were first identified for the treatment of both solid and hematologic neoplasms. The important question being evaluated in this commentary, a question that is likely to herald a new and welcome frontier in optimizing therapeutic decision making, is: Which agents do patients prefer? And, second, in assessing such preferences, are the measurement metrics adequate, reliable, validated, and reproducible? Although these two questions, on the surface, should be simple and straightforward to answer, the lack of standardized methodologies to assess patient preferences as well as the heterogeneous procedures by which safety data are collected in large randomized studies make answering them a much more complicated matter. In the article that accompanies this editorial, Escudier et al report the results of a randomized, controlled, double-blind cross-over trial that evaluated patient preference for pazopanib versus sunitinib in patients with metastatic renal cell cancer. Importantly, some preliminary results of this study were alluded to in the recently published phase III study by Motzer et al that compared these two agents as first-line therapy and concluded that both had similar efficacy. The study by Escudier et al now provides more thorough data and complements the findings of the study by Motzer et al: that pazopanib boasts a better safety and quality-of-life profile. Taken together, the results of these two studies seem to demonstrate that pazopanib is preferred by both patients and physicians. A double-blind, cross-over design was used in which tyrosine kinase inhibitor–naive patients were randomly assigned to receive either pazopanib (800 mg per day for 10 weeks) followed by a 2-week washout phase (period 1), and then sunitinib 50 mg (4 weeks on, 2 weeks off) for 10 weeks (period 2), or the reverse sequence. A questionnaire that measured patient preference for a specific treatment at the end of the two treatment periods was the basis for the primary end point. More patients preferred pazopanib to sunitinib, citing less fatigue and better overall quality of life, whereas those preferring sunitinib cited less diarrhea as the reason for their treatment preference. Although Escudier et al was creative in designing and executing this complicated study, there are several weaknesses that must be considered before accepting the conclusions at face value. First and foremost, to meet prespecified and hypothesized end points that were thought to be meaningful, the study required a relatively small sample size of 160 patients. Unfortunately, even with this small, predefined sample, 33% of the randomly assigned patients were never evaluable for the primary end point because of withdrawals, adverse events, lack of efficacy, deaths, and other miscellaneous reasons, severely limiting meaningful prespecified analyses. The data on which the pazopanib preferences were based included a modified intention-to-treat population, again severely curtailing the robustness of the originally intended analyses. In addition, one of the most important adverse events leading to patient preference was fatigue. Again, the difficulty in measuring this adverse event underscores the inherent problems that sponsors face in both collecting and evaluating safety data; more precision is needed in the coding for safety issues and in the nomenclature related to safety and adverse events. For example, the incidence of any grade of fatigue in the study by Motzer et al was 55% for pazopanib and 63% for sunitinib; asthenia was not even listed. In the study by Escudier et al, the numbers for fatigue and asthenia for pazopanib are 29% and 16%, and for sunitinib, 30% and 24%. Evaluation of global data from the approved product labels for indications other than renal cancer (using comparable dosing as used for renal cell carcinoma, with larger representative patient numbers than in the study by Escudier et al) underscores the difficulties in assessing comparable safety terms. For example, for the use of sunitinib in GI stromal tumors, the dictionary of adverse event terms reports the frequency of asthenia, but makes no mention of fatigue. Fatigue is reported with the use of pazopanib for soft tissue sarcoma, but no mention of asthenia appears. Interestingly, the fatigue percentage JOURNAL OF CLINICAL ONCOLOGY E D I T O R I A L VOLUME 32 NUMBER 14 MAY 1