|
|
||
 |
|
|||||
|
 |
||||||
|
||||||
|
||||||
|
Originally published as JCO Early Release 10.1200/JCO.2005.11.910 on February 14 2005 © 2005 American Society of Clinical Oncology.
Standards of Proof, Standards of Practice, and Proof of Standards: A Tale of Two TrialsTom Baker Cancer Centre and University of Calgary, Calgary, Alberta, Canada In this issue of the Journal of Clinical Oncology, Chang et al1 and Witzig et al2 report two randomized controlled trials evaluating the role of recombinant human erythropoietin (epoetin alfa) for the management of anemia in cancer patients receiving chemotherapy. Previous rigorous trials and systematic reviews of randomized trials have demonstrated the efficacy of this agent in maintaining hemoglobin levels and reducing transfusion requirements in cancer patients receiving chemotherapy.3-5 Recommendations from high-quality, evidence-based clinical practice guidelines confirm the role for epoetin alfa in anemia management in this clinical setting.4,5 The drug is already commonly used in practice. Furthermore, the public and patients welcome the availability of a therapeutic option that can reduce exposure to what they perceive as the risks of blood transfusion. Both trials1,2 reported in this issue confirm, yet again, that epoetin alfa works in achieving the objectives of maintaining hemoglobin levels and reducing transfusion requirements in cancer patients receiving myelosuppressive therapy. So, what do these trials add to our current knowledge, and how should they affect routine practice? Both trials indirectly address the issue of drug scheduling. Whereas virtually all previous randomized trials tested epoetin alfa using an inconvenient three-times-per-week schedule, clinicians had already migrated towards the more convenient weekly schedule, and the wording of recommendations from clinical practice guidelines reflected this situation.4 Now, as a result of these trials, we can confirm both the efficacy of weekly dosing in maintaining hemoglobin levels and the reduction of transfusion requirements across a spectrum of cancer patient populations undergoing chemotherapy. Considering the current clinical context of the usage of routine epoetin alfa and the total body of available evidence, it would seem appropriate for updates of current guidelines and for future guidelines to endorse the weekly schedule for hemoglobin maintenance without demanding the higher standard of proof, which would involve a direct comparison to three-times-per-week treatment. The trial by Witzig et al2 also addressed the clinical utility of a tool for better selection of patients who might benefit from epoetin alfa treatment. This is important because one of the practical difficulties of using epoetin alfa is that, once it is started, the clinical effect takes several weeks to be observed. This means that timing the initiation of therapy must be done carefully and proper selection of those patients who can truly benefit from treatment should be performed because not all patients will develop a degree of anemia requiring intervention. The secondary question posed by this trial is specifically targeted to a corrective intervention strategy for anemia. The trial by Chang et al1 addressed a clinical question with more far-reaching implications, involving an expanded indication for epoetin alfa as a preventative, as opposed to corrective, intervention to manage anemia. Waiting for significant decreases in hemoglobin after initiation of myelosuppressive therapy to determine a patient’s eligibility for treatment with epoetin alfa may unnecessarily expose patients to an increased amount of time with symptoms of anemia. The preventative approach can be expected to minimize the burden of anemia symptoms by reducing the time spent in a symptomatic state. The Chang et al study included a relatively homogeneous population of breast cancer patients (79% receiving adjuvant chemotherapy) who were randomly assigned to receive the experimental therapy (epoetin alfa once weekly) or best supportive care when the hemoglobin decreased to or just below 12 g/dL. This threshold for initiation of treatment is higher than that used in previous trials (and in the study by Witzig et al2). Given this potential expansion of clinical indications for this relatively costly drug, it seems appropriate that the standard of proof for clinical benefit for any trial ought to meet two more stringent criteria. First, the trial should include outcomes that are more clinically compelling than the hemoglobin level alone, as the trials in this issue1,2 do. Second, the choice of control group should provide a direct comparison with the more traditional corrective strategy tested in previous trials and use strategies such as blinding to minimize bias in the assessment of clinically important outcomes. To satisfy the first criterion, the Chang et al1 trial included quality of life (QOL) as the primary outcome, using validated instruments that were demonstrated previously in a rigorous trial to detect statistically significant and presumably clinically important improvements for epoetin alfa over placebo, albeit in a population with more advanced disease.6 The Chang et al trial demonstrated that, compared with best supportive care, once weekly epoetin alfa, when used as a preventative strategy, improves QOL, effectively maintains hemoglobin level, and reduces transfusion requirements. However, the publication strategy that juxtaposes the Chang et al1 trial with that of Witzig et al2 raises other important issues about standards of proof and proof of standards when determining how clinical trials ought to affect changes in standards of practice. The Chang et al trial falls short of meeting the standard of proof for influencing the standard of practice beyond weekly dosing because of the choice of control group. First, payers and providers are likely to want evidence of at least equivalent benefits in QOL from a direct comparison of the preventative strategy with the current corrective standard before accepting expansion of the clinical indications for this relatively expensive agent. Second, the discordance between the trial results in relation to the QOL outcomes raises the specter of potential bias for an open-label positive-result study in the face of a negative-result placebo-controlled trial, despite the differences in the patient populations that participated in the two trials. Readers contrasting these two trials will wonder whether the use of an open-label design in the former study contributed to biases that could have led to a positive-result study. Although investigators of recent studies have questioned the therapeutic power of the placebo effect itself, they do acknowledge its importance as a methodologic tool to minimize bias in controlled clinical trials, especially when subjective end points are being evaluated.7,8 This dilemma leads to consideration of the proof of the standard for measuring QOL in these studies. Recent systematic reviews of the efficacy of epoetin alfa have pointed to the inconsistencies in the results of randomized trials with respect to QOL outcomes.3-5 Proof of standard in the QOL measurement field refers to the process of validation, for which there are established psychometric methods. Fortunately, there is considerable overlap in the QOL instruments used in these two studies and in the battery of instruments reported in the well-known positive study by Littlewood et al.6 But an instrument that is validated in one population does not necessarily perform as well in a population with different characteristics. Therefore, readers might reasonably attribute the discordance of the QOL results in the two trials in this issue1,2 as reflecting differences in the populations studied, with patients in the Witzig et al2 trial having much more advanced disease across a more heterogeneous group of disease sites. However, the population studied in the Littlewood et al trial (placebo-controlled, positive result) also included patients with advanced disease across a spectrum of disease sites and more closely resembles patients in the Witzig et al trial (placebo-controlled, negative result). Furthermore, the way epoetin alfa was used as an anemia-corrective strategy in the trials by Littlewood et al and Witzig et al was similar. Thus, given the available body of evidence, it would be difficult to conclude that the difference in QOL outcomes in the trials by Chang et al1 and Witzig et al is due mainly to the advantages of the preventative strategy tested in the former study or to the differences in the populations studied. These observations raise issues about the outcome measures themselves. QOL investigators are concerned about how to approach the analytic problems associated with data collected prospectively at multiple time points and the inevitable challenge of missing data. The two trials in this issue1,2 attempt to minimize such potential biases by carefully defining the different populations eligible for analyses of the different outcomes of interest. The analysis was simplified in the Chang et al1 trial by comparing changes in QOL scores from baseline to a time point at 12 weeks, with treatment assignment as the main factor corrected for baseline QOL score as a covariate. The analytic strategy for the trial reported by Witzig et al2 seems less straightforward because patients completed QOL assessments monthly for 16 weeks, and it is unclear to this reader how the different scores were combined to yield mean QOL scores after treatment. Despite an analytic approach designed to include as many relevant patients as possible in the prospective QOL assessments of patients in the Witzig et al trial, overall QOL was not improved. However, it is encouraging that well-done comparative studies of epoetin alfa therapy from different groups are beginning to use the same battery of QOL measures, thus enhancing the likelihood that QOL investigators will soon resolve the methodologic issues plaguing the measurement of QOL in prospective trials using longitudinal data with multiple time points. Given the inconsistent results of previous trials in demonstrating improved QOL or symptom improvement associated with epoetin alfa therapy and given the results of the trial reported by Witzig et al,2 Chang et al1 probably regret not using a placebo control arm or doing a direct comparison of preventative epoetin against the current standard of practice. Despite the apparent high quality and transparency of methods in the Chang et al trial in all other respects, failure to use as rigorous a control arm as possible under the circumstances is even more regrettable because of concerns raised through well-documented studies suggesting the risk of systematic bias for trials supported and conducted by industry.9 Standards of proof need to be higher for changing standards of practice, especially when expanding the market for costly therapies where acceptable alternatives exist. Author's Disclosures of Potential Conflicts of Interest The author indicated no potential conflicts of interest. REFERENCES
Related Articles
This article has been cited by other articles:
|
|||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
|
|
|||||||||||
|
|||||||||||
|
|
Copyright © 2005 by the American Society of Clinical Oncology, Online ISSN: 1527-7755. Print ISSN: 0732-183X
|
