|
Advertisement |
|
|
||
 |
|
|||||
|
|||||
|
 |
||||||
|
||||||
|
||||||
|
Journal of Clinical Oncology, Vol 24, No 24 (August 20), 2006: pp. 3821-3822 © 2006 American Society of Clinical Oncology. DOI: 10.1200/JCO.2006.06.9856
Some Thoughts on the Reporting of Adverse Events in Phase II Cancer Clinical TrialsNational Surgical Adjuvant Breast and Bowel Project Biostatistical Center, University of Pittsburgh Graduate School of Public Health, Pittsburgh, PA In this issue, Scharf and Colevas1 discuss discrepancies in adverse event (AE) reporting for phase II oncology trials in peer-reviewed publications and the National Cancer Institute (NCI) Clinical Data Update System (CDUS).2 Many published articles do not provide an appropriate lexicon for AE assessment, and published reports of AEs in phase II studies are problematic in the following three areas: underreporting of low-grade AEs, underreporting of recurrent AEs, and inconsistent and incomplete characterization and reporting of high-grade AEs.1 The authors conclude that there is a lack of consistency between the definition of AEs in published trial results compared with the original study protocols, and published AE data markedly differ from the data available in CDUS. The authors suggest that journals set a standard of quality assurance for AE information. The authors based their conclusions on an analysis of 22 published reports of single-agent CDUS-monitored phase II clinical trials, culled from a potential pool of 355 single- and multiple-agent CDUS-monitored phase II studies active between March 1998 and October 2003. In that time, the NCI Common Toxicity Criteria version 2.03 was the mandated lexicon for reporting and grading AEs in NCI-sponsored trials. Nevertheless, Scharf and Colevas1 point out that only 16 (73%) of the 22 publications identified a specific AE criteria system, and only 13 publications (59%) cited the NCI Common Toxicity Criteria as the method of AE classification.
However, the main focus of the analysis involves a comparison of the AEs reported in the publications versus the AEs reported in the CDUS database. When all grade 3 to 5 AEs were considered, regardless of whether or not they were attributable to an agent, there were 611 grades 3 to 5 AEs in the CDUS database compared with 413 reported in publications. The authors do not mention if these differences are distributed relatively evenly across the 22 studies or if a small proportion of the studies account for the bulk of the differences. Most of the rest of the analysis is focused on AEs that were agent attributable, as summarized in Figures 2A and 2B. Figure 2A shows that the reporting of AEs of grade These results from Scharf and Colevas1 raise questions about quality assurance in the peer-reviewed reporting of safety information in phase II clinical trials. However, their methods and conclusions warrant further comment. First, the methods do not address whether the 22 single-agent studies included in the analysis are representative of all published phase II trials. In fact, Mahoney et al4 have reported that single-agent phase II trials have higher percentages of AEs unrelated to the study agents than observed in multiagent phase II clinical trials. In their first conclusion regarding the underreporting of low grade AEs in phase II trials, Scharf and Colevas1 do not report the disease stage of the patients in phase II clinical trials and the nature of the agents being tested. In most phase II oncology studies, aggressive therapies are piloted in patients who may have metastatic cancer to demonstrate improved response rates, disease-free survival and survival. Length constraints of journal articles force authors to prioritize the reporting of information to the most clinically important information about treatment outcomes. Therefore, when reporting the results of phase II trials, most investigators emphasize end points which exclude non life-threatening AEs. Importantly, most therapies being tested in phase II studies cause low-grade AEs in a high proportion of patients. Consequently, some authors have proposed strategies which reduce the collection of some low-grade AEs in phase II settings because of the amount of resources required to collect this informaton.5 Furthermore, Mahoney et al4 assert that the stage of drug development associated with the stage of the disease of the patients being studied may influence the ability to recognize drug-induced reactions. A more thorough analysis of the relationship between low- and high-grade AEs and disease-free and overall survival could perhaps be better achieved in the phase III setting where patient populations are larger and more information about agent attribution is available before the beginning of the studies. The authors' second conclusion relates to the underreporting of recurrent AEs. This is a valid point because recurrent AEs are often not evaluated in published reports. To reliably characterize recurrent AE data, one must construct precise definitions of what is meant by recurrent. Even with appropriate definitions, the characterization of recurrent AEs can be complicated because AE grades can vary over the courses of therapy as the doses of the agent(s) being studied are modified. The third conclusion by Scharf and Colevas1 concerns the incomplete reporting of severe AEs (SAEs). Three points should be made regarding this conclusion. First, it should be emphasized that the CDUS and other AE databases change over time. This issue was not addressed by the authors. The CDUS database is updated quarterly,2 and updates can continue after initial reports of results have been published. Quarterly changes in the database are a result of new AE information, ongoing medical reviews of previously reported AEs, resolution of the attribution of previously reported AEs, or updated delinquent information. Also, in journal articles, the time between analysis of AE data and actual publication of the results can be lengthy. Hence, small discrepancies between the published reports and the CDUS database at the time of publication may occur. Even larger differences may occur between current information in the CDUS database and reports published months or years previously. Therefore, all published reports of safety information and subsequent evaluations of those reports should indicate the cutoff date on which the AE information is based. A second point regarding the inconsistent reporting of SAEs is the lack of resources available at coordinating or data collection centers, leading to difficulty in the assessment of AE attribution to study agents and contributing to variation in the reporting of SAEs over time. For example, Mahoney et al4 reported that, in a review of 24 phase II and two phase III trials, 11% of 75,598 routine AEs required queries for missing information about the grade or attribution of AEs. Furthermore, the consistency of reviewers' attribution assessments has been shown to be poor in some clinical trials,6,7 contributing to the burden of assessing agent-induced reactions in a timely and reliable fashion. Third, some authors tend to only include those AEs that, in their clinical judgment, are most relevant to a patient's well-being. This approach may eliminate information that is subjective, again leading to underreporting of AE information. As suggested, a link from a journal article to an online toxicity table could alleviate this problem. However, such information should be limited to the greatest toxicities experienced only for attributable AE categories because, as indicated by Mahoney et al,4 a table summarizing all AEs regardless of their attribution may be voluminous and unreliable. Accordingly, any table made available for public dissemination should be limited to prespecified categories that are reliably assessed. Also, links to recurrent toxicities would be cumbersome to implement. Finally, the authors suggest that journals should provide an extra level of quality assurance. This suggestion is appropriate, but it is not clear that publication resources are available to do this or that journals are responsible for quality assurance of submitted information. Any mandated standard would have to be constructed with care, accommodating different trial designs and striking a balance between the necessity of reporting patient safety data and the burden of collecting and potentially publishing unnecessary, clinically irrelevant or biased AE information. In conclusion, Scharf and Colevas1 aptly point out that, in peer-reviewed reports of phase II results, it is the primary responsibility of investigators to provide a well-defined lexicon for the definition of AEs and accurate reporting of AEs. However, published reports should also provide the most balanced perspective on how the contribution of AE information relates to the overall assessment of the clinical benefit of promising agents for the treatment of patients in phase II cancer trials. Author's Disclosures of Potential Conflicts of Interest The author indicated no potential conflicts of interest. ACKNOWLEDGMENTS Supported by Public Health Service Grants No. U10-CA-69651 and CA-69974 from the National Cancer Institute, Department of Health and Human Services. REFERENCES
1. Scharf O, Colevas AD: Adverse event reporting in publications compared with sponsor database for cancer clinical trials. J Clin Oncol 24: 3933-3938, 2006 2. National Cancer Institute: Clinical Data Update System (CDUS), 2003. http://ctep.cancer.gov/reporting/cdus.html 3. National Cancer Institute: Common Toxicity Criteria v2.0 (CTC), 1999. http://ctep.cancer.gov/reporting/CTC-3.html 4. Mahoney MR, Sargent DJ, O'Connell MJ, et al: Dealing with a deluge of data: An assessment of adverse event data on North Central Cancer Treatment Group trials. J Clin Oncol 23: 9275-9281, 2005 5. Goldberg RM, Sargent DJ, Morton RF, et al: Early detection of toxicity and adjustment of ongoing clinical trials: The history and performance of the North Central Cancer Treatment Group's real-time toxicity monitoring program. J Clin Oncol 20: 4591-4596, 2002 6. Thomas EJ, Studdert LLB, Brennan TA: The reliability of medical record review for estimating adverse event rates. Ann Intern Med 136: 812-816, 2002 7. Rothenberg ML, Meropol NJ, Poplin EA, et al: Mortality associated with irinotecan plus bolus fluorouracil/leucovorin: Summary findings of an independent panel. J Clin Oncol 19: 3801-3807, 2001 This article has been cited by other articles:
|
|||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
|
|
|||||||||||
|
|||||||||||
|
|
Copyright © 2006 by the American Society of Clinical Oncology, Online ISSN: 1527-7755. Print ISSN: 0732-183X
|
3 in articles differs somewhat from the information they found in the CDUS database. Because of the small number of grade 3 to 5 AEs in many studies, some of the seemingly large relative differences in the text represent small differences in terms of absolute numbers. Interestingly, in eight of the 22 publications, investigators reported more attributable grade 3 to 5 AEs than were present in the CDUS database, whereas 12 others reported less attributable grade 3 to 5 AEs than were present in the CDUS database; two studies had concordant information. In Figure 2B, when all AE grades more than 0 are considered, significant underreporting of AE information was provided in the peer-reviewed articles. When Figures 2A and 2B are viewed together, one concludes that much of the underreporting of the AE information is a result of either not reporting or vastly underreporting grades 1 and 2 AEs. 
