|
Advertisement |
|
|
||
 |
|
|||||
|
|||||
|
 |
||||||
|
||||||
|
||||||
|
Journal of Clinical Oncology, Vol 23, No 36 (December 20), 2005: pp. 9275-9281 © 2005 American Society of Clinical Oncology. DOI: 10.1200/JCO.2004.00.0588
Dealing With a Deluge of Data: An Assessment of Adverse Event Data on North Central Cancer Treatment Group TrialsFrom the Mayo Clinic and Mayo Foundation, Rochester, MN; Allegheny Cancer Center, Pittsburgh, PA; The University of North Carolina at Chapel Hill, Chapel Hill, NC; and Toledo Community Hospital Oncology Program CCOP, Toledo, OH Address reprint requests to Michelle Mahoney, MS, Mayo Clinic, 200 First St SW, Rochester, MN 55905; e-mail: mahoneym{at}mayo.edu
PURPOSE: Adverse events (AEs) are monitored in clinical trials for patient safety, to satisfy reporting requirements, and develop safety profiles. Recently, much attention has been placed on the reporting of serious AEs (SAEs) that are either life threatening or lethal in clinical trials. However, SAEs comprise a small subset of all AE data collected for trials; the majority of AE data collected are routine AEs (RAEs) regarding non–life-threatening events. We assessed the utility of the RAE data collected, relative to the volume. PATIENTS AND METHODS: We surveyed the RAE data from 26 North Central Cancer Treatment Group coordinated trials. RESULTS: A total of 8,318 (11%) of 75,598 of RAEs required queries. Of these, 86% were protocol-required RAEs, 83% of RAEs required per protocol were within normal limits (eg, platelets) or not present, and 61% of extra AEs were mild. One fifth of RAEs were considered unlikely to be related or unrelated to treatment. Overall, 3% of events were severe, life threatening, or caused death. Only 1% of RAE data reported required expedited reporting (eg, via Adverse Event Expedited Reporting System). Results indicate that 72% of RAEs would be eliminated if only the maximum severity per patient and type were required. These results were validated in a large phase III trial. CONCLUSION: The majority of RAEs identified, transcribed, and entered are not clinically important. Our data suggest that reducing the number of AEs monitored will affect substantially neither overall patient safety nor compromise evaluation of regimens undergoing testing. We present several considerations for such a reduction in data collection, as well as a policy that we have used to address the deluge of RAE data.
During the last decade, the conduct and monitoring of the safety in clinical trials has received great attention. Issues relating to these serious, life-threatening, or adverse events (AEs) causing death (ie, serious adverse events [SAEs]), and potential improvements to the current system(s) have been discussed vigorously. Clearly, careful monitoring of SAEs is mandatory for safe and ethical conduct of clinical trials. However, it must be recognized that SAEs comprise only a small subset of all AE data collected in a clinical trial. The focus of this report is the routine AE (RAE) data collected for clinical trials, which covers all AEs, and of which SAEs comprise a smaller subset. Several criteria exist for standardizing and objectifying the classification and assessment of both SAEs and RAEs.1,2 Our analysis is focused on phase II and III trials, a majority of which are noninvestigational agents (described in the Results). The National Cancer Institute (NCI) Common Toxicity Criteria (CTC) Version 1.0 (V1.0) was implemented in 1982, seeking to standardize the collection of AEs deemed at least possibly related to treatment. In 1998, V2.0 was implemented, standardizing data collection further and expanding the number of AE categories (eg, vomiting, diarrhea). The most notable update associated with CTC V2.0 was the addition of attribution (ie, relationship) to study treatment. The addition of attribution provided an opportunity to collect AEs deemed unrelated or unlikely to be related to treatment, which previously were documented in the patient medical history, but not collected or simply noted as a comment (free-text v predefined values) via CTC V1.0 for clinical trials research purposes. NCI Common Terminology Criteria of Adverse Events V3.0 has been implemented recently, with an additional increase in the number of classifications. The conduct and collection of data for clinical trials have been highlighted recently.3,4 In the multi-institutional and community-based setting, the procedures for collecting RAE data involve considerable effort and resources. The local treating physician evaluates a patient for AEs, reporting SAEs rapidly to regulatory entities (eg, NCI, US Food and Drug Administration) via an expedited reporting mechanism (eg, Adverse Event Expedited Reporting System,5 Vaccine Adverse Event Reporting System,6 and/or MedWatch7). At a later date, a clinical research associate (CRA) abstracts all AEs from each patient's medical history, and transcribes the AE data onto case report forms (CRFs). These data reported via CRFs (which may be either paper or electronic8) are what we refer to as RAEs, and include any previously reported SAEs. The data from these CRFs are then managed at a central data management office. Shortly after the implementation of CTC V2.0, questions arose based on the addition of attribution, definitions, processes used, and workload necessary to collect RAE data. Specifically, the following questions were included. Do the additions of attribution and the expansion of the definition of an AE affect the volume of data collected? What is gained by the addition of collecting unrelated and unlikely to be related AEs? How do these additions and changes influence the workload of the CRAs and central office personnel in terms of data collection, clean-up, and monitoring. For example, determining the attribution of RAEs often requires several interactions between the CRA, oncology nurse, pharmacist, and treating physician. As another example, there are different philosophies among researchers regarding the inclusion of unrelated and unlikely related AEs in monitoring trials. To address these concerns, we set out to ensure that we were collecting the RAE data necessary to address patient safety and study end points; identify the relative utility of the information the process of identifying, transcribing, and entering RAE data produced; and determine whether we were collecting too much RAE data.
The North Central Cancer Treatment Group (NCCTG) is one of the NCI's cooperative cancer treatment research groups, and conducts primarily phase II and III clinical trials in the community setting. It has more than 30 membership institutions, affiliated with more than 315 community-based treating locations throughout the United States, Canada, and Mexico. The Mayo Clinic Comprehensive Cancer Center serves as the research base for the NCCTG. To further understand the process of collecting RAE data on NCCTG trials, we describe the two distinct tasks CRAs must perform at the time of data abstraction and transcription. The first task is to identify the RAEs that have been prespecified for monitoring by the treatment protocol. These RAEs are preprinted onto CRFs based on the known AE profile of the agent(s) under study. Even though such events are known to be attributed to the study agents, they continue to be monitored for additional incidence and prevalence information in larger populations of patients. All grades (ie, 0 to 5) and attributions (defined in the next paragraph) are recorded for the prespecified RAEs. The second task of the CRA is to review the remainder of the patient record for RAEs experienced beyond the prespecified list. Here, the RAE type(s) are handwritten onto our CRFs. We note for completeness that in mid-2003, our CRAs began using a remote data entry system,8 which will eventually eliminate paper forms and centralize data entry. However, the process of reviewing the patient chart to identify RAEs will remain identical, with the CRA entering data (versus transcribing onto a CRF). For classification of RAEs, the following definitions are provided within the CTC V2.0 manual.9 AE is defined as any unfavorable or unintended symptom, sign, or disease (including an abnormal laboratory finding), temporally associated with the use of a medical treatment or procedure that may or may not be considered related to the medical treatment or procedure. Grades are assigned for each AE (eg, grade 0 = not present or within normal limits, grade 1 = mild, grade 2 = moderate, grade 3 = severe and undesirable, grade 4 = life threatening or disabling, grade 5 = death). For each AE, the attending physician or clinician in conjunction with the research nurse who examined and evaluated the patient should assign attribution (ie, 1 = clearly not [unrelated], 2 = doubtfully [unlikely], 3 = may be [possibly], 4 = likely [probably], or 5 = clearly [definitely] related to the investigational agents).
We assessed the RAE data via simple summary statistics (eg, mean, median, quartiles) and graphical methods. Frequency tables were used to explore the association of the distribution of variables on the basis of classifications of the agents (ie, commercial v investigational) or disease setting.
Samples and Characteristics The data set comprises 75,598 RAEs collected on each treatment cycle for 1,181 patients between January 1999 and November 2001. This time period and trials were selected such that we had data collected both pre- and postimplementation of CTC V2.0, the trials were NCCTG coordinated, and the trials were open to accrual during this period. Two of the 28 NCCTG trials open to accrual during this period were excluded because attribution was not collected, leaving 26 trials (24 phase II, two phase III) used in the analysis. Five of the trials used NCI-supplied investigational agent(s). Of the 24 phase II trials, nine were investigating single agents; the remaining 15 were testing combination therapies. Eleven percent (8,318 of 75,598) of RAEs required queries of the membership because of missing grade or attribution. Complete data triplicates of RAE type, grade, and attribution are available on 67,280 patient events, and this set forms the basis for this analysis (Table 1; Fig 1). A lower percentage of RAEs were prespecified to monitor for the hematology and lung trials, as these trials used primarily commercial agents of which the AE profiles are well known.
Characteristics of RAE Data An average of 2,588 RAEs were reported per study (range, 176 to 23,055). One percent of RAEs reported were also reported through expedited systems (eg, Adverse Event Expedited Reporting System). An average of 13 RAEs were transcribed per patient evaluation (range, one to 45). During the life of a trial, the CRAs transcribed an average of 58 RAEs per patient (range, one to 484). Fifty-four percent of patients experienced at least one grade 3+ AE. A higher percentage (31% v 21%) of unrelated and unlikely to be related RAEs were reported for those RAEs grade 3 and above (Table 2; Fig 2) as compared to grades 1 and 2. Similarly, and in the subset of 15 phase II studies involving combination regimens (v single agents), a higher percentage of unrelated and unlikely to be related RAEs were observed (ie, 37% v 21%) for grade 3 and higher RAEs (v grades 1 and 2.). We note that a larger percentage of unrelated (or unlikely) RAEs were observed in brain trials and those trials classified as miscellaneous. This may be due to the difficulty in distinguishing AEs from events due solely to disease progression, of which the latter are not to be reported as AEs per CTC V2.0 definitions. The brain trials in our analysis included patients with newly diagnosed grade 3/4 disease as well as one trial including both recurrent and primary disease. The miscellaneous trials enrolled patients with metastatic disease.
Prespecified RAEs Eighty five percent (57,033 of 67,280) of the RAEs were prespecified. This percentage did not differ on the basis of the agent(s) investigational status (P = .55), but did differ by disease status (P < .0001), specifically with regard to hematology and lung studies. These results did not change on the basis of either phase of study (83% phase II, 87% phase III) or of using single versus combination agents (82% v 85%). Overall, there was an average of 12 AE classifications prespecified per study (range, 3 to 20 classifications), which did not differ on the basis of the agent(s) investigational status or disease setting. Of the RAEs prespecified to be monitored, 54% were never recorded to be more severe than grade 2 during a study. Per patient visit, 85% of the prespecified RAEs that were required to be monitored and reported did not occur. RAEs most frequently included in the prespecified list were diarrhea (13%), nausea (12%), neutropenia (15%), vomiting (13%), and infection with neutropenia (10%).
Extra RAEs
Validation Sample
The ethical care of patients, and good clinical practice, require that treating physicians carefully monitor their patients for AEs caused by treatment. However, the conduct of sound clinical research has different requirements. In the area of AE monitoring and reporting, we believe that our data suggest that we are collecting more data than are needed for high-quality clinical research. Of more than 75,000 RAE data triplicates (ie, type, grade, attribution) collected, 11% required additional information regarding grade or attribution. Clarifying these issues alone is a burden on both the local site and the coordinating center. An average of 12 (range, three to 20) RAEs were prespecified for transcription at each patient visit; 85% (47,754 of 57,033) of these RAEs were grade 0 (ie, did not occur in that patient cycle). Overall, only 3% of RAEs transcribed were grade 3, 4, or 5. We believe that a balance needs to be identified between the responsibility of monitoring patient safety, developing a safety profile of a regimen, and satisfying reporting requirements versus the process used to identify, collect, transcribe, enter into a computer, correct, analyze, and ultimately publish potentially unnecessary RAE data. In view of the volume of RAE data collected, research organizations need to more fully embrace technology to streamline the data collection processes.10 A recent publication11 reported an average of 18 minutes (maximum, 4 hours) of CRA workload to conduct an AE assessment for a patient visit, requiring in addition an average of 35 minutes (maximum, 6 hours) to transcribe the RAEs and document treatment. As described earlier, the NCCTG and other Mayo Clinic Comprehensive Cancer Center clinical trials are now using remote data entry8,12 which will eliminate the transcription and centralized entry of paper CRFs, presumably improving efficiency. Other institutions are using the Internet as a tool to develop forms,13 manage specimens,14 or manage trials.15,16 Alternatively, optical scanning, digital photography, and facsimiles17-20; direct entry from an electronic medical record20-22; or digital assistants (palm pilots) are being used to document and submit patient medical information directly into a computerized medical record.23-26 Although this technology is available, several limitations exist, primarily because of differing computer systems, nonstandardization of data items, security of patient data, and the time needed for data documentation and abstraction.12,15,22,26-28 In general, this technology addresses how rapidly we obtain retrievable data, but does not address the sheer volume of data collected. The remainder of this discussion focuses on factors that need to be considered in data reduction strategies. We believe that careful consideration must be given to limiting the collection of RAE data, much of which will depend on the severity and level of attribution of RAEs observed. Recent literature supports the notion that variable physician experience and expertise in AE reporting significantly influences the ability to collect accurate AE data. Severity is clearly defined via the CTC and must be collected, which leaves us to consider the importance of collecting events considered not attributable to the treatment(s) investigated or collecting only a subset of data. Although not a goal of our study, we expand on the notion of subjectivity in assigning attribution, given that it clearly impacts any decision in terms of which data not to collect. Thereafter, we provide examples of strategies that may be used to reduce excessive data collection. Consistent assignment of attribution may be difficult because of observer variability. Several reports have examined the differences across reviewers of AEs,29,30 finding poor inter-rater reliability. A classic example of such variability is the death of Hanna Greener in 1848, after she received chloroform as anesthesia for the removal of a toenail31; more than 150 years later, the cause of her death is still debated. In lieu of the technology available today, the authors note that it is "interesting and instructional to observe how one's frame of reference can lead to an observer bias and cause even the most learned and respected physician to interpret facts to fit the individual's prejudices."31 A more recent example of variability between observers occurred on a recent phase III Intergroup advanced colorectal trial. Of 23 grade 5 AEs reported,32 10 were initially considered treatment related by the local treating physicians. However, a central, independent review classified 20 of the events as treatment related.32 This review panel recommended a new scale for attribution: treatment related, treatment unrelated, or treatment exacerbated. Accurately recognizing attribution is also difficult for several reasons. The stage of drug development (phase I v II v III) influences the ability to recognize drug-induced reactions. Even after marketing, recognizing attribution may remain difficult due to other confounding factors.29-45 Furthermore, the AEs documented within the medical history represent only a subset of all events experienced, given that the patient is not necessarily systematically questioned regarding all possible health aspects.46-49 Fellowes et al50 found that 89% of women receiving adjuvant tamoxifen or goserelin for breast cancer had recorded AEs in their medical notes, compared with 99% reporting AEs at interview. Regarding data reduction strategies, if consistent and accurate attribution of AEs to study treatment were possible, it could be argued that we do not need to collect AEs deemed unrelated or unlikely to be related to study treatment. However, current NCI reporting requirements mandate the collection of some such events (probably for the reasons mentioned in the previous two paragraphs). Quarterly reports required by NCI5 include unrelated and unlikely grade 3, 4, and 5 AEs. By definition, cooperative groups would be noncompliant if only related events were collected. In the time period covered in this report, our CRFs did not contain the level of detail necessary for CRAs to be aware of which combinations of grade and attribution to report. Therefore, our CRAs completed the CRFs with inconsistent reporting, which is evident in our data. For example, our data suggest that a higher percentage of unrelated and unlikely to be related AEs are reported in single-agent phase II trials than in phase II trials evaluating combination agents (38% v 27%). This observation in part may be because single agents have a less clearly defined AE profile. Overall, we are more likely to receive all RAEs for grades 3 to 5 than we are to receive RAEs for grades 1 and 2, as suggested by a higher percentage of unrelated and unlikely RAEs reported for grades 3+ (Fig 2). Clearly, more direction is needed with regard to which combinations of grade and attribution are necessary to report via CRFs. There is a great difference between the volume of data collected and the condensed summaries reported in literature.51-53 Thus, a second strategy to reduce RAE data collection would be to collect only what is reported in published literature. For example, typically only the maximum severity (v all events) per patient is reported, which ignores the frequency and occurrences of repeated AEs. In our study, only 28% of RAEs would have been reported if only the maximum severity per patient during their entire course of treatment were required, thus eliminating the reporting of approximately 48,244 AEs. However, the sites would still need to monitor and record all events to identify the patients' maximum severity of each AE, which would not eliminate this part of the process of collecting AEs. This strategy would limit the ability to detect quality control issues with respect to study compliance (ie, dose modifications), and would place an added burden on auditing team(s) for verification of data against the medical history. In addition, important characteristics of the AEs (eg, cumulative v immediate effects), and the time to resolution would be difficult to track. This strategy may be acceptable for larger phase III trials; however, we believe that it is inadequate in earlier stages of development (ie, phase I and II studies). A final hazard of a strategy of only collecting summary AE data at the end of a trial is that, under such a strategy, the coordinating data center may fail to detect rapidly an increased incidence of known (ie, expected) AEs for studies using reduced SAE reporting requirements (such as for commercial agents.) Here, the NCCTG Real-Time Toxicity Monitoring Program (for which all grade 4 and 5 events and hospitalizations not required to be reported via other expedited systems are reported in real-time fashion to our research base on a simple one-page form) may alleviate this problem.54,55 An alternative would be to collect more intensive RAE data on a subset of patients on a phase III trial. We currently are using this approach on the initial 300 patients per arm of the current NCCTG phase III colon adjuvant trial (N0147). Patients enrolled beyond the 300th patient (per arm) will have only maximum severity per type reported at the end of active treatment. The data for this trial are too immature to report at this time. A third strategy would be to collect only SAE data. In our trial, only 1% of RAEs required SAE reporting. This may not be acceptable, especially in studies using commercial agents, for which less extensive SAE reporting is required; such studies are also more subject to subtleties associated with attribution. Figueiras et al33 performed a study showing that physician attitudes are associated with the probability of SAE reporting, based on beliefs that SAEs were well documented before marketing; it is nearly impossible to determine attribution; reporting is required only if the AE is definitely related to the drug(s); and a single report cannot contribute to medical knowledge. Our own Real-Time Toxicity Monitoring Program55 has detected increased rates of expected AEs in more than six trials that would not have been reported via expedited systems. Almost a dozen trials had new toxicity patterns identified on the basis of collecting AE data cycle by cycle and during the time period reviewed in this study. Collecting only SAE data on such trials would have failed to detect such anomalies. Therefore, this strategy may not be optimal. In an effort to strike a balance between ensuring patient safety and relieving the data burden, the NCCTG has developed a possible solution: the NCCTG Routine AE Data Submission Policy. Under this policy, all grades and attributions are transcribed for the prespecified AEs. However, for AEs observed beyond those prespecified, the CRAs transcribe based on the combination of grade and attribution. CRAs transcribe and submit grade 2 AEs considered at least possibly related to treatment, and all grade 3 or greater events regardless of relationship to treatment. An optional version is available, which includes grade 1 AEs with attribution of possible, probable, or definite, and satisfies reporting requirements for NCI agent-supplied trials.5 This policy clarifies the RAE data reporting requirements. More importantly, the policy eliminates 69% (7,067 of 10,247) of unnecessary transcription by removing all triplicates (ie, type, grade, attribution) of grade 1 and unrelated and unlikely grade 2 AEs. We also strive to limit the number of prespecified RAEs to only those absolutely necessary. This policy was recently implemented in the NCCTG for studies using the NCI Common Terminology Criteria of Adverse Events (V3.0). We believe that the new NCCTG policy strikes a balance between maintaining patient safety and recording clinically important data. The policy clearly states the RAE data items that are required. Our plans are to evaluate and report the success of this policy formally in future report(s). This policy is NCCTG's effort and the first step in improving the process of collecting RAE data.
The authors indicated no potential conflicts of interest.
Supported in part by Public Health Service Grants No. CA-25224, CA-37404, and CA-35103 and conducted as a collaborative trial of the North Central Cancer Treatment Group and Mayo Clinic. Authors' disclosures of potential conflicts of interest are found at the end of this article.
1. http://www.fda.gov/cder/cancer/toxicityframe.html: 2. Brown EG: Effects of coding dictionary on signal generation: A consideration of use of MeDRA compared with WHO-ART. Drug Saf 25:445-452, 2002 (abstr)[Medline] 3. Weinstein JN, Deyo RA: Clinical research: Issues in data collection. Spine 25:3104-3109, 2000[Medline] 4. Dolan G, Broomfield J, Lewith G, et al: Operational and resource management of an RCT: Some of the pitfalls and lessons learned. J Clin Nurs 8:389-395, 1999[Medline] 5. http://ctep.info.nih.gov/reporting/adeers.html: 6. http://www.fda.gov/cber/gdlns/vaers-l.txt: 7. http://www.fda.gov/medwatch/index.html: 8. de Groen PC, Barry JA, Schaller WJ: Applying world wide web technology to the study of patients with rare diseases. Ann Intern Med 129:107-113, 1998 9. http://ctep.info.nih.gov/reporting/ctc.html: 10. Hardison CD, Schnetzer T: Using information technology to improve the quality and efficiency of clinical trial research in academic medical centers. Pharmaceut Pract Manag Q 19:49-57, 1999 (abstr) 11. Roche K, Paul N, Smuck B, et al: Factors affecting workload of cancer clinical trials: Results of a multicenter study of the National Cancer Institute of Canada Clinical Trials Group. J Clin Oncol 20:545-556, 2002 12. Geyer S, Sargent D, Daiss M, et al: Mayo Clinic Phase 2 Consortium: Challenges and triumphs in establishing a multi-institution cancer clinical trials consortium. Control Clin Trials 23:59S, 2002 (abstr A44)[CrossRef] 13. Brenner D, Nelson KM: Web Form Builder: Design your protocol specific data capture. Control Clin Trials 23:53S, 2002 (abstr A34) 14. Brandt DK, Tikhomirov A, Stein A, et al: A web-based specimen tracking system. Control Clin Trials 23:55S, 2002 (abstr A37)[CrossRef] 15. Kleinman B, Darke A, Hartline JA: The challenges and benefits of a web-based trial. Control Clin Trials 23:91S, 2002 (abstr P62) 16. Heger M: Exploring new technologies in clinical research: Experiences from multinational online studies using the internet. Int J Pharmaceut Med 16:34-35, 2002 17. Khan ZM, Rascati KL, Koeller JM, et al: Fax technology for collecting outcomes data in a computer database. Am J Health Syst Pharm 56:2540-2542, 1999 18. Galdino GM, Swier P, Manson PN, et al: Converting to digital photography: A model for a large group or academic practice. Plast Reconstr Surg 106:119-124, 2000[Medline] 19. Davidson LJ, Ryan WJ, Rohay JM, et al: Technological advances in data entry and verification: Is Teleform for you? Nurs Res 45:373-376, 1996[Medline] 20. Denwood R: Data capture for quality management nursing opportunity. Comput Nurs 14:39-44, 1996[Medline] 21. Darbo DA, Eichenauer J, Darbo MJ: Computerized medical record war story: Practical experiences of computer implementation at the front. J Ambul Care Manag 23:38-42, 2000 (abstr)[Medline] 22. Gutierrez G: Medicare, the Internet, and the future of telemedicine. Crit Care Med 29:N144-N150, 2001[Medline] 23. Thomas JD, Nissen SE: Digital storage and transmission of cardiovascular images: What are the costs, benefits and timetable for conversion? Heart 76:13-17, 1996 24. Information Pa: PalmVue System. Gastroenterol Nurs 19:40, 1996 25. Lee JJ, Zee B, Lai M, et al: A mobile logging/screening system using handheld computing device. Control Clin Trials 23:54S, 2002 (abstr A35) 26. Wells M, Chambliss L, Doerzbacher J: Handing off a handheld data collection device to the researcher's hand. Control Clin Trials 23:54S, 2002 (abstr A36) 27. Donaldson LJ: From black bag to black box: Will computers improve the NHS? Evidence must shape implementation. BMJ 312:1371-1372, 1996 28. Wetzel G: Future challenges in pediatric cardiology: The increasing role of medical informatics. Curr Opin Cardiology 15:209-210, 2000[Medline] 29. Hofer TP, Bernstein SJ, DeMonner S, et al: Discussion between reviewers does not improve reliability of peer review of hospital quality. Med Care 38:152-161, 2000[CrossRef][Medline] 30. Thomas EJ, Studdert LLB, Brennan TA: The reliability of medical record review for estimating adverse event rates. Ann Intern Med 136:812-816, 2002 31. Knight PR, Bacon DR: An unexplained death: Hannah Greener and chloroform. Anesthesiology 96:1250-1253, 2002[Medline] 32. 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 33. Figueiras A, Tato F, Fontainas J, et al: Influence of physicians' attitudes on reporting adverse drug events: A case-control study. Med Care 37:809-814, 1999[CrossRef][Medline] 34. Moore TJ, Psaty BM, Furberg CD: Time to act on drug safety. JAMA 279:1571-1573, 1998 35. Chaikof EL: Hear no evil, see no evil, speak no evil: Postmarket monitoring, underreporting, and estimating the prevalence of endograft-related adverse events. J Vasc Surg 35:1299-1300, 2002[CrossRef][Medline] 36. US Food and Drug Administration, Summary F: Managing the risk from medical product use. Creating a risk management framework, 1999. http://www.fda.gov/oc/tfrm/executivesummary.pdf. 37. Niu MT, Salive ME, Ellenberg SS: Neonatal deaths after hepatitis B vaccine: The vaccine adverse event reporting system, 1991-1998. Arch Pediatr Adolesc Med 153:1279-1282, 1999 38. Rosenthal S, Chen R: The reporting sensitivities of two passive surveillance systems for vaccine adverse events. Am J Public Health 85:1706-1709, 1995 39. Moride Y, Haramburu F, Requejo AA, et al: Under-reporting of adverse drug reactions in general practice. Br J Clin Pharmacol 43:177-181, 1997[CrossRef][Medline] 40. Nutescu E, Hunt C, Teeters J: Multidisciplinary approach to improving allergy documentation. Am J Health Syst Pharm 55:364-368, 1998 41. Davis C, Ralevski E, Kennedy SH, et al: The role of personality factors in the reporting of side effect complaints to moclobemide and placebo: A study of healthy male and female volunteers. J Clin Psychopharmacol 15:347-352, 1995[CrossRef][Medline] 42. Currier JS, Spino C, Grimes J, et al: Differences between women and men in adverse events and CD4 responses to nucleoside analogue therapy for HIV infection. J Acquir Immune Defic Syndr 24:316-324, 2000[Medline] 43. Walker JS, Carmody JJ: Experimental pain in healthy human subjects: Gender differences in nociception and in response to ibuprofen. Anesthes Analges 86:1257-1262, 1998[Abstract] 44. Sloan JA, Goldberg RM, Sargent DJ, et al: Women experience greater toxicity with Fluorouracil-based chemotherapy for colorectal cancer. J Clin Oncol 20:1491-1498, 2002 45. Barsky AJ, Saintfort R, Rogers MP, et al: Nonspecific medication side effects and the nocebo phenomenon. JAMA 287:622-627, 2002 46. Colebunders R, De Roo A, Schrooten W: Safety reporting in clinical trials. JAMA 285:2076-2078, 2001 47. Hanlon JT, Maher RL, Lindblad CI, et al: Comparison of methods for detecting potential adverse drug events in frail elderly inpatients and outpatients. Am J Health Syst Pharm 58:1622-1626, 2001 48. Di Giulio P, Arrigo C, Gall H, et al: Expanding the role of the nurse in clinical trials: The nursing summaries. Cancer Nurs 19:343-347, 1996[CrossRef][Medline] 49. Hyland ME, Crocker GR: Validation of an asthma quality of life diary in a clinical trial. Thorax 50:724-730, 1995 50. Fellowes D, Fallowfield LJ, Saunders CM, et al: Tolerability of hormone therapies for breast cancer: How informative are documented symptom profiles in medical notes for ‘well-tolerated’ treatments? Breast Cancer Res Treat 66:73-81, 2001 (abstr)[CrossRef][Medline] 51. Rothenberg ML, Meropol NJ, Poplin EA, et al: Mortality associated with irinotecan plus bolus fluorouracil/leucovorin. J Clin Oncol 20:1145-1146, 2002 52. Perrone F, De Maio E, Maione P, et al: Survey of modalities of toxicity assessment and reporting in noncomparative prospective studies of chemotherapy in breast cancer. J Clin Oncol 20:52-57, 2001[CrossRef] 53. Ioannidis JPA, Lau J: Completeness of safety reporting in randomized trials: An evaluation of 7 medical areas. JAMA 285:437-443, 2001 54. Sargent DJ, Goldberg RM, Mahoney MR, et al: Rapid reporting and review of an increased incidence of a known adverse event. J Natl Cancer Inst 92:1011-1013, 2002 55. 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 Submitted August 25, 2004; accepted September 22, 2005.
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
|
TOP
ABSTRACT
INTRODUCTION
2 and Fisher's exact tests were used. Analyses were conducted using SAS (SAS Institute, Cary, NC) and S-plus (V3.3; Statistical Sciences, Seattle, WA). 
