|
Advertisement |
|
|
||
 |
|
|||||
|
|||||
|
 |
||||||
|
||||||
|
||||||
|
Originally published as JCO Early Release 10.1200/JCO.2008.18.2154 on September 22 2008 © 2008 American Society of Clinical Oncology.
Missing Patient-Reported Outcome Data in an Adjuvant Lung Cancer StudyDepartment of Biostatistics, Graduate School of Public Health, University of Pittsburgh, Pittsburgh, PA In this issue of Journal of Clinical Oncology, Bezjak et al1 describe important results regarding patient-reported outcomes (PROs) in adjuvant therapy for non–small-cell lung cancer. Because the clinical outcomes for their study proved favorable for the chemotherapy, patients and their care providers will welcome the patient-reported information. The PROs will help future patients anticipate how the therapy will affect their lives. It will also help them distinguish symptoms associated with treatment from symptoms of cancer recurrence, an all-important distinction for an adjuvant patient. As important as this article is for the treatment of lung cancer, it is also of methodologic interest because it illustrates several issues regarding missing data in PROs. There were four sources of missing data by design: patients assigned to the observation group were not assessed at weeks 5 and 9; patients were not expected to complete assessments if they did not receive their assigned therapy; patients were not expected to continue assessments after cancer recurrence; and patients were not expected to continue in the PRO substudy if they did not complete the baseline assessment. There were also missing data as a result of low adherence. Regarding the data that were missing as a result of low adherence, the authors should be applauded for providing information about compliance at each time point and for examining compliance differences according to patient characteristics and treatment group. However, a more complete analysis would have used logistic mixed-effects models to include all time points rather than just baseline in the analysis of patient characteristics (eg, as in Land et al2). Patients with different characteristics might have been more likely to drop out over time. In addition, it is unfortunate that the significant patient differences they found in baseline compliance were not further explored to determine how they might have affected the treatment group comparisons and overall conclusions. The authors also compared compliance between treatment groups at each time point and declared them comparable. However, in the long-term assessments, the observation group had consistently higher compliance. The differences were nontrivial in magnitude (up to 10%) and reached the level of a statistical trend (P < .1) at both 9 and 12 months. A longitudinal analysis of compliance might well have found a significant difference during the long-term period. The provision of PRO completion rates by time point, including available sample sizes and the percentages submitted of those expected, should become standard for PRO articles. This can be in the form of a table such as that provided in Bezjak et al1 or, if the missing data problem is mild, a text summary of the range of completion rates. I submit that a longitudinal analysis of compliance differences (eg, using a logistic mixed-effects model) should also be standard. Furthermore, when there are significant differences or when the adherence rates are low, the potential biases in the treatment comparisons should be formally explored. If information is collected prospectively regarding the causes of missing data, that might be used to infer that data were missing at random. If missing data are potentially informative, there are a host of methods to pursue, such as pattern mixture models or selection models.3,4 How might the missing data have affected the results in the Bezjak et al1 article? One consideration is that the primary end point is based on an increase of 10 points over baseline at any point (or a decrease in 10 points, if no increase is observed). The more assessments observed, the greater the chance of observing such an increase. The omitted assessments at weeks 5 and 9 in the observation group might also have biased toward worse quality of life (QOL) in the chemotherapy arm, depending on the natural history of QOL in the weeks after surgery. It is unfortunate that the observation group could not be assessed at the same time points as the chemotherapy group, although I certainly recognize the difficulty of doing so. Missing data also biased the conclusion that patients’ QOL returned to baseline in the long-term, as demonstrated in Figure 4 in Bezjak et al.1 That figure includes a large number of patients at baseline who were not assessed at 36 months (186 – 50 = 136 in the chemotherapy group; 173 – 39 = 134 in the observation group), many of whom had experienced disease recurrence or had died or who had not received assigned treatment. The patients in the 36-month average are largely those who were healthier and had higher QOL at baseline. (We know from this article that those patients with higher baseline QOL survived longer). This particular study also demonstrates another common cause of missing data in multicentered trials: poor adherence at the institution level. The lead cooperative group was the National Cancer Institute of Canada Clinical Trials Working Group (NCIC-CTG), but the study was conducted throughout the North American Intergroup. Completion of QOL was only mandatory for NCIC-CTG sites. A secondary subset analysis including only the participants in high-adherence institutions (NCIC-CTG or otherwise) would be free of some of the biases resulting from missing data when all participants are included. A final point: I have used the Bezjak et al1 article as a case study of missing data not because it was unique, but because these are problems that arise often. In highlighting the problems, I do not wish to overshadow the quality and importance of the study and its results, without which it would not have been selected for publication. AUTHOR'S DISCLOSURES OF POTENTIAL CONFLICTS OF INTEREST The author(s) indicated no potential conflicts of interest. REFERENCES
1. Bezjak A, Lee CW, Ding K, et al: Quality-of-life outcomes for adjuvant chemotherapy in early-stage non–small-cell lung cancer: Results from a randomized trial, JBR.10. J Clin Oncol doi:10.1200/JCO.2007.12.6094 [epub ahead of print on September 22, 2008] 2. Land SR, Ritter MW, Costantino JP, et al: Compliance with patient-reported outcomes in multicenter clinical trials: Methodologic and practical approaches. J Clin Oncol 25:5113-5120, 2007 3. Fairclough D: Design and Analysis of Quality of Life Studies in Clinical Trials (Kindle ed). London, United Kingdom, Chapman & Hall, 2002 4. Donaldson GW, Moinpour CM: Learning to live with missing quality-of-life data in advanced-stage disease trials. J Clin Oncol 23:7380-7384, 2005
Related Article
|
|||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
|
|
|||||||||||
|
|||||||||||
|
|
Copyright © 2008 by the American Society of Clinical Oncology, Online ISSN: 1527-7755. Print ISSN: 0732-183X
|
