Jan 18, 2008 - index is best for research purposes is unclear. Although the ... Registrar General's office, and has been shown to be ... The functional domain.
1043
Is There an Optimal Comorbidity Index for Prostate Cancer? Shabbir M. H. Alibhai, MD, MSc1,2,3,4 Marc Leach, MA1 George A. Tomlinson, PhD1,3,4,5 Murray D. Krahn, MD, MSc1,3,4 Neil E. Fleshner, MD, MPH6 Gary Naglie, MD1,2,3,4
BACKGROUND. Comorbidity is an important consideration in oncology practice, particularly among older patients. Although a variety of comorbidity indices have been employed in research studies, it is unclear whether any one index is preferred.
METHODS. An age-stratified random sample of 345 men (mean age of 69 years) who were newly diagnosed with prostate cancer were identified from a cancer registry in Ontario, Canada. Comorbidity and treatment information were obtained from chart review. Four comorbidity indices were utilized: Charlson Index, Diag-
Division of General Internal Medicine and Clinical Epidemiology, University Health Network, Toronto, Ontario, Canada.
nosis Count, Index of Coexistent Disease (ICED), and number of medications.
2
Geriatric Program, Toronto Rehabilitation Institute, Toronto, Ontario, Canada.
tectomy or radiotherapy) and overall 6-year survival. Multivariable model perform-
3
the area under the receiver operating characteristic curve (AUROC).
4
RESULTS. Among men with localized disease (n 5 231), in models adjusted for age, Gleason score, and prostate-specific antigen level, only the Charlson Index
1
Department of Medicine, University of Toronto, Toronto, Ontario, Canada. Department of Health Policy, Management and Evaluation, University of Toronto, Toronto, Ontario, Canada. 5
Department of Public Health Sciences, University of Toronto, Toronto, Ontario, Canada. 6
Department of Surgery, University of Toronto, Toronto, Ontario, Canada.
Logistic regression analysis was used to compare the performance of comorbidity measures with respect to predicting receipt of curative treatment (radical prostaance including each of the comorbidity measures was compared by calculating
was found to be a statistically significant predictor of receipt of curative treatment (P < .05), although all comorbidity indices had similar AUROC in adjusted models. After a median follow-up of 6.5 years, 116 of 345 men (33.6%) had died. In adjusted models, all 4 comorbidity indices performed similarly in predicting overall survival.
CONCLUSIONS. Although comorbidity is an important predictor of both curative treatment and overall survival in prostate cancer, the optimal comorbidity index for use in research remains unclear. Selecting the optimal comorbidity index may depend on both the specific patient population and the outcome being considered. Cancer 2008;112:1043–50. 2008 American Cancer Society.
Supported in part by a grant from the Physicians Services Incorporated Foundation (Grant 99-14). Support was also provided by research fellowships from the Queen Elizabeth Hospital Research Foundation and the Department of Medicine at the University of Toronto (to S.M.H.A.), by the Toronto Rehab Institute (to S.M.H.A.), by unrestricted funds from the Ontario Ministry of Health (to S.M.H.A.), by the F. Norman Hughes Chair in Pharmacoeconomics (to M.D.K), and by the Mary Trimmer Chair in Geriatric Medicine Research at the University of Toronto (to G.N.). Dr. Alibhai is a Research Scientist with the National Cancer Institute of Canada. We thank Cancer Care Ontario for preparing the study file and the clinicians, hospitals, and regional cancer centers that participated in this study.
ª 2008 American Cancer Society
KEYWORDS: prostatic neoplasms, aged, comorbidity, survival, treatment, risk adjustment.
omorbidity has been defined as the coexistence of 2 chronic conditions or impairments in the same person.1 Comorbidity is an important predictor of survival for a variety of common cancers.2 In the setting of prostate cancer, increasing comorbidity, independent of age, is associated with a lower likelihood of receiving potentially curative therapy,3,4 a shorter overall survival,5 a greater likelihood of dying from other causes,6,7 and a greater risk of various short-term postoperative complications.8
C
Address for reprints: Shabbir M. H. Alibhai, MD, MSc, Division of General Internal Medicine, University Health Network, Rm. EN 14-214, 200 Elizabeth Street, Toronto, Ontario, M5G 2C4 Canada; Fax: (416) 595-5826; E-mail: shabbir.alibhai@ uhn.on.ca
DOI 10.1002/cncr.23269 Published online 18 January 2008 in Wiley InterScience (www.interscience.wiley.com).
Received July 18, 2007; revision received August 28, 2007; accepted September 24, 2007.
1044
CANCER
March 1, 2008 / Volume 112 / Number 5
Despite the widespread use of comorbidity measures in clinical research, which comorbidity index is best for research purposes is unclear. Although the importance of including comorbidity when performing risk adjustment in clinical studies is now standard practice, to our knowledge, there is a dearth of head-to-head studies examining the performance of different comorbidity measures, both in the setting of prostate cancer9 and in other malignancies. The objective of the current study was to examine whether any 1 of 4 commonly used comorbidity measures was superior in terms of its predictive ability of specific outcomes. We selected 2 clinically important outcomes for this analysis—receipt of curative therapy (radical prostatectomy or radiation therapy) and overall survival—and focused on men with newly diagnosed prostate cancer.
MATERIALS AND METHODS Study Design We identified a cohort of men with newly diagnosed prostate cancer from a large cancer registry. We performed detailed chart reviews to obtain clinical and comorbidity information on a stratified random sample of patients from this larger cohort. Ethics approval was obtained from the University of Toronto and participating hospitals. Database Cohort All 5192 patients in the Ontario Cancer Registry (OCR) with a new diagnosis of prostate cancer between April 1, 1995 and March 31, 1996 were selected. The OCR is a comprehensive province-wide cancer database with case ascertainment rates in excess of 95%.10,11 The OCR contains limited treatment and comorbidity information but does not include tumor-specific (stage and grade) information. Vital status information in the OCR comes from the Registrar General’s office, and has been shown to be >99% accurate.12,13 Chart Review Cohort A stratified random sample of patients with prostate cancer was generated from the database cohort. Sampling was stratified by age into 4 categories (age
