© 2004 by Oxford University Press
© 2004 Oxford University Press
CORRESPONDENCE |
RESPONSE: Re: Is Patient Travel Distance Associated With Survival on Phase II Clinical Trials in Oncology?
Correspondence to: Elizabeth B. Lamont, MD, MS, Massachusetts General Hospital Cancer Center, 55 Fruit St., Suite: Zero Emerson Place, Boston, MA 02114 (e-mail: elamont{at}partners.org)
We appreciate Drs. Muñoz and Samet's careful review of our manuscript and the opportunity to provide additional details and results from our work. They have one main methodologic concern and several related empiric statistical questions regarding our study. Their overarching methodologic concern is that multivariable regression may not be able to adjust appropriately for the extreme differences in prognostically important covariates (e.g., race, income, tumor site, extent of disease) between the two groups of patients (i.e., those patients living 
15 miles from the center and those living >15 miles from the center). With respect to empirical questions, Drs. Muñoz and Samet ask for a number of statistical details to evaluate the possibility that the association we found between travel distance and survival is spurious, the result of confounding by race.
To address the methodologic concern, Drs. Muñoz and Samet recommend stratifying the analysis by each covariate. Indeed, in our original analyses (not presented in our article), we carried out stratified analyses and evaluated interaction effects for several covariates. Here we provide information for one important covariate, patient race. However, we note that the use of multivariable methods in this setting and decisions regarding which models to evaluate and present are often a matter of taste and debate.
Given that white race and travel distance (15-mile cut point) were highly associated, the concern of Drs. Muñoz and Samet that the association between travel distance and survival may be confounded by race is appropriate. Among those living within 15 miles of the treating facility, 28% were white, and among those living more than 15 miles from the treating institution 91% were white (P<.001). Of the 110 patients analyzed, 44 patients (40%) died during the observation period. Grouping African American and white patients together, the univariate hazard of death for those living more than 15 miles from the treating facility relative to those living closer was 0.81 (95% confidence interval [CI] = 0.44 to 1.48) (Fig. 1). The univariate hazard of death among African American patients was equivalent to that of the white patients (hazard ratio = 1.06, 95% CI = 0.56 to 2.01). Among African American patients, the univariate hazard of death for those living more than 15 miles from the treating institution relative to those living closer was 0.53 (95% CI = 0.07 to 4.04). Among white patients, the univariate hazard of death for those living more than 15 miles from the treating institution relative to those living closer was 0.78 (95% CI = 0.29 to 2.07). We did not find a statistically significant interaction between race and travel distance (Table 1); however, given the small sample size, statistical power for such tests is clearly limited. In summary, despite the strong association between race and travel distance, results of our stratified analysis and interaction test suggest at least a qualitatively consistent effect of distance on outcomes in African American and white patients and that the association between travel distance and survival that we reported originally is not confounded by race.
|
2(1) = 0.76, P = .383.
|
Among other empiric questions, Drs. Muñoz and Samet found an inconsistency between the number of white patients in Tables 2 and 3 of the original article. In evaluating the analytic dataset to understand their finding, we discovered that race was coded incorrectly for three patients. The correct race distribution is as follows: 73 white patients, 35 African American patients, and two patients of the designation "other." We corrected the variable, repeated the analyses, found that our results were robust to the change, and notified the editors. All analyses reported here use the corrected race variable.
REFERENCE
1 Lamont EB, Hayreh D, Pickett KE, Dignam JJ, List MA, Stenson KM, et al. Is patient travel distance associated with survival on phase II clinical trials in oncology? J Natl Cancer Inst 2003;95:1370–5.
CiteULike 
Connotea 
Del.icio.us What's this?
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||