• © 2003 by American Society of Clinical Oncology

In Reply:

We would like to thank Dr Goel for carefully reading our article¹ and commenting on some missing statistical information. Indeed there is a typing error in Figure 1 of our article: The number of patients who received salvage treatment categorized as “other” did not add up to 16 patients, as erroneously cited in the figure, but to only 11 patients. We would like to apologize for this error.

We do not agree, however, to exclude the 17 patients who did not undergo restaging transurethral resection (TUR) from the denominator when calculating the percentage of salvage cystectomy applied for tumor progression or relapse. Although we do not have information with respect to the immediate tumor response for these 17 patients, we do have follow-up data. None of these 17 patients were treated with salvage cystectomy during the follow-up period. Thus the percentage of patients among the entire group that underwent salvage cystectomy remains 83 (20%) of 415 patients. This does not imply, however, that these 17 patients did not experience a local relapse or local tumor progression, but simply indicates that cystectomy was not a treatment option in these patients, mainly because of advanced age or patient refusal.

The most important point raised by Dr Goel refers to the problem of multicollinearity. Of course T stage, N stage, and R status of residual tumor after TUR (R0, R1, R2) are not completely independent variables, statistically or clinically. For T4 tumors we observed some more nodal positive cases (as judged clinically by computed tomography scan, not histopathologically in the resected specimen) than for T1 to T3 tumors, but this difference was not significant. On the other hand, there was an association between the degree of residual tumor and T stage as well as between R stage and N stage, but the calculated values of r² (Spearmen correlation) did not indicate a risk of multicollinearity (T stage v R status, r² = 0.14; N stage v R status, r² = 0.02).

Candidates for joining the multivariate model as independent variables were selected using the results of univariate analysis. All selected variables were included in the regression model (no stepwise method). The differences in the significance of the three factors (T stage, N stage, R status) in the multivariate models might reflect their unequal clinical impact on the respective end points. For the end point of complete response at restaging TUR, it is obvious that the degree of residual tumor before radiotherapy or radiochemotherapy had the strongest impact, as it is very unlikely that a patient with a complete initial TUR (R0) will have residual tumor at restaging TUR 6 weeks after completion of radiotherapy or radiochemotherapy. The risk of developing distant metastases is biologically highest among patients with advanced disease (T3/4 or N+), whereas the variable R status also reflects some treatment-related constraints (eg, quality of surgery, topography of tumor site, irresectability of superficial but large tumors) that are not directly related to the risk of developing distant disease.