Response Assessment Classification in Patients with Advanced Renal Cell Carcinoma Treated on Clinical Trials

1611 Response Assessment Classification in Patients with Advanced Renal Cell Carcinoma Treated on Clinical Trials Effect of Measurement Criteria and Other Parameters Lawrence H. Schwartz, M.D. 1,2 Madhu Mazumdar, Ph.D. 3 Liang Wang, M.D. 1 Alex Smith, M.S. 3 Stephanie Marion, B.S. 4 David M. Panicek, M.D. 1,2 Robert J. Motzer, M.D. 2,4 1 Department of Radiology, Memorial Sloan-Kettering Cancer Center, New York, New York. 2 Weill Medical College, Cornell University, New York, New York. 3 Department of Epidemiology and Biostatistics, Memorial Sloan-Kettering Cancer Center, New York, New York. 4 Department of Medicine, Memorial Sloan-Kettering Cancer Center, New York, New York. Supported in part by the Byrne Foundation and the National Institutes of Health Cancer Chemotherapy Program Project (CA05826-35); by Mr. William H. Goodwin, Mrs. Alice Goodwin, and the Commonwealth Cancer Foundation for Research (Richmond, Virginia); and by the Experimental Therapeutics Center of Memorial Sloan-Kettering Cancer Center (New York, New York). The authors thank Carol Pearce for her review of the article. Address for reprints: Lawrence H. Schwartz, M.D., Department of Radiology, Memorial Sloan-Kettering Cancer Center, 1275 York Avenue, New York, NY 10021; Fax: (212) 794-4010; E-mail: schwartl@mskcc.org Received March 14, 2003; revision received July 14, 2003; accepted July 21, 2003. BACKGROUND. The objective of the current study was to evaluate the effect on response assessment classification in patients with metastatic renal cell carcinoma (RCC) using unidimensional (Response Evaluation Criteria in Solid Tumors) and bidimensional (World Health Organization) criteria, including or excluding measurements of the primary renal tumor and using a new index to compensate for the disproportionate effect of large renal tumors relative to their metastases. METHODS. Fifty-three imaging studies involving a total of 44 patients with metastatic RCC who were treated on clinical trials of interferon- analogue and/or thalidomide were reviewed retrospectively. The best overall response assessment and progression free survival were calculated with both unidimensional and bidimensional tumor measurements. Patients were then stratified into two groups: patients with primary renal tumors in situ and patients who underwent resection of their primary renal tumors. The best overall response and the time to disease progression were calculated based on the sum of measurements (conventional methodology), both including and excluding the primary tumor. A new method of response assessment, the normalized lesion index, which equalizes the differences in tumor size for an individual patient, was evaluated and compared with the conventional response assessment. RESULTS. There was an 11% disagreement rate in the best overall response assessment between unidimensional and bidimensional measurements. The time to progression was 9.2 months measured unidimensionally, compared with 6.4 months assessed bidimensionally. In the group of patients who had primary renal tumors in situ, using the conventional sum of measurements method, the apparent time to progression was an average of 4.2 months longer compared with measurements that did not include the primary renal tumor. The use of the normalized lesion index method resulted in an improved concordance in best overall response assessments and similar time to progression assessments when the primary renal tumor was included compared with patients who did not have primary renal tumors in situ. CONCLUSIONS. The use of unidimensional measurements in RCC therapy assessment results in significantly different time to progression classification compared with the use of bidimensional measurements. Response assessment classification in patients with RCC is affected by the exclusion or inclusion of measurements of the primary renal tumor. The normalized lesion index warrants further study in assessing response in patients with metastatic RCC and other solid tumor malignancies that often show substantial differences in sizes of measurable lesions. Cancer 2003;98:1611 9. 2003 American Cancer Society. 2003 American Cancer Society DOI 10.1002/cncr.11712

1612 CANCER October 15, 2003 / Volume 98 / Number 8 Advanced renal cell carcinoma (RCC) is characterized by resistance to systemic therapy. 1 4 Interferon- and interleukin-2 are associated with a poor response. 1 4 Although two randomized trials have shown that interferon- treatment modestly improves survival, there were few long-term survivors. 5,6 Several recent reports have described stabilization in the size of metastases and rare partial responses after treatment with thalidomide. 7,8 The study of novel treatments in clinical trials remains a high priority. Because the potential activity of the new therapies is compared with historic data sets and published clinical trials of cytokines or other therapies, 9 response assessment is critical in identifying agents with potential activity even activity of only a modest degree. In patients with RCC, tumor response usually is assessed by computed tomography (CT) imaging. 10,11 The standard criteria for assessing therapy response with CT imaging are evolving. In the past, the criteria outlined by the World Health Organization (WHO) were used to assess the best response and disease progression. The WHO recommends measurement of the greatest tumor dimension and greatest perpendicular dimension; these measurements are multiplied, and the resulting cross product of all tumor deposits measured is summed. 12 A more recently released guideline by the Response Evaluation Criteria in Solid Tumors (RECIST) Group has recommended measuring only the greatest dimension of a tumor (i.e., a unidimensional measurement) and using the sum of these measurements to evaluate response assessment. 13,14 Several aspects of response assessment are particularly relevant to the assessment of new therapies in clinical trials for patients with advanced RCC. 15 Metastases associated with RCC often are not round. Consequently, there may be differences in response assessment classifications obtained with unidimensional and bidimensional measurements. Second, for patients with primary renal tumors in situ, there is often a significant discrepancy between the size of the renal mass and the size of its metastases. 15 Because many primary RCC lesions are large relative to their metastatic disease, inclusion of the primary tumor in the sum of measurements may greatly (and disproportionately) impact therapy response assessment. In addition, the primary RCC tumor frequently does not change substantially in size compared with metastatic disease at follow-up. Time to disease progression as an endpoint of treatment with novel agents on clinical trials has been cited as particularly relevant to patients KEYWORDS: neoplasms, metastases, World Health Organization, carcinoma, renal cell, effectiveness evaluation of treatment, clinical trials. with RCC, given the low response and partial response of RCC to all therapies, the natural history of RCC, and the nature of new agents that may induce tumor stabilization as their primary effect. 15,16 The objectives of the current study were to evaluate two response assessment parameters, the time to disease progression and the best overall response, using the conventional WHO and RECIST criteria, and to study the impact of inclusion or exclusion of measurements of the primary renal tumor on response assessment classification. Finally, we propose a novel system of normalizing response assessment for variably sized tumors, so that both large and small lesions (and changes in their sizes) are given equal weight in the response assessment. We compare this novel system with the conventional sum of measurements technique, which is a weighted response assessment. MATERIALS AND METHODS Patient Population Forty-four patients who were treated on clinical trials with pegylated interferon- or thalidomide 17 19 were the subject of this retrospective review. Thirty-five patients were treated on 1 trial, and 9 patients were treated on 2 trials (1 trial with pegylated interferon- and 1 trial with thalidomide). The response assessment for the 9 patients was considered independently for each of the 2 trials, for a total of 53 assessments. All patients on study had assessment of disease by CT scan, which was present for evaluation in the Picture Archiving and Communication System (PACS) of Memorial Sloan-Kettering Cancer Center. Information on eligibility, treatment plans, and outcomes for these institutional review board approved trials has been reported elsewhere. 17 19 Eligibility criteria for the trials included histologic confirmation of RCC, Stage IV disease, and bidimensionally measurable metastatic lesions. During treatment, CT scans were obtained every 8 weeks to assess response, and WHO response criteria were used prospectively during the trials. Tumor Measurements on CT Scans All tumor sites in each patient were measured at baseline and on all follow-up CT scans. CT scans were obtained with contiguous 7-mm-thick slices that usually were obtained after administration of both oral and intravenous contrast. If the patient could not tolerate intravenous contrast (generally due to abnormal renal function), then a noncontrast study was obtained. Tumor measurements were obtained on a PACS workstation (GE Medical Systems, Chicago, IL)

Response Assessment in Renal Cell Carcinoma/Schwartz et al. 1613 with an electronic caliper tool that allows the user to draw a thin electronic line on the computer monitor. Images could be magnified and window/level settings adjusted at the radiologist s discretion to best display each tumor deposit. The largest perpendicular dimensions were obtained for each tumor deposit and were recorded by one radiologist (L.W.). The site of each tumor deposit was recorded. Response Categories Unidimensional (greatest dimension) percent changes were categorized according to the following RECIST classification criteria: complete response (CR), the disappearance of all disease; partial response (PR), a decrease 30% in the sum of the greatest dimensions of all measurable disease; stable disease (SD), a decrease 30% and an increase 20% in the sum of greatest tumor dimensions; and progressive disease (PD), an increase 20% in the sum of the greatest tumor dimensions. Bidimensional (cross-product) percent changes were categorized according to the following WHO classification criteria: CR, the disappearance of all disease; PR, a decrease 50% in the sum of the cross products; SD, a decrease 50% and an increase 25% in the sum of the cross products; and PD, an increase 25% in the sum of the cross products. FIGURE 1. Diagrammatic explanation of the normalized lesion index response assessment method and comparison with conventional bidimensional response assessment (World Health Organization criteria). Assessment of Outcome Endpoint The best overall response was calculated for each patient based on the best response category over all follow-up scans using the standard response categories and the rules of categorization for both the conventional WHO and RECIST criteria. Possible overall responses, in decreasing order of goodness, were CR, PR, SD, and PD. Overall disagreement rates and the disagreement rate for each response category were calculated. The time to disease progression was calculated from the date of the baseline scan to the date of the first scan that showed PD using standard WHO and RECIST criteria. Progression based on RECIST criteria was an increase 20% in the sum of the greatest tumor dimension. Progression based on WHO criteria was an increase 25% in the sum of the cross products. Kaplan Meier curves for the time to disease progression were plotted. 20 A paired Prentice Wilcoxon test 21 was used to compare the time to disease progression between any two curves. Patients were divided into two groups based on whether their primary renal tumor still was in situ or had been removed. For patients who had primary tumors in situ, analyses were performed both by including the measurements of their primary renal tumor (as well as all measurable disease) at baseline and at follow-up and by excluding the primary renal tumor measurements. For patients without primary tumors in situ, response assessment was calculated based on all measurable metastatic disease. To account for variability in size of tumor deposits for each patient, both the best overall response and the time to disease progression were calculated using a novel response parameter, which we call the normalized lesion index. The normalized lesion index compensates for the fact that the primary renal tumor is generally a dominant mass that does not change in size substantially during therapy and, thus, may dampen the impact of disease progression or response of smaller lesions when those smaller lesions simply are summed. For this method, the percent change was calculated individually for each observed tumor; then, the average of these percentages at each follow-up was categorized as a response assessment using both the RECIST and WHO criteria, as described above. The differences between the new normalized lesion index and the conventional WHO weighted sum of the product response assessment are shown in Figure 1. The best overall response and the time to disease progression were evaluated with the conventional RE- CIST and WHO criteria and were compared with similar unidimensional and bidimensional measurements analyzed with the normalized lesion index. This was done for the both the entire data set of assessments and for the subset of assessments with the primary tumor intact, in which it is anticipated that the difference is greatest between the conventional RECIST and WHO versus the normalized lesion index.

1614 CANCER October 15, 2003 / Volume 98 / Number 8 RESULTS Patients and Lesions We evaluated 36 male patients and 8 female patients with a mean age of 62 years (range, 31 81 years). Evaluations were performed after treatment with pegylated interferon- in 29 patients and thalidomide in 24 patients. A total of 236 scans were analyzed: 53 baseline scans and 183 follow-up scans (range, 1 20 follow-up scans). An average of 4.45 scans per patient (range, 2 21 scans per patient) were measured. Scans were obtained at the same time intervals in all patients. There was an average of 4.0 tumor deposits (lesions) at baseline (range, 1 19 deposits per patient). Sites of metastases included lung, liver, lymph node, peritoneum, and renal fossa. Unidimensional Analysis versus Bidimensional Analysis Classifications of the best overall response and the time to disease progression were calculated for the 53 assessments in 44 patients both unidimensionally (with RECIST criteria) and bidimensionally (with WHO criteria) (Table 1). The best overall response classification differed in 6 of 53 assessments (11%). For 4 of 9 assessments (44%), the discrepancy was categorized as SD with RECIST criteria and PD with WHO criteria. There were fewer discrepancies in the partial response category. The classification of time to disease progression was significantly longer when lesions were measured with unidimensional parameters compared with bidimensional parameters (P 0.001). The mean time to disease progression assessed with unidimensional criteria was 9.2 months, compared with 6.4 months assessed with bidimensional criteria (Fig. 2). This shift in response favored the SD category with RECIST criteria and the PD category with WHO criteria. FIGURE 2. Time to progression obtained using unidimensional (Response Evaluation Criteria in Solid Tumors [RECIST]; solid line) and bidimensional (World Health Organization [WHO]; dashed line) measurement techniques. Impact of the Primary Tumor Twenty-four of 53 assessments (45%) involved primary renal tumors in situ. The analysis of tumor assessment classification with and without the primary tumor was performed on this subset of patients. The mean greatest dimension of the primary tumor at baseline was 8.9 cm (range, 2.2 13.1 cm). The mean size of the metastases at baseline was 2.6 cm (range, 1.0 8.7 cm). The best overall response was calculated for each patient both including and excluding measurements of the primary tumor. Table 2 lists the best overall response classifications with and without the primary tumor for both unidimensional and bidimensional criteria. For unidimensional measurements, there was a difference in best overall response categorization in 6 of 24 patients, resulting in an overall disagreement rate of 25% with and without the primary tumor measurement. For bidimensional measurements, 10 of 24 patients had a different response assessment categorization, resulting in an overall disagreement rate of 42%. In general, more patients were categorized with SD when the primary RCC was included in the measurement. Specifically, when measuring unidimensionally and including the primary RCC tumor, 20 patients were classified with SD; however, without the primary RCC tumor, only 17 patients were classified with SD. Similarly, when measuring bidimensionally and including the primary RCC tumor, 19 patients were classified

Response Assessment in Renal Cell Carcinoma/Schwartz et al. 1615 with SD; however, without including the primary RCC tumor, only 14 patients were classified with SD. In the same manner, the time to disease progression was calculated for each patient both including and excluding the primary renal tumor. For unidimensional measurements, the median time to progression increased from 5.4 months without the primary tumor to 9.6 months with the primary tumor included in the measurement (difference, 4.2 months). For bidimensional measurements, the median time to progression increased from 4.0 months without the primary tumor to 9.6 months with the primary tumor included in the measurement (difference, 5.6 months). For patients with primary tumors, inclusion of the primary tumor resulted in a significantly longer median time to disease progression when the lesions were measured either unidimensionally (P 0.013) or bidimensionally (P 0.006). New Method: The Normalized Lesion Index The best overall response and the time to disease progression were calculated for all 53 assessments (44 patients) using the normalized lesion index. A percentage change in tumor size (using either unidimensional or bidimensional measurements) was calculated for each lesion, and the percentage changes were averaged. A categoric response based on RECIST or WHO criteria then was determined for the average percentage change. Figure 1 illustrates the difference between the conventional response assessment and assessment with the new normalized lesion index. The best overall response for all 53 assessments (44 patients) using both the conventional WHO and RECIST methods of therapy response assessment and the normalized lesion index method is shown in Table 3. There is a tendency, both in unidimensional measurements and in bidimensional measurements, to classify patients with SD using the conventional method and with PD using the normalized lesion index method. Similarly, one of two patients who had a PR according to the normalized lesion index method would have been classified with SD according to the conventional method. This is because the primary renal tumor is generally a dominant mass that does not change in size and lessens the impact of disease progression or response of smaller lesions when the lesions are summed; however, in the normalized lesion index method, each individual lesion measurement is counted equally. Figure 3 illustrates the comparison of the time to disease progression calculated unidimensionally and bidimensionally using the con-

1616 CANCER October 15, 2003 / Volume 98 / Number 8 FIGURE 3. Time to disease progression obtained using conventional World Health Organization weighted (dashed line) and normalized lesion index (solid line) bidimensional response measurement techniques in (A) the total cohort of 53 measurements in 44 patients and (B) a subset of 24 patients who had intact primary renal tumors (B). ventional technique of response assessment and using the normalized lesion index. Table 4 lists the best overall response classification with bidimensional measurements, comparing the normalized lesion index (including the kidney) with the conventional weighted average (excluding the kidney) in the 24 patients who had their primary RCC tumors in situ. The overall disagreement rate was 17%. This demonstrates similarity in the best FIGURE 4. Time to disease progression obtained using World Health Organization bidimensional measurements in 24 patients with primary renal tumors in situ, comparing the normalized lesion index response including the primary and conventional weighted response excluding excluding the primary tumor. overall response between the conventional method without the kidney included in the sum of measurements and the normalized lesion index with the kidney included in the sum of measurements. The time to disease progression according to each of these two scenarios is plotted in Figure 4. There was no significant difference in the time to progression comparing these two methods; therefore, this new parameter may be a useful technique for assessing patients with differences in the size of measurable

Response Assessment in Renal Cell Carcinoma/Schwartz et al. 1617 lesions. In our series, this group was comprised of patients who still had their primary renal tumors in situ. DISCUSSION Standardization of response assessment criteria is critical for judging the success or failure of a therapy in a clinical trial and in directing patient care. The results of studies are compared to identify new agents that may be beneficial and possibly better than a current standard therapy. The only way to gauge this success or failure accurately is to assure that the response assessment criteria used with new agents are applied in the same fashion as in the comparative studies. Although comparison among individual studies is not definitive in comparing treatments, it may provide guidance in selecting a novel treatment strategy for a Phase III trial. Specialized response assessment criteria have been developed for certain tumors. For example, an international workshop defined response parameters for non-hodgkin lymphoma. 22 Some of the response parameters used in this specialized set of criteria differ from standards developed in the WHO criteria and modified in the RECIST criteria. There is significant interobserver variability in RCC response assessment. 23 Thiesse et al. identified a number of factors that accounted for disagreement in response assessments, including errors in tumor measurements, errors in selection of measurable targets, and technical radiologic problems. Our data identify two additional potential sources for disagreement in RCC therapy response assessment classification. Despite the endorsement of unidimensional measurements by the RECIST group and others, 24 acceptance of these criteria is not universal. 25 Our study demonstrates that there is both a moderate difference in best overall response classification (Fig. 5) and a significant difference in time to progression classification when assessment is made with unidimensional measurements versus bidimensional measurements. Judging which measurement criteria are superior is beyond the scope of our study; however, we have demonstrated that these two measurement systems may yield different response classifications; results obtained with one method may not be comparable with results generated by the other method. Neither the WHO criteria nor the RECIST criteria address whether measurements of a patient s primary tumor should be included in the assessment of therapy response. In many instances, such as in RCC, the primary tumor already may have been resected and thus cannot be included in the measurement. In those patients who have a primary tumor that is intact and FIGURE 5. Axial computed tomography scans of the chest at (A) baseline and (B) 4 months of follow-up. The mediastinal lymph node increased by 18% unidimensionally and by 53% bidimensionally, indicating a difference in response assessment at this time point for this lesion. can be measured, our data show that inclusion of this measurement tends to lead to an apparent increase in time to progression and an increase in the number of patients classified with SD, because a primary RCC tumor tends to be substantially larger than its metastases. In addition, selection of patients according to prior nephrectomy may influence the apparent outcome assessment for a new agent studied in a clinical trial. Differences in patient population regarding extent of metastases and presence or absence of renal primary tumor likely contribute to the wide variation in response rates that has been reported among patients treated with cytokine combination therapy with interferon- plus interleukin-2. 26 In one series of patients undergoing nephrectomy, it was reported that the average size of resected primary tumors was 7.0 cm, and the size was correlated with the measurements obtained from the preoperative CT scan. 27 Most metastases from RCC, however, are smaller than 7.0 cm. Therefore, when summing either the greatest dimensions or the cross products, the primary RCC tumor usually will play a dominant role mathematically in the response assessment classification. Frequently, the primary tumor changes relatively little in size, despite an increase or decrease in the size of its metastatic lesions. This often results in the categorization of these patients as having SD. Although these patients may actually have SD,

1618 CANCER October 15, 2003 / Volume 98 / Number 8 other investigators may choose not to include measurements of the primary renal tumor and thus would report that similar patients either had progressive disease or responded to a similar or different therapy. In addition, comparison of patient responses within the same trial or cohort may be difficult if some have a primary renal tumor included in the measurement method and others do not. We have devised a new method for data analysis in studies and trials in which patients may or may not have their primary tumor intact or, in a more general sense, in patients with large differences in the size of metastatic lesions used in response assessment. This new method must be validated in larger trials; the objective in this study was to define a method of finding equivalence in response assessment classification when including patients who may or may not have their primary renal tumor intact. Instead of summing the measurements (obtained either unidimensionally or bidimensionally), we propose averaging the percentage changes in the tumor sizes. This method creates an unweighted average response assessment that we have termed the normalized lesion index. Because percentages, rather than absolute values, are averaged, the measurement of each tumor deposit will have equal weight in the response assessment. Our data showed that the normalized lesion index response assessment methodology was equivalent to the conventional method when the primary tumor was excluded. It is not desirable to exclude the primary RCC, because this is a part of the patient s extent of disease. However, including the primary and summing the measurements according to conventional methods will result in significantly different times to disease progression for these patients. The outcome of trials that have varying numbers of patients with their primary tumors intact and measured will influence the reported effect of the therapeutic agent on both the best overall response and the time to progression. The normalized lesion index method equalizes the response assessment classification for individual tumors and patients who may or may not have their primary tumor intact. It is possible that larger tumors grow more slowly and that rapidly growing metastases may cause clinical decline. By simply summing the lesions, the smaller metastases would be overshadowed by the larger lesions. This normalized lesion index may address this problem by equalizing the impact of changes in the size of these lesions. Further study of this technique is warranted, including its use in larger patient cohorts, correlation with other response parameters, and study in other types of solid tumor malignancies in which either the primary tumor or metastatic lesions vary substantially in size. Another potential approach would be to report the data analysis in two ways: one with and one without the primary mass included. In summary, our results underscore the proposition that response assessment classification obtained with unidimensional measurements may not be the same as the classification obtained with bidimensional measurements and that inclusion of the primary renal tumor in the measurement of tumor burden can affect profoundly the classification of a patient s response to therapy. This information should be considered in clinical trial design, in clinical trial interpretation, and in the interpretation of a given patient s response (or apparent lack of response) to therapy. We also have defined the normalized lesion index, which accounts for differing tumor sizes. This index needs further validation and also may be used to examine response assessment methods used for other types of tumors. REFERENCES 1. Motzer RJ, Bander NH, Nanus DM. Renal cell carcinoma. N Engl J Med. 1996;335:865 875. 2. Vogelzang NJ, Stadler WM. Kidney cancer. Lancet. 1998;352: 1691 1696. 3. Bukowski R. Immunotherapy in renal cell carcinoma. Oncology. 1999;13:801 813. 4. Figlin RA. Renal cell carcinoma: management of advanced disease. J Urol. 1999;161:381 387. 5. Pyrhonen S, Salminen E, Ruutu M, et al. Prospective randomized trial of interferon alfa-2a plus vinblastine versus vinblastine alone in patients with advanced renal cell carcinoma. J Clin Oncol. 1999;17:2859 2867. 6. Medical Research Council and Collaborators. Interferonalpha and survival in metastatic renal carcinoma: early results of a randomized controlled trial. Lancet. 1999;353:14 17. 7. Amato D. Thalidomide for recurrent renal cell cancer in a 40-year-old man. Oncology. 2000;12(Suppl 13):33 36. 8. Eisen T, Boshoff C, Sapunar F, et al. Continuous low-dose thalidomide: a Phase II study in advanced melanoma, renal cell, ovarian and breast cancer. Br J Cancer. 1999;82:812 817. 9. Motzer RJ, Bacik J, Murphy BA, Russo P, Mazumdar M. Interferon-alfa as a comparative treatment for clinical trials of new therapies against advanced renal cell carcinoma. J Clin Oncol. 2001;20:289 296. 10. Motzer RJ, Schwartz L, Law TM, et al. Interferon alfa-2a and 13-cis retinoic acid in renal cell carcinoma: antitumor activity in a Phase II trial and interactions in vitro. J Clin Oncol. 1995;13:1950 1957. 11. Motzer RJ, Murphy B, Bacik J, et al. Phase III trial of interferon alfa-2a with or without 13-cis retinoic acid for patients with advanced renal cell carcinoma. J Clin Oncol. 2000;18: 2972 2980. 12. World Health Organization Staff. WHO handbook for reporting results of cancer treatment. Geneva: World Health Organization, 1979.

Response Assessment in Renal Cell Carcinoma/Schwartz et al. 1619 13. Therasse P, Arbuck SG, Eisenhauer EA, et al. New guidelines to evaluate the response to treatment in solid tumors. J Natl Cancer Inst. 2000;92:205 216. 14. James K, Eisenhauer E, Christian M, et al. Measuring response in solid tumors: unidimensional versus bidimensional measurement. J Natl Cancer Inst. 1999;91:523 528. 15. Motzer RJ, Russo P. Systemic therapy for renal cell carcinoma. J Urol. 2000;163:408 417. 16. Stadler WM, Kuzel T, Dumas M, et al. A multicenter Phase II trial of interleukin-2, interferon-alfa, and 13 cis-retinoic acid in patients with metastatic renal cell carcinoma. J Clin Oncol. 1998;16:1820 1825. 17. Motzer RJ, Berg W, Ginsberg M, et al. Phase II trial of thalidomide for patients with advanced renal cell carcinoma. J Clin Oncol. 2002;20:302 306. 18. Motzer RJ, Rakhit A, Ginsberg M, et al. Phase I Trial of 40-kD branched pegylated interferon alfa-2a for patients with advanced renal cell carcinoma. J Clin Oncol. 2001;19:1312 1319. 19. Motzer RJ, Rakhit A, Thompson J, et al. Phase II branched pegylated interferon-alfa 2a (40 kda) for patients with advanced renal cell carcinoma. Ann Oncol. 2002;13:1799 1805. 20. Kaplan EL, Meier P. Nonparametric estimation from incomplete observations. J Am Stat Assoc. 1958;53:457 481. 21. Woolson RF, O Gorman TW. A comparison of several tests for censored paired data. Stat Med. 1992;11:193 208. 22. Cheson AD, Horning SJ, Coiffer B, et al. Report of an international workshop to standardize response criteria for non- Hodgkin s lymphomas. J Clin Oncol. 1999;17:1244 1253. 23. Thiesse P, Ollivier L, Di Stefano-Louineau D, et al. Response rate accuracy in oncology trials: reasons for interobserver variability. J Clin Oncol. 1997;15:3507 3514. 24. Dachman AH, MacEneaney PM, Adedipe A, et al. Tumor size on computed tomography scans: is one measurement enough? Cancer. 2001;91:555 560. 25. Saini S. Radiologic measurement of tumor size in clinical trials: past, present, and future [opinion]. AJR Am J Roentgenol. 2001;176:333 334. 26. Vogelzang NJ, Lipton A, Figlin RA. Subcutaneous interleukin-2 plus interferon alfa-2a in metastatic renal cell cancer; an outpatient multicenter trial. J Clin Oncol. 1993;11:1809 1816. 27. Irani J, Humbert M, Lecocq B, et al. Renal tumor size: comparison between computed tomography and surgical measurements. Eur Urol. 2001;39:300 303.