Response Assessment in Pediatric Rhabdomyosarcoma: Can Response Evaluation Criteria in Solid Tumors Replace Three-dimensional Volume Assessments?

Transcription

1 Note: This copy is for your personal non-commercial use only. To order presentation-ready copies for distribution to your colleagues or clients, contact us at Original Research n Pediatric Imaging Response Assessment in Pediatric Rhabdomyosarcoma: Can Response Evaluation Criteria in Solid Tumors Replace Three-dimensional Volume Assessments? 1 Reineke A. Schoot, MD Kieran McHugh, MD Rick R. van Rijn, MD, PhD Leontien C. M. Kremer, MD, PhD Julia C. Chisholm, MD, PhD Huib N. Caron, MD, PhD Johaes H. M. Merks, MD, PhD Purpose: Materials and Methods: To investigate (a) interobserver variability for three-dimensional (3D) (based on European Pediatric Soft-Tissue Sarcoma Study Group [EpSSG] guidelines) and one-dimensional (1D) (based on Response Evaluation Criteria in Solid Tumors [RECIST]) response assessments, (b) intermethod variability between EpSSG guidelines and RECIST, and (c) clinically relevant consequences of interobserver and intermethod variability in pediatric patients with rhabdomyosarcoma. The study was approved by the Academic Medical Center Ethics Committee and the Great Ormond Street Hospital Ethics Committee; both committees waived the requirement for informed consent because of the retrospective nature of the study. Data were analyzed from 124 consecutive male and female children and young adults (age range, 1 18 years) with rhabdomyosarcoma at two institutions ( ) with relevant imaging studies. Tumors were measured by two radiologists (1D and 3D measurements) at diagnosis and after induction chemotherapy. Interobserver variability was analyzed by using three different tests, and the intermethod variation was calculated. 1 From the Departments of Pediatric Oncology (R.A.S., L.C.M.K., H.N.C., J.H.M.M.) and Pediatric Radiology (R.R.v.R.), Emma Children s Hospital-Academic Medical Center (EKZ-AMC), PO Box 22660, 1100 DD Amsterdam, the Netherlands; Department of Radiology, Great Ormond Street Hospital for Children, London, England (K.M.); and Children and Young People s Department, Royal Marsden Hospital, Sutton, England (J.C.C.). Received December 15, 2012; revision requested January 27, 2013; revision received May 13; accepted May 21; final version accepted June 6. Supported by a grant from F. Hoffma-La Roche, the Tom Voȗte Foundation, and the National Health Service through funding of the National Institute for Health Research Biomedical Research Centre of the Royal Marsden Hospital. Address correspondence to J.H.M.M. ( j.h.merks@amc.uva.nl). Results: Sixty-four eligible patients were included (median age, 4.6 years). Agreement between observers for EpSSG guidelines and RECIST was moderate (k = and 0.592, respectively); interobserver variation led to different potential treatment decisions in nine (14%) and 11 (17%) of the 64 patients, respectively. Comparison of EpSSG guidelines and RECIST resulted in 13 discrepant response classifications (20%), which were equally distributed (under- and overestimation of response) and led to consequences for treatment choice in five patients (8%). Conclusion: EpSSG guidelines and RECIST are not interchangeable; neither technique demonstrated superiority in this study. These findings should be taken into account in future study protocol design. q RSNA, 2013 Supplemental material: /suppl/doi: /radiol /-/dc1 q RSNA, radiology.rsna.org n Radiology: Volume 269: Number 3 December 2013

2 Historically, response in pediatric soft-tissue sarcoma trials has been defined by means of two-dimensional and, later, three-dimensional (3D) tumor measurements (1 4). In current adult oncology trials, however, response assessment is defined according to one-dimensional (1D) measurements based on the Response Evaluation Criteria In Solid Tumors (RECIST) guidelines (5,6). As the European Union introduced a new pediatric regulation to facilitate the development and availability of new agents for children in 2006, pharmaceutical industries have been using RECIST methodology for response assessment (7,8). However, these guidelines were developed for, and validated in, adult oncology patients, the majority of whom had metastatic disease (6,9). It remains unclear whether it is appropriate to use the 1D Advances in Knowledge Neither three-dimensional (3D) measurement (based on European Pediatric Soft-Tissue Sarcoma Study Group [EpSSG] guidelines) nor one-dimensional (1D) measurement (based on Response Evaluation Criteria in Solid Tumors [RECIST]) of tumor response showed superiority in this cohort study. Interobserver and intermethod variability for both EpSSG guidelines (3D measurement) and RECIST (1D measurement) was significant and resulted in different response classifications, with consequences for treatment in a considerable proportion of patients; interobserver variability was 14% for 3D measurement and 17% for 1D measurement; intermethod variability was 8%. Response rates according to EpSSG guidelines and RECIST were 63% and 59%, respectively (P =.86), indicating that 1D measurement did not underestimate response in pediatric patients with rhabdomyosarcoma, as had been feared by many pediatric oncologists. RECIST guidelines for pediatric solid tumor response assessments, which are currently based on 3D measurements (10,11). The rationale of RECIST, that is, the correlation between diameter and tumor load, is based on the mathematic approximation of tumors as spherical lesions. In spherical lesions, the logarithm of cell number correlates with diameter, independently of the initial tumor size (9). Many pediatric solid tumors are irregular in shape and lack a spherical appearance. When the length of an irregularly shaped tumor becomes more than twice its width, diameter becomes inaccurate as an estimation for tumor size (12,13). Furthermore, when treated with chemotherapy these tumors can remain stable in the longest diameter, whereas the tumor regresses in one or two other dimensions (11,14). Such a response to chemotherapy is clinically important because it may facilitate surgical resectability, which is a major component of local treatment in pediatric soft-tissue sarcoma. To gain further insight into the effect of using RECIST response assessment, we initiated a retrospective cohort study in pediatric patients with rhabdomyosarcoma treated according to International Society of Pediatric Oncology Malignant Mesenchymal Tumor (SIOP-MMT) and, later, European Pediatric Soft-Tissue Sarcoma Study Group (EpSSG) protocols (4,15). The aims of the study were to investigate (a) interobserver variability for EpSSG (3D measurement) and RECIST (1D measurement) response assessments individually, (b) intermethod variability between EpSSG and RECIST, and (c) Implications for Patient Care Tumor response assessments in pediatric rhabdomyosarcoma with use of either EpSSG 3D criteria or RECIST should be judged with caution. Care should be taken in choosing tumor assessment methods because different methods can lead to different treatment decisions. clinically relevant consequences of interobserver and intermethod variability for both methods in pediatric patients with rhabdomyosarcoma. Materials and Methods Travel and accommodation costs were funded by Roche. The authors had control of the data and the information submitted for publication. Those authors who are not consultants for Roche had control of inclusion of any data and information that might present a conflict of interest for those authors who are consultants for Roche. Patients This study was approved by the Academic Medical Center Ethics Committee and the Great Ormond Street Hospital Ethics Committee; both committees waived the requirement for informed consent because of the retrospective nature of the study. We used clinical data from the medical records of consecutive patients (aged 0 18 years) presenting to Hospital 1 (Emma Children s Hospital Academic Medical Center, Amsterdam, the Netherlands) or Hospital 2 (Great Ormond Street Hospital for Children NHS Trust, Published online before print /radiol Content code: Radiology 2013; 269: Abbreviations: CI = confidence interval EpSSG = European Pediatric Soft-Tissue Sarcoma Study Group 1D = one-dimensional RECIST = Response Evaluation Criteria in Solid Tumors SIOP-MMT = International Society of Pediatric Oncology- Malignant Mesenchymal Tumor 3D = three-dimensional Author contributions: Guarantors of integrity of entire study, R.A.S., K.M., J.H.M.M.; study concepts/study design or data acquisition or data analysis/interpretation, all authors; manuscript drafting or manuscript revision for important intellectual content, all authors; manuscript final version approval, all authors; literature research, R.A.S., K.M., J.H.M.M.; clinical studies, R.A.S., R.R.v.R., L.C.M.K., J.H.M.M.; statistical analysis, R.A.S., K.M., L.C.M.K., J.H.M.M.; and manuscript editing, all authors Conflicts of interest are listed at the end of this article. Radiology: Volume 269: Number 3 December 2013 n radiology.rsna.org 871

3 Table 1 Response Classification for 3D and 1D Measurements Response Class EpSSG (3D Measurement) RECIST (1D Measurement) Adjusted RECIST (1D Measurement) CR 100% decrease 100% decrease 100% decrease PR 66% but,100% decrease 30% but,100% decrease 30% but,100% decrease mpr.33% but,66% decrease NA.12%* but,30% decrease SD Neither mpr nor PD Neither PR nor PD Neither mpr nor PD PD 40% increase 20% increase 12%* increase Note. CR = complete remission, mpr = minor partial response, NA = not applicable, PD = progressive disease, PR = partial response, SD = stable disease. * Adjusted to EpSSG guidelines. Equivalent changes for maximum diameter and volume in spherical tumors: r = radius, diameter = 2r, volume = 4/3 pr 3. Considering a decrease in volume of 33%: Old volume = 4/3 pr 3 ; the new volume after 33% decrease = (4/3 pr 3 ), where 4/3 p = a constant = k, k 3 r 3 = k (0.67r 3 ). So the new r = = 0.88 of the original, that is, a decrease in 12% in diameter. London, England) with a biopsy-proved rhabdomyosarcoma. We included patients with measurable disease at the primary tumor site after initial surgery (primarily incisional biopsy) (determined by J.H.M.M. and J.C., with 11 and 14 years of experience as consultant pediatric oncologists, respectively) diagnosed between January 1999 and December Tumors were classified according to the clinical TNM system (16) and the Intergroup Rhabdomyosarcoma Study postsurgical staging system (17). Localization of the primary tumor was defined as described in the EpSSG Rhabdomyosarcoma 2005 protocol (4). Patients were treated according to the guidelines of SIOP-MMT 95 from 1999 to 2005 and according to the EpSSG Rhabdomyosarcoma 2005 protocol from 2006 to 2009 (4,15). Tumor Measurements The maximal computed tomographic (CT) section thickness permitted was 5 mm. When possible, the same magnetic resonance (MR) imaging sequence was used for sequential measurements in the following order of preference: contrast material enhanced T1- and T2- weighted sequences and unenhanced T1-weighted sequences (Appendix E1 [online]). Primary tumors were measured on CT or MR images at diagnosis and after induction chemotherapy (two or three courses). Imaging studies were prospectively and independently evaluated by two experienced pediatric radiologists (K.M. and R.R.v.R., with 22 and 10 years of experience as pediatric radiologists, respectively). To minimize interobserver variability, a concise guideline was developed summarizing the EpSSG and RECIST measurement protocols (Appendix E2 [online]). Tumors were measured according to the EpSSG Rhabdomyosarcoma 2005 and RECIST protocols (4 6). For EpSSG, the two maximal perpendicular diameters of the primary tumor were measured in the axial plane on the section with the largest tumor surface area. The craniocaudal tumor dimension was measured at MR imaging with the sagittal or coronal sequences and, if available in the original dataset, on coronal reconstructions at CT. Otherwise, this was derived from the sum of the number of sections containing visible tumor (calculated as X times the section thickness and interval minus one section interval). The longest diameter for 1D RECIST was measured in any plane on MR images or CT scans. For CT scans without coronal reconstructions, the longest diameter was measured in the axial plane only (6). Response Classification Response of the primary tumor after induction chemotherapy was calculated as the proportional increase or decrease in volume (3D measurement) or diameter (1D measurement) compared with the size of the tumor at diagnosis. EpSSG response assessment is based on the volume of the primary tumor (in cubic centimeters), which is calculated as follows: p/6 3 a 3 b 3 c = a 3 b 3 c, where a = height (in centimeters), b = width (in centimeters), and c = depth (in centimeters) (4). Because the change in volume is often larger than an equivalent change in diameter, the proportional change in tumor volume does not show a linear relationship with the proportional change in diameter. EpSSG and RECIST protocols use different response cutoffs. Therefore, to facilitate a direct comparison between 3D and 1D tumor response assessment, RECIST was adjusted to EpSSG cutoff values and definitions (Table 1). Equivalent cutoff values that do not exist in RECIST were calculated with a translation formula (Table 1). Statistical Analyses Interobserver variability and clinical consequences for EpSSG and RECIST measurements. The interobserver variability of response assessments according to EpSSG and RECIST was calculated by using three different methods: the intraclass correlation coefficient, limits of agreement, and k statistics. First, we visualized the correlation between the two observers in response assessment (as continuous outcome variables; proportional change in size) in a direct correlation plot. To correct for differences between the observers (under- or overestimation), the intraclass correlation coefficient was calculated (18). Meaningful agreement is attained if the intraclass correlation 872 radiology.rsna.org n Radiology: Volume 269: Number 3 December 2013

4 coefficient is at least 0.75 (18). Second, we visualized the scale of the difference between the observers for EpSSG and RECIST measurements in a Bland- Altman plot by plotting the mean value of both observers against the difference between the observers (19). To estimate the 95% confidence interval (CI) of the difference between the observers, we calculated the corresponding limits of agreement (mean difference between observers 6 2 standard deviations). Third, we used k statistics to assess the correlation between both observers for categoric outcomes (complete remission, partial response, stable disease, progressive disease) for EpSSG and RECIST measurements separately. k values of indicate almostperfect agreement, substantial agreement, moderate agreement, fair agreement, and slight agreement (20). To evaluate the clinical consequences of the interobserver variability, we calculated the proportion of patients who were classified differently along with the corresponding 95% CI. Subsequently, we compared the proportion of response classifications that would have resulted in different treatment decisions owing to interobserver variability: responders (minor partial response, partial response, complete remission) versus nonresponders (stable disease, progressive disease). In current EpSSG protocols, this achievement of response is decisive in treatment continuation (responder) or switch to second-line therapy (nonresponder). Intermethod variability and clinical consequences. Discrepant measurements (for each method) between the two observers were resolved in a consensus meeting. Measurements from this meeting were used in the analysis of intermethod variability. To evaluate the clinical relevance of the differences between the two methods, we calculated the proportion of patients who were classified differently along with the corresponding 95% CI. Subsequently, we analyzed the proportion of patients who would require a different treatment decision when another response assessment method Figure 1 was used. For these comparisons, we used the definitions of EpSSG (3D measurement) and the above-mentioned adjusted RECIST (1D measurement) (Table 1). We also investigated the consequences of implementing unadjusted RECIST, as requested in pharmaceutical trials, in EpSSG soft-tissue sarcoma protocols and its influence on responsebased treatment decisions. Response rate was defined as the proportion of patients achieving at least partial response. Results Patient Characteristics During a period of 10 years, 124 consecutive patients were identified. Sixty of these patients were excluded, leaving 64 patients (Fig 1). There were 33 boys (median age, 4.5 years; age range, years) and 31 girls (mean age, 5.0 years; age range, Figure 1: Flowchart shows patient enrollment. FUP = follow-up, IRS group I II = Intergroup Rhabdomyosarcoma Study clinical groups I and II (ie, patients with microscopic and macroscopic radical surgery at diagnosis, respectively [17]). 5 Images were not available digitally or in hard copy because they were sent back to referring hospitals, were not retrievable from archives, or were not obtained at prescribed time point. 5 An imaging modality other than MR imaging or CT was used for follow-up. 5 Images were of poor quality because hard copies were not suitable for adequate measurements, images were obtained elsewhere, often not expecting malignant disease, and therefore not adequate for rhabdomyosarcoma imaging. Poor image quality was determined by K.M. and R.R.v.R. (with 22 and 10 years of experience in pediatric radiology, respectively). years). One hundred seventeen MR images and 11 CT scans were analyzed by the two observers. Eight patients had undergone imaging with a different imaging modality for response assessment. Coronal reconstructions were not available in four of the CT 11 scans (36%). The median interval between studies was 9.1 weeks (interquartile range, weeks). The median age at diagnosis was 4.6 years (range, years), and 54 patients were younger than 10 years (Table 2). Thirtyone patients were treated according to the EpSSG Rhabdomyosarcoma 2005 protocol (including all five patients with metastases) and 33 were treated according to SIOP-MMT 95. Interobserver Variability and Clinical Consequences for EpSSG and RECIST Measurements The interobserver variability, which was investigated with different statistical methods, is presented in Figure 2. Agreement between observers for Radiology: Volume 269: Number 3 December 2013 n radiology.rsna.org 873

5 Table 2 Patient Characteristics Parameter Hospital 1 (n = 37) Hospital 2 (n = 27) All (n = 64) Sex M F Age No. of patients,10 y Age (y)* 5.4 ( ) 3.2 ( ) 4.6 ( ) Histologic finding Embryonal Alveolar Not otherwise specified Localization Parameningeal Orbit Head and neck Extremity Genitourinary Nonbladder prostate Bladder prostate Other TNM status T T N M Time between studies (wk)* 8.7 ( ) 9.7 ( ) 9.1 ( ) Note. Except where indicated, data are numbers of patients. * Data are medians. Numbers in parentheses are ranges. Nonparameningeal nonorbit head and neck. both the EpSSG and RECIST response assessments was moderate (intraclass correlation coefficient: and 0.732, respectively; 95% CI: 0.518, and 0.594, 0.828; mean difference 6 2 standard deviations: and ; k = and 0.592). For EpSSG, the mean difference was for responders and for nonresponders. For RECIST, the mean difference was for responders and for nonresponders. The correlation for EpSSG response assessments seemed lower for poor versus good responders, whereas correlation for RECIST was similar for responders and nonresponders (Fig 2a, 2b). A similar pattern was seen in the Bland-Altman plots (Fig 2c, 2d) and corresponding limits of agreement. This difference in distribution between EpSSG and RECIST protocols is inherent to the formula used to calculate volume (p/6 3 a 3 b 3 c). The same measurement error in a good responder results in a smaller absolute difference in the percentage decrease in volume than in a patient with stable disease or progression of the lesion. Clinical consequences of the interobserver variability of both individual methods are presented in Figure 2a and 2b. Lines in the correlation plots mark the consecutive response cutoffs (progressive disease, stable disease, minor partial response, and partial response) for the EpSSG and adjusted RECIST response assessments. Interobserver variability led to a similar proportion of patients with differences in response categories for both methods: 22 of 64 patients with EpSSG (34%; 95% CI: 23, 46) and 24 of 64 patients with adjusted RECIST (38%; 95% CI: 26, 49). These interobserver differences would lead to different treatment decisions (responders vs nonresponders) in nine of the 64 patients (14%; 95% CI: 6, 23) with EpSSG and in 11 of the 64 patients (17%; 95% CI: 8, 26) with adjusted RE- CIST (Fig 2a, 2b). Intermethod Variability and Clinical Relevance Comparison of the two response assessment methods with use of the EpSSG definitions for 3D measurements and the adjusted RECIST definitions for 1D measurements (Table 1) resulted in 13 discrepant classifications (20%; 95% CI: 10, 30) (Table 3). In seven patients (11%; 95% CI: 3, 19), response assessed with adjusted RECIST resulted in a worse response category than that assessed with EpSSG. In six other patients (9%; 95% CI: 2, 17), adjusted RECIST resulted in a better response category than EpSSG. These findings would lead to different treatment decisions in five patients (8%; 95% CI: 1, 14), switching from responders (at least minor partial response) to nonresponders (stable or progressive disease) or vice versa. The 13 discrepant cases between 3D EpSSG and adjusted RECIST were evaluated for common characteristics predisposing for discrepant measurements. Median tumor size did not predispose for discrepancies between 1D and 3D measurements at diagnosis (P =.28) or follow-up (P =.72). Forty tumors (63%) had a nonspherical shape (ratio 1.5). Discrepancies were seen in five of 40 patients (13%) with a ratio of at least 1.5 and four of 24 patients (17%) with more spherical tumors (P =.59). Discrepant cases were located in the extremities (two cases), orbit (two cases), and parameningeal area (five cases) (P =.065, not significant). Implementation of 1D measurements with use of the unadjusted RECIST definitions resulted in radiology.rsna.org n Radiology: Volume 269: Number 3 December 2013

6 Figure 2 Figure 2: Interobserver variability between EpSSG and adjusted RECIST response assessments. (a, b) Direct correlation plots and (c, d) Bland-Altman plots show interobserver variability in response assessment in 64 tumors obtained with 3D (a, c) and 1D (b, d) measurements. Horizontal and vertical lines in a and b differentiate between response classes of partial response (PR), minor partial response (mpr), stable disease (SD), and progressive disease (PD). Horizontal lines in c and d represent mean difference between observers 1 and 2 and limits of agreement. differences in response classifications (28%; 95% CI: 17, 39) compared with EpSSG (Table 4). These findings would lead to different treatment decisions in 14 patients (22%; 95% CI: 12, 32), who were classified as responders (five with partial response and nine with minor partial response) according to EpSSG but with stable disease according to RECIST; these patients would be switched to secondline chemotherapy according to current EpSSG rhabdomyosarcoma treatment protocols. Response rates according to EpSSG (where partial response is defined as 66% but,100% decrease) and RE- CIST (where partial response is defined as 30% but,100% decrease) measurements were not significantly different (63% and 59%, respectively; P =.86). Discussion The results of this study call into question the reliability of tumor response assessments in pediatric patients with rhabdomyosarcoma and demonstrate that both EpSSG (3D measurement) and RECIST (1D measurement) methodologies should be judged with caution. Interobserver variability for individual guidelines led to different treatment decisions in a considerable number of patients. Nevertheless, response rates according to EpSSG and RECIST protocols were similar in our study, indicating that 1D response assessments did not underestimate response in pediatric patients with rhabdomyosarcoma, as had been feared by many pediatric oncologists and radiologists (10,11,21). Radiology: Volume 269: Number 3 December 2013 n radiology.rsna.org 875

7 Table 3 Intermethod Variability for EpSSG and Adjusted RECIST (n = 64) EpSSG Guidelines CR (100% Decrease) PR ( 30% but,100% Decrease) Agreement between observers was moderate for both EpSSG and RECIST methods. Similar interobserver variability has been reported in adult oncology patients with metastatic disease (22 24). However, some of the articles discussing radiologic response assessment did not analyze interobserver variability (9,25 29). The development and validation of the RECIST 1.0 and 1.1 criteria, for instance, were based on measurements retrieved from a database (9,25,29). Furthermore, although three study groups evaluated the prognostic value of early response, a key issue in pediatric rhabdomyosarcoma protocols, none reported on interobserver variability (26 29). We did not assess intraobserver agreement in our study because several studies in other malignancies (22 24,30) have shown that interobserver variability Adjusted RECIST mpr (.12% but,30% Decrease) SD (Neither mpr nor PD) CR (100% decrease) PR ( 66% but,100% decrease) mpr (.33% but,66% decrease) SD (neither mpr nor PD) PD ( 40% increase) PD ( 12% Increase) Note. CR = complete remission, mpr = minor partial response, PD = progressive disease, PR = partial response, SD = stable disease. Table 4 Intermethod Variability for EpSSG and Unadjusted RECIST (n = 64) EpSSG Guidelines CR (100% Decrease) Unadjusted RECIST PR ( 30% but,100% Decrease) SD (Neither PR nor PD) CR (100% decrease) PR ( 66% but,100% decrease) mpr (.33% but,66% decrease) SD (neither mpr nor PD) PD ( 40% increase) PD ( 20% Increase) Note. CR = complete remission, mpr = minor partial response, PD = progressive disease, PR = partial response, SD = stable disease. is consistently greater than intraobserver variability and, therefore, more relevant for clinical decision making. A potential limitation of our study is the retrospective nature of our cohort. Images were acquired during a 10-year period and, inherent to the retrospective nature of our study, there was no imaging protocol implemented specifically for this study on the comparison of response assessment methods. However, all patients were included in the same consecutive international prospective trials, all with guidelines for imaging protocols. Furthermore, all original MR imaging and CT studies were prospectively and independently studied by two experienced pediatric radiologists. Therefore, we believe that the retrospective nature of our study will have had only minimal impact on the results. EpSSG guidelines (3D measurement) and RECIST (1D measurement) were not interchangeable in our study. This can be partly explained by differences in response classifications. Minor partial response, an important response category in EpSSG protocols, does not exist in RECIST. Similar subcategories of partial response are also used in other nonrhabdomyosarcoma, pediatric oncology trials (eg, neuroblastoma). In EpSSG, patients with minor partial response continue first-line treatment, whereas those with less than minor partial response switch to second-line therapy. International cooperative pediatric soft-tissue sarcoma study groups use different response requirements to continue with standard chemotherapy or switch to second-line treatment. This is the result of a continuing debate on the prognostic value of early response in patients with rhabdomyosarcoma. The EpSSG treatment regimen requires tumor reduction to at least minor partial response (.33% decrease in volume) to allow treatment continuation, whereas the North American Clinical Oncology Group only advises to switch to second-line therapy in the case of progressive disease. Implementation of early response is standard practice in European soft-tissue sarcoma protocols since Koscielniak et al (29) reported that treatment reduction in Intergroup Rhabdomyosarcoma Study III patients showing early response did not compromise survival. This practice was further continued and confirmed in the German Collaborative Soft-Tissue Sarcoma 91 study (27). Thus, if RECIST was to be implemented in current EpSSG protocols, RECIST response classification would need adjustment to include minor partial response. Concerns regarding the ease with which a patient may be considered mistakenly to have progressive disease has led several groups to adopt a higher cutoff for progressive disease in order not to erroneously deny patients effective chemotherapy (9,11,31). So perhaps in future protocols, only patients with clear and significant progressive disease should be switched to 876 radiology.rsna.org n Radiology: Volume 269: Number 3 December 2013

8 Disclosures of Conflicts of Interest: R.A.S. Financial activities related to the present article: institution received a grant from Roche for travel and accommodation costs. Financial activities not related to the present article: none to disclose. Other relationships: none to disclose. K.M. Financial activities related to the present article: is a paid consultant for Roche; received travel expenses from Roche. Financial activities not related to the present article: none to disclose. Other relationships: none to disclose. R.R.v.R. Financial activities related to the present article: institution received a grant from Roche for travel and accommodation expenses; institution received money for writing assistance, medicines, equipment, and administration support from Roche. Financial activities not related to the present article: receives royalties from Springer and Thieme. Other relationships: none to disclose. L.C.M.K. No relevant conflicts of interest to disclose. J.C.C. Financial activities related to the present article: is a consultant for the BERNIE trial (sponsored by Roche) and member of the steering committee. Financial activities not related to the present article: is a paid consultant for Roche. Other relationsecond-line chemotherapy and minor partial response should be abolished as a response category, as is current practice in adult oncology and Children s Oncology Group rhabdomyosarcoma protocols (32 34). It is debatable whether results from this study, which deals with chemosensitive tumors, can be translated into phase II settings where patients generally have refractory, relapsed, or metastatic disease. Our findings suggest that RECIST measurements have lower interobserver variability than EpSSG in poor responders (ie, stable or progressive disease). However, this is based on very small patient numbers owing to the nature of the cohort under investigation. In addition, the selection of target lesions in patients with metastatic disease causes further interobserver variability (22,24). In phase II studies, response serves a different purpose; response rate (complete remission + partial response) is often the primary or secondary endpoint because it indicates some degree of biologic antitumor activity (6). Conclusions regarding potential efficacy of a new drug are sometimes based on only a few responding patients. Therefore, a reliable method for response assessment is of utmost importance in phase II studies. The considerable interobserver variability of both EpSSG and RECIST response assessment methods shown in this study underlines the need for central review of imaging at decisive time points in phase II studies. Functional imaging techniques, such as fluorine 18 fluorodeoxyglucose positron emission tomography, are promising as predictors of outcome and might assist in future response assessments (35,36). Simplification of methodology was one of the goals of RECIST (6), and indeed using only 1D measurement is slightly less time consuming. It is of interest to note that there are software packages, although not widely available or used in daily practice, that can assess both RECIST and volume measurements in a semiautomated fashion. In conclusion, although neither method came out as superior, interobserver variability for both EpSSG (3D measurement) and RE- CIST (1D measurement) was significant in this retrospective cohort study, resulting in different response classifications with treatment consequences in a considerable proportion of patients. In chemosensitive patient cohorts, switching to second-line therapy only in the case of progressive disease could prevent decisions being made on the basis of arbitrary errorprone radiologic response assessments in a substantial proportion of patients. A large cohort study of consecutive patients with rhabdomyosarcoma treated according to uniform and modern guidelines, with recommendations to decrease the risk of inter- and intraobserver variability, is needed and should test the hypothesis that radiologic response is correlated with survival. For consistency and standardization of response assessments in future pediatric oncology trials, central review of imaging is strongly advocated. Acknowledgments: We thank Raphael Rousseau, MD, PhD, of F. Hoffma-La Roche, for critically reviewing this manuscript and for making this fundamental and important research possible through funding. In addition, we acknowledge support for third-party editorial assistance from Charlotte Keerley, PhD, of Gardiner-Caldwell Communications funded by F. Hoffma-La Roche Ltd. ships: none to disclose. H.N.C. No relevant conflicts of interest to disclose. J.H.M.M. Financial activities related to the present article: institution received a grant from Roche for travel and accommodation costs; institution received money for provision of writing assistance, medicines, equipment, or administrative support from Roche. Financial activities not related to the present article: institution receives money for consultancy from Roche. Other relationships: none to disclose. References 1. Flamant F, Rodary C, Rey A, et al. Treatment of non-metastatic rhabdomyosarcomas in childhood and adolescence: results of the second study of the International Society of Paediatric Oncology MMT84. Eur J Cancer 1998;34(7): Stevens MCG, Rey A, Bouvet N, et al. Treatment of nonmetastatic rhabdomyosarcoma in childhood and adolescence: third study of the International Society of Paediatric Oncology SIOP Malignant Mesenchymal Tumor 89. J Clin Oncol 2005;23(12): World Health Organization. WHO handbook for reporting results for cancer treatment (reporting of response, section 5 P22-27). WHO Offset Publication No. 48. Geneva, Switzerland: World Health Organization, Bisogno G, Bergeron C, Jeey M, et al. European paediatric soft tissue sarcoma study group RMS 2005: a protocol for non-metastatic rhabdomyosarcoma, search?query=eudract_number: Eisenhauer EA, Therasse P, Bogaerts J, et al. New response evaluation criteria in solid tumours: revised RECIST guideline (version 1.1). Eur J Cancer 2009;45(2): Therasse P, Arbuck SG, Eisenhauer EA, et al. New guidelines to evaluate the response to treatment in solid tumors. European Organization for Research and Treatment of Cancer, National Cancer Institute of the United States, National Cancer Institute of Canada. J Natl Cancer Inst 2000;92(3): Bosch X. Pediatric medicine: Europe follows U.S. in testing drugs for children. Science 2005;309(5742): Chisholm JC, Casanova M, Geoerger B, et al. A phase II study evaluating addition of bevacizumab to chemotherapy in childhood and adolescent patients presenting with metastatic rhabdomyosarcoma (RMS) and Radiology: Volume 269: Number 3 December 2013 n radiology.rsna.org 877

9 non-rhabdomyosarcoma soft tissue sarcoma (NRSTS) [abstr]. J Clin Oncol 2010;28(15 Suppl 1):63s. 9. James K, Eisenhauer E, Christian M, et al. Measuring response in solid tumors: unidimensional versus bidimensional measurement. J Natl Cancer Inst 1999;91(6): Barnacle AM, McHugh K. Limitations with the response evaluation criteria in solid tumors (RECIST) guidance in disseminated pediatric malignancy. Pediatr Blood Cancer 2006;46(2): McHugh K, Kao S. Response evaluation criteria in solid tumours (RECIST): problems and need for modifications in paediatric oncology? Br J Radiol 2003;76(907): Kao SC, Strabala NM, Berbaum KS, et al. CT volume estimation of primary abdominal tumours using ellipsoid model in children. Pediatr Radiol 2001;31(Suppl 1):S26 S Spears CP. Volume doubling approach to tumor measurement. J Clin Oncol 1985; 3(11): Husband JE, Schwartz LH, Spencer J, et al. Evaluation of the response to treatment of solid tumours: a consensus statement of the International Cancer Imaging Society. Br J Cancer 2004;90(12): Oberlin O, Rey A, Sanchez de Toledo J, et al. Randomized comparison of intensified six-drug versus standard three-drug chemotherapy for high-risk nonmetastatic rhabdomyosarcoma and other chemotherapysensitive childhood soft tissue sarcomas: long-term results from the International Society of Pediatric Oncology MMT95 study. J Clin Oncol 2012;30(20): Harmer M. The case for TNM. Clin Oncol 1977;3(2): Maurer HM, Beltangady M, Gehan EA, et al. The Intergroup Rhabdomyosarcoma Study-I: a final report. Cancer 1988;61(2): Lee J, Koh D, Ong CN. Statistical evaluation of agreement between two methods for measuring a quantitative variable. Comput Biol Med 1989;19(1): Bland JM, Altman DG. A note on the use of the intraclass correlation coefficient in the evaluation of agreement between two methods of measurement. Comput Biol Med 1990;20(5): Landis JR, Koch GG. The measurement of observer agreement for categorical data. Biometrics 1977;33(1): McHugh K, Kao S. Can paediatric radiologists resist RECIST (response evaluation criteria in solid tumours)? Pediatr Radiol 2003;33(11): Erasmus JJ, Gladish GW, Broemeling L, et al. Interobserver and intraobserver variability in measurement of non-small-cell carcinoma lung lesions: implications for assessment of tumor response. J Clin Oncol 2003;21(13): Hopper KD, Kasales CJ, Van Slyke MA, Schwartz TA, TenHave TR, Jozefiak JA. Analysis of interobserver and intraobserver variability in CT tumor measurements. AJR Am J Roentgenol 1996;167(4): Suzuki C, Torkzad MR, Jacobsson H, et al. Interobserver and intraobserver variability in the response evaluation of cancer therapy according to RECIST and WHO criteria. Acta Oncol 2010;49(4): Bogaerts J, Ford R, Sargent D, et al. Individual patient data analysis to assess modifications to the RECIST criteria. Eur J Cancer 2009;45(2): Burke M, Anderson JR, Kao SC, et al. Assessment of response to induction therapy and its influence on 5-year failure-free survival in group III rhabdomyosarcoma: the Intergroup Rhabdomyosarcoma Study-IV experience a report from the Soft Tissue Sarcoma Committee of the Children s Oncology Group. J Clin Oncol 2007;25(31): Dantonello TM, Int-Veen C, Harms D, et al. Cooperative trial CWS-91 for localized soft tissue sarcoma in children, adolescents, and young adults. J Clin Oncol 2009;27(9): Ferrari A, Miceli R, Meazza C, et al. Comparison of the prognostic value of assessing tumor diameter versus tumor volume at diagnosis or in response to initial chemotherapy in rhabdomyosarcoma. J Clin Oncol 2010;28(8): Koscielniak E, Harms D, Henze G, et al. Results of treatment for soft tissue sarcoma in childhood and adolescence: a final report of the German Cooperative Soft Tissue Sarcoma Study CWS-86. J Clin Oncol 1999;17(12): Shah GD, Kesari S, Xu R, et al. Comparison of linear and volumetric criteria in assessing tumor response in adult high-grade gliomas. Neuro-oncol 2006;8(1): Green S, Weiss GR. Southwest Oncology Group standard response criteria, endpoint definitions and toxicity criteria. Invest New Drugs 1992;10(4): Hawkins D. Study protocol ARST 0531: randomized study of vincristine, dactinomycin and cyclophosphamide (VAC) versus VAC alternating with vincristine and irinotecan (VI) for patients with intermediate-risk rhabdomyosarcoma (RMS), NCT ?id=arst+0531&rank= Kao S. Evaluation criteria. In: Study protocol ARST 0431: intensive multi-agent therapy, including dose-compressed cycles of ifosfamide/etoposide (IE) and vincristine/ doxorubicin/cyclophosphamide (VDC) for patients with high-risk rhabdomyosarcoma, T ?id=arst+0431&rank= Malempati S. Study protocol ARST 08P1: a pilot study to evaluate novel agents (temozolomide and cixutumumab [IMC-A12, Anti-IGF-IR monoclonal antibody, IND #100947, NSC #742460) in combination with intensive multi-agent interval compressed therapy for patients with high-risk rhabdomyosarcoma, P1&rank= Baum SH, Frühwald M, Rahbar K, Wessling J, Schober O, Weckesser M. Contribution of PET/CT to prediction of outcome in children and young adults with rhabdomyosarcoma. J Nucl Med 2011;52(10): Eugene T, Corradini N, Carlier T, Dupas B, Leux C, Bodet-Milin C. 18 F-FDG-PET/CT in initial staging and assessment of early response to chemotherapy of pediatric rhabdomyosarcomas. Nucl Med Commun 2012; 33(10): radiology.rsna.org n Radiology: Volume 269: Number 3 December 2013