- Email: [email protected]
Assessment of lung cancer response after nonoperative therapy: tumor diameter, bidimensional product, and volume. A serial ct scan-based study
Int. J. Radiation Oncology Biol. Phys., Vol. 51, No. 1, pp. 56 – 61, 2001 Copyright © 2001 Elsevier Science Inc. Printed in the USA. All rights reserved 0360-3016/01/$–see front matter
PII S0360-3016(01)01615-7
CLINICAL INVESTIGATION
Lung
ASSESSMENT OF LUNG CANCER RESPONSE AFTER NONOPERATIVE THERAPY: TUMOR DIAMETER, BIDIMENSIONAL PRODUCT, AND VOLUME. A SERIAL CT SCAN–BASED STUDY MARIA WERNER-WASIK, M.D., YING XIAO, PH.D., EDWARD PEQUIGNOT, M.S., WALTER J. CURRAN, M.D., AND WALTER HAUCK, PH.D. Kimmel Cancer Center of Jefferson Medical College, Philadelphia, PA Purpose: Tumor response after nonoperative lung cancer therapy is traditionally evaluated by bidimensional measurement of maximum tumor diameters. The purpose of this analysis is to investigate whether tumor largest dimension (based on RECIST [Response Evaluation Criteria In Solid Tumors]), bidimensional tumor product, and volume correlate with each other in evaluating tumors of patients with locally advanced non–small-cell lung cancer (NSCLC). In addition, the pace of locally advanced NSCLC volumetric response over time, as well as the prognostic value of tumor size, was assessed in this report with software-assisted evaluation of sequential tumor measurement. Methods and Materials: Patients with locally advanced NSCLC treated with thoracic radiotherapy (RT) with or without chemotherapy were included, if the following were available: a pretreatment computed tomography (CT) simulation and at least two follow-up diagnostic thoracic CT scans taken at our institution after 1996 that were available in Dicom format for electronic transfer of images from diagnostic radiology to a computer terminal with commercial statistics software (AcQsim/CMS Focus). Primary lung tumor and grossly involved lymph nodes were contoured manually on pre-RT axial images and on all follow-up CT scans. Tumor/lymph node largest dimensions, bidimensional products (BP), and volumes were measured using the same software. Data were presented as percent change in volume or unidimensional and bidimensional measurements, with the CT simulation measurements serving as baseline. Results: A total of 22 patients were evaluated. The median thoracic RT dose was 62.4 Gy (range: 50.0 – 69.6), and all patients had a Karnofsky performance status >80. Chemotherapy (mostly carboplatin/paclitaxel) was given to 17 patients. Nineteen patients had Stage III NSCLC; 1 patient was in Stage I, 1 was in Stage IV, and 1 was recurrent. A total of 107 thoracic CT scans (22 pretreatment and 85 follow-up), averaging 4.9 scans per patient, were analyzed. Tumors reached the smallest volume at a median of 11.0 months from RT completion in all patients, 8.5 months in patients who subsequently failed locally (n ⴝ 8), and 11.9 months in those who did not fail locally. Failure rates were as follows: in-field, 36% (8/22); intrathoracic (lung nodules, effusion, pleura), 55% (12/22); and distant, 50% (11/22). Eleven patients are still alive, 4 free of disease. Overall median survival time (MST) is 27.3 months. The median initial tumor volume was 88.0 cc (range: 3.8 –218) for all patients; median BP was 33.0 cm2 (range: 3.1–112.1), and median tumor largest dimension was 7.6 cm (range: 2.2–13.5). The MST of patients with initial tumor volume <63.0 cc (n ⴝ 9) was >53.0 months and of those with tumor volume >63.0 cc was 17.3 months. The MST of patients (n ⴝ 6) with initial bidimensional tumor product <16 cm2 was >53.0 months and of those with tumor product >16 cm2 was 17.3 months. The MST of patients with largest initial dimension <4 cm was >53.1 months and of those with largest dimension >4 cm was 25.0 months. At 24 months, 79% of patients with a tumor volume <124.0 cc (n ⴝ 18) had locally controlled tumors, vs. 0% of patients with tumor volumes >124.0 cc. At the same time point, 93% of patients with BP <40 cm2 were locally controlled, vs. 0% of those with BP >40 cm2; 100% of patients with tumor dimensions <7.5 cm were locally controlled, vs. 40% of those with dimensions >7.5 cm. The partial responses in our series (assessed as the best response obtained during observation period) were as follows: 4 patients assessed based on either dimension only, product only, or volume only; 15 partial responses based on dimension or product; 16 partial responses based on volume alone; 3 cases of no tumor response, based on dimension or product; and 2 cases based on tumor volume alone. That represents good to excellent agreement among all three methods of measurement. Conclusions: (1) The response of locally advanced NSCLC to nonoperative therapy is a slow process, with tumor volumes reaching their nadir several months after treatment. (2) Smaller initial tumor size, as measured by largest tumor dimension, bidimensional product, or tumor volume, is associated with better local control and survival than larger initial measurements. (3) Any of the three tumor measurements (largest dimension,
Reprint requests to: Maria Werner-Wasik, M.D., Department of Radiation Oncology, Kimmel Cancer Center of Jefferson Medical College, 111 South 11th Street, Philadelphia, PA 19107. Tel: (215) 955-7679; E-mail: [email protected] Presented at the RSNA Annual Meeting, Chicago, IL, Novem-
ber 26 –December 30, 2000. Acknowledgment—We gratefully acknowledge the expert technical assistance of Michael Albert. Received Dec 11, 2000, and in revised form Mar 28, 2001. Accepted for publication Apr 17, 2001. 56
Volumetric tumor response in lung cancer
●
M. WERNER-WASIK et al.
57
bidimensional product, or volume) can be used as a reliable tool in assessing lung cancer response to nonoperative therapy. This confirms further the validity of RECIST and does not suggest that tumor volume is significantly superior for response evaluation. © 2001 Elsevier Science Inc. Lung cancer, Tumor volume, Tumor response, Radiation therapy.
INTRODUCTION The recently proposed RECIST (Response Evaluation Criteria In Solid Tumors) (1) raises the question whether a simple unidimensional tumor measurement is equivalent to the more complicated bidimensional measurements with regard to tumor response. Tumor response after nonoperative lung cancer therapy is currently evaluated by bidimensional measurement of maximum tumor diameters on thoracic computed tomography (CT) scans, based on the World Health Organization’s criteria (2). The optimal time interval between completion of therapy and performance of such measurements is not well defined. To our knowledge, no longitudinal study on the pace of lung cancer response after radiotherapy (RT) has been performed. In addition, any potential factors affecting such a response have not been well studied. The converse question is whether bidimensional measurements alone are precise enough to measure tumor response, or whether tumor volume may be a more sensitive tool in doing so. For spherical tumors, tumor product (area) and tumor volume are fully correlated. Lung tumors, however, are highly irregular in shape, and therefore we hypothesize that measurement of tumor volume provides a more precise assessment of response than measurement of bidimensional product.
METHODS AND MATERIALS Patient population Patients with locally advanced non–small-cell lung cancer (NSCLC) who underwent a CT simulation in preparation for thoracic RT at the Department of Radiation Oncology at Thomas Jefferson University Hospital between 1997 and 1999 were identified. The patients were eligible for the analysis if they underwent a CT simulation before RT, had a long follow-up with 2–9 post-treatment thoracic CT scans, all performed at Thomas Jefferson University Hospital, and if their follow-up images were transferable electronically from the Department of Radiology to the Kimmel Cancer Center computer, which has a commercial contouring software. All patients received thoracic RT to a median dose of 62.4 Gy (range: 50.0 – 69.6). Thoracic RT dose had to be at least 50.0 Gy with treatment breaks not exceeding 2 weeks for the patient to be eligible for analysis. Elective nodal irradiation was delivered to the immediately adjacent mediastinum, but the supraclavicular areas were not irradiated electively. No special maneuvers were performed to account for respiratory motion, but patients underwent fluoroscopy to visualize tumor motion and allow wider margins during RT delivery.
Initial tumor measurements The outline of the primary lung tumor and grossly involved hilar/mediastinal lymph nodes was contoured manually on each transaxial image derived from the commercial planning CT simulator. Intravenous contrast was used for both planning CT scans and follow-up scans. Involved lymph nodes were defined as those measuring ⱖ1.5 cm in the largest dimension. Soft-tissue windows were used for defining mediastinal masses, and lung windows were used for lesions bordering/surrounded by lung parenchyma. Involved supraclavicular lymph nodes were omitted from evaluation because of the difficulty in defining their outlines. The largest tumor and lymph node dimension, as well as dimension perpendicular to it, were measured on transaxial CT images using statistics software available in the commercial CT simulator (AcQsim, Picker International). Bidimensional tumor product was calculated as a product of the largest dimension and the dimension perpendicular to it. For each patient, a sum (or products) of the largest dimensions of the primary tumor and all involved lymph nodes was recorded as the initial largest dimension (or bidimensional tumor product). Tumor volume was also measured using statistics software in the CT simulator. Initial tumor volume was defined as the sum of the primary tumor volume and all involved lymph nodes. Tumor measurements after therapy Post-therapy CT scan images of the chest/upper abdomen were transferred electronically from the Department of Radiology at Thomas Jefferson University Hospital to a computer terminal at the Department of Radiation Oncology at Kimmel Cancer Center of Jefferson Medical College. Again, tumor/lymph node outlines were contoured on CT slices with a commercially available contouring/planning software (Focus/CMS). Tumor largest dimensions, perpendicular dimensions, and volumes were measured using the statistics package of the same software. Bidimensional tumor product (BP), the product of its largest perpendicular dimensions, was calculated for each tumor. Again, follow-up tumor largest dimension/product/volume were defined as the following: sum of the largest dimensions/products/volumes of the primary tumor and all involved lymph nodes. Assessment of patient survival and local control Survival time in months, as well as follow-up time in months, was measured from the date of CT simulation performed for RT planning. Of note, patients scheduled to receive induction chemotherapy underwent their CT simulation procedure before chemotherapy. Local failure was defined as occurring within RT fields.
58
I. J. Radiation Oncology
●
Biology
●
Physics
Volume 51, Number 1, 2001
For a local failure to be scored, tumor size had to increase on follow-up scans. In addition, lung failure (within lung parenchyma, pleura, or pleural cavity), as well as distant failure (outside chest), was scored. Statistical methods Time to death and time to local failure were estimated by the product limit method of Kaplan and Meier (3); equality of medians was compared by the log-rank test (4). Computations were performed by SAS 6.12. Agreement among responses for dimension and product, dimension and volume, and product and dimension was measured by kappa (5, 6). Computations were performed by StatXact 4.0.1. To identify initial tumor dimension, product, and volume as the best cut point, with regard to its predictive value for survival and time to local failure, successive values between the minimum and maximum divided patients into small and large groups. The cut point associated with the smallest p value (log-rank test, testing equality between small and large groups) determined the best predictor. Although all three of these differences were statistically significant, they were chosen to be optimal for these data and need to be validated in other data sets. Therefore, p values are not reported.
RESULTS Patient characteristics A total of 65 patients with locally advanced NSCLC (Stage III) underwent CT simulation between January 1997 and December 1999. Twenty-two patients with NSCLC were identified as eligible for the analysis, as defined in “Methods.” Nineteen had Stage III disease, and one each had Stage I, Stage IV, and recurrent disease. Their average age was 66 years (range: 50 – 80), and 10 patients were female. All patients had Karnofsky performance status of at least 80. Seventeen patients received chemotherapy (mostly carboplatin/paclitaxel) in addition to thoracic RT. Chemotherapy was administered as induction only in 4 patients, as concurrent only in 4 patients, and as both induction and concurrent with thoracic RT in 9 patients. Overall survival and failure patterns Median survival time (MST) of all patients was 27.3 months (range: 6.2–53.8⫹). Eleven patients are still alive, 4 without evidence of disease. All failures noted during follow-up time were scored. Failure patterns were as follows: crude local failure, defined as occurring in the RT field, 36% (8/22); lung failure, 55% (12/22); and distant failure, 50% (11/22). Actuarial local control at 24 months was 60% and at 48 months was 40% (Fig. 1). Tumors reached the smallest volume at a median of 11.0 months from RT completion in all patients, a median of 8.5 months in those patients who subsequently failed locally (n ⫽ 8), and a median of 11.9 months in those who did not fail locally (Fig. 2 and Fig. 3).
Fig. 1. Local tumor control in all patients.
Response evaluation and local control by tumor volume, bidimensional product, and largest dimension A total of 107 thoracic CT scans were reviewed in 22 patients. Twenty-two scans, one per patient, were obtained directly before initiation of therapy and the remainder, during follow-up. On average, 4.9 scans per patient were evaluated. A total of 4 (18%) patients achieved a complete radiographic response at some point of their observation; 15 (68%) patients had a partial response (by product as endpoint), and 3 (14%) had no response. Median initial tumor volume was 88.0 cc (range: 3.8 –218.1). While trying to identify an initial tumor volume as best cut point with regard to value in predicting local control, a range of tumor volumes between 63.0 and 125.0 cc seemed predictive, with significant p values. We decided to choose a volume associated with the lowest p value, namely 124.0 cc. Similarly, for BP the best cut point seemed to be 40 cm2, and for largest tumor dimension, 7.5 cm.
Fig. 2. Tumor volume response in locally controlled patients.
Volumetric tumor response in lung cancer
●
M. WERNER-WASIK et al.
59
Table 1. Best tumor response by type of measurement
Largest dimension Bidimensional product Volume Average % of all patients
CR
PR
NR
4 4 4 18%
15 15 16 69%
3 3 2 12%
Abbreviations: CR ⫽ complete response; PR ⫽ partial response; NR ⫽ no response.
Fig. 3. Tumor volume response in patients failing locally.
At 24 months, 79% of patients with tumor volumes ⱕ124.0 cc (n ⫽ 18) had locally controlled tumors, vs. 0% of patients with tumor volumes ⬎124.0 cc. Median initial BP was 33.0 cm2 (range: 3.1–112.1). At 24 months, 93% of patients with BP ⱕ40 cm2 were locally controlled, vs. 0% of those with BP ⬎40 cm2. Median initial largest dimension was 7.6 cm (range: 2.2–13.5). At 24 months, 100% of patients with dimensions ⱕ7.5 cm were locally controlled, vs. 40% of those with dimensions ⬎7.5 cm. The effect of initial tumor volume, initial tumor bidimensional product, and initial largest dimension on patient survival While trying to identify an initial tumor volume as best cut point with regard to its predictive value for patient survival, a range of tumor volumes between 30 cc and 90 cc seemed predictive, with significant p values. We decided to choose a volume associated with the lowest p value, namely 63.0 cc. The MST of patients whose initial tumor volume was ⱕ63.0 cc (n ⫽ 9) was ⬎53.0 months and of those with tumor volume ⬎63.0 cc, 17.3 months. Similarly, the MST of patients (n ⫽ 6) with an initial tumor product of ⱕ16 cm2 was ⬎53.0 months and of those with tumor product ⬎16 cm2, 17.3 months. The MST of patients with an initial largest dimension of ⱕ4 cm was ⬎53.1 months, and of those with a largest dimension of ⬎4 cm was 25.0 months. Agreement among initial tumor measurements by different methods RECIST (2) defines partial response as a decrease of at least 50% in BP, at least 30% in largest dimension, or at least 65% in tumor volume. The partial responses in our series were achieved in 15 patients as measured by largest dimension, in 15 patients as measured by bidimensional product, and in 16 patients as measured by tumor volume (Table 1). Complete responses were achieved in 4 patients
by each of the three measurements. No responses were seen in 3 patients as measured by largest dimension, in 3 patients by bidimensional product, and in 2 patients by volume. That represents good to excellent agreement among all three methods of measurement. Kappa (5, 6) for agreement between volumetric response and BP response was 0.776 (95% CI 0.357–1.0, substantial agreement)(Fig. 4)). It was 0.776 between volumetric response and dimension response (95% CI 0.357–1.0, substantial agreement)(Fig. 5), and it was 1.0 between BP response and dimension response (perfect agreement, data not shown). DISCUSSION The AJCC staging system for lung cancer (7) does not incorporate size into its tumor descriptions, except for the T1 subset, which specifies a tumor of ⱕ3 cm in largest diameter. A T2, T3, or T4 tumor can be of any size, as long as it invades the requisite structures in the chest. Statistics reported for locoregional control in the thorax after definitive RT vary widely, depending on how sophisticated the assessment of the response is and when it takes place. When posterior and lateral chest radiographs are used (RTOG 73-01) (8), and a cross-section of the tumor or the pulmonary shadow is recorded, a complete response occurs in
Fig. 4. Relationship between changes in bidimensional product and tumor volume.
60
I. J. Radiation Oncology
●
Biology
●
Physics
Fig. 5. Relationship between changes in tumor diameter and tumor volume.
24% of patients treated with 60.0 Gy, and a partial response occurs in 32%. Another 35% of patients have stable disease, and only 9% have progressive disease. These two-dimensional data, however, may not reflect the true volumetric responses. A more rigorous definition of complete response is absence of tumor documented by clinical, radiographic, and bronchoscopic examinations and endoscopic biopsy (9, 10). When response is evaluated by these criteria, only 16 –20% of patients have a complete response; 15% have a partial response; 16 –20%, stable disease; and 45–53%, progressive disease at 3 months after the completion of radiation therapy. If this evaluation is repeated every 6 months, local control at 3 years is only 7– 8%. The assessment of how tumors respond to treatment has traditionally been performed by measuring the tumor’s two dimensions (the longest diameter and its perpendicular diameter) in vivo or on radiographic images. The partial response has been defined as 50% or greater decrease in the sum of the products of diameters of all measured lesions persisting for a minimum of 4 weeks. An objective system of evaluating tumor response is mandatory to compare the effectiveness of various treatment modalities. The less prone such a system is to human error, the more valuable it is in clinical trials and daily oncologic practice. It seems that volumetric tumor measurement may be a more objective tool for assessing tumor response and represents an improvement over the traditional bidimensional measurements. However, the use of volume requires that a tumor be outlined on each transaxial CT/MRI slice, rather than measured on one selected slice, as for bidimensional evaluation. This can be accomplished either by manual “digitizing” of tumor margin by physician or, more conveniently, by automatic selection of contrast-enhancing tumor regions on each slice by the specialized software. Such software in general uses tissue segmentation techniques, and its application will
Volume 51, Number 1, 2001
probably be limited for a while to the institutions performing clinical research. A modern radiation therapy planning tool, a CT simulator, is equipped with a statistics package that allows the automatic display of tumor volume and its three largest dimensions after tumor outline has been digitized on each CT slice. However, volume is notoriously difficult to measure on post-therapy images. This difficulty is well illustrated in our study, where only a portion of all patients in our database could have their follow-up CT scans forwarded electronically to a computer with software that measures tumor volume. The newly introduced RECIST (1, 11), which relies on a single largest dimension of tumor rather than on the product of perpendicular diameters, is intended to further simplify the assessment of tumor response. RECIST is based on the assumption that “tumors are spherical and that responding patients have equivalent percentage reductions in the measures of length, width and depth of the tumor, which makes no difference in defining a partial response based on changes in largest dimension or the product of perpendicular diameters” (11). RECIST has been validated in a retrospective analysis of more than 4,000 patients treated on 14 clinical trials run by the National Cancer Institute, National Cancer Institute of Canada, Bristol-Myers Squibb, and Rhone-Poulenc Rorer. A partial tumor response is described as a 50% decrease in bidimensional product, 65% decrease in volume, and 30% decrease in tumor largest dimension (2). Separately, Watanabe et al. (12) concluded, based on evaluation of 99 patients with NSCLC, that unidimensional measurements may be sufficient for evaluating tumor response to chemotherapy, even though bidimensional measurements reflected tumor volume changes better than unidimensional measurement. Our data indicate that all three tumor size evaluation criteria—largest dimension, product of perpendicular dimensions, and volume—are approximately equally accurate when evaluating tumor response to either combined chemotherapy and radiotherapy or radiotherapy alone. That is a very important observation, because product of perpendicular diameters was in use as a surrogate for tumor volume, the presumed “truest” measure of tumor burden. Therefore, demonstrating that all three tumor measurements fare equally well as tools that evaluate the response to treatment relieves us from pursuing a cumbersome volumetric assessment as a “gold standard.” Time to “best” tumor response may likely vary, depending on which treatment modality is used, chemotherapy alone or radiation therapy alone. The pace of tumor response may differ yet again for combined modality approaches. RECIST does not specify how often patients should be imaged when receiving radiotherapy and recommends evaluation after only “every other cycle” of chemotherapy while on study and at “twice-longer intervals” after the end of treatment (2). Knowledge of the average time it takes for the irradiated tumor to shrink maximally after radiotherapy is crucial to the design of clinical trials using
Volumetric tumor response in lung cancer
radiotherapy. It allows for ordering of the imaging studies for the evaluation of “best” tumor response at the appropriate time and justifies repeated imaging of patients after therapy, a practice challenged by some. Our data demonstrate that the time required for the lung tumor to respond maximally after thoracic radiotherapy with or without chemotherapy is quite prolonged, on average 5–11 months. We also found that only smaller tumors can be controlled longterm with good probability after a combined modality regimen. In addition, larger tumor volumes seemed to be predictive of shorter median survival time than smaller volumes. It would be essential to ask this question again in a larger group of patients. A similar observation with regard to prognostic significance of tumor volume in lung cancer has been recently presented by the group from Wu¨rzburg, Germany (13). A total of 784 CT scans of 136 patients who received thoracic radiation therapy for lung cancer were evaluated with regard to pre- and post-therapeutic tumor volume. For tumors of more than 100 cc, the 1-year local control rate did not exceed 42%, irrespective of the RT dose applied (30 –74 Gy). Only 2/48 tumors of more than 100 cc were controlled
●
M. WERNER-WASIK et al.
61
more than 2 years after treatment. Tumor volume did not retain prognostic significance with regard to survival in multivariate analysis, which may possibly be explained by separately evaluating primary tumors and involved lymph nodes. In contrast, in limited-stage small-cell lung cancer, Japanese investigators established a tumor volume of less than 100 cc to be associated in a multivariate analysis of 77 patients with extended survival (31 months) when compared with volumes of more than 100 cc (survival 12 months) (12). It is interesting to note that the cutoff point of a “more favorable” initial tumor volume of lung cancer seems to be in the vicinity of 100 cc, which roughly corresponds to the largest tumor dimension, 6 –7 cm. Assessment of response in irradiated tissue is sometimes fraught with difficulty, mostly due to the treatment-related fibrosis obscuring measured tumor and to displacement of tumor and normal structures caused by scarring, etc. In the lung, anatomic considerations such as pleural effusion or atelectasis may make measurements difficult, if not sometimes impossible. It is hoped that the advent of functional imaging, such as PET scanning, will help resolve those issues.
REFERENCES 1. 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. 2. Miller AB, Hoogstraten B, Staquet M, et al. Reporting results of cancer treatment. Cancer 1981;47:207–214. 3. Kaplan EL, Meier P. Nonparametric estimation from incomplete observations. J Am Stat Assoc 1958;53:457– 481. 4. Peto R, Pike MC. Design and analysis of randomized clinical trials requiring prolonged observation of each patient. II. Analysis and examples. Br J Cancer 1977;35:1–39. 5. Agresti A. Categorical data analysis. New York: John Wiley & Sons; 1990. 6. Landis JR, Koch GG. The measurement of observer agreement for categorical data. Biometrics 1977;33:159 –174. 7. Fleming ID, Cooper JS, Henson DE, et al. AJCC cancer staging manual. 5th edition. Philadelphia: Lippincott-Raven Publishers; 1997. p. 135–137. 8. Perez CA, Stanley K, Rubin P, et al. A prospective randomized study of various irradiation doses and fractionation schedules in the treatment in inoperable non-oat-cell carcinoma of the lung. Cancer 1980;45:2744 –2753.
9. Le Chevalier T, Arriagada R, Quoix E, et al. Radiotherapy alone versus combined chemotherapy and radiotherapy in nonresectable non–small-cell lung cancer: First analysis of a randomized trial in 353 patients. J Natl Cancer Inst 1991;83: 417– 423. 10. Le Chevalier T, Arriagada R, Tarayre M, et al. Significant effect of adjuvant chemotherapy on survival in locally advanced non-small-cell lung carcinoma. J Natl Cancer Inst 1992;84:58. 11. Gehan EA, Tefft MC. Will there be resistance to the RECIST (Response Evaluation Criteria in Solid Tumors)? J Natl Cancer Inst 2000;92:179 –181. 12. Watanabe H, Sekine I, Sawada M, et al. Measuring response of small cell lung cancer to chemotherapy: Comparison of unidimensional and bidimensional measurements. Proc ASCO 2000;19:498a. 13. Willner J, Caragiani E, Tschlammer A, et al. The influence of total dose and tumor volume on local control and survival in non-small cell lung cancer. Int J Radiat Oncol Biol Phys 2000;48:327.
PII S0360-3016(01)01615-7
CLINICAL INVESTIGATION
Lung
ASSESSMENT OF LUNG CANCER RESPONSE AFTER NONOPERATIVE THERAPY: TUMOR DIAMETER, BIDIMENSIONAL PRODUCT, AND VOLUME. A SERIAL CT SCAN–BASED STUDY MARIA WERNER-WASIK, M.D., YING XIAO, PH.D., EDWARD PEQUIGNOT, M.S., WALTER J. CURRAN, M.D., AND WALTER HAUCK, PH.D. Kimmel Cancer Center of Jefferson Medical College, Philadelphia, PA Purpose: Tumor response after nonoperative lung cancer therapy is traditionally evaluated by bidimensional measurement of maximum tumor diameters. The purpose of this analysis is to investigate whether tumor largest dimension (based on RECIST [Response Evaluation Criteria In Solid Tumors]), bidimensional tumor product, and volume correlate with each other in evaluating tumors of patients with locally advanced non–small-cell lung cancer (NSCLC). In addition, the pace of locally advanced NSCLC volumetric response over time, as well as the prognostic value of tumor size, was assessed in this report with software-assisted evaluation of sequential tumor measurement. Methods and Materials: Patients with locally advanced NSCLC treated with thoracic radiotherapy (RT) with or without chemotherapy were included, if the following were available: a pretreatment computed tomography (CT) simulation and at least two follow-up diagnostic thoracic CT scans taken at our institution after 1996 that were available in Dicom format for electronic transfer of images from diagnostic radiology to a computer terminal with commercial statistics software (AcQsim/CMS Focus). Primary lung tumor and grossly involved lymph nodes were contoured manually on pre-RT axial images and on all follow-up CT scans. Tumor/lymph node largest dimensions, bidimensional products (BP), and volumes were measured using the same software. Data were presented as percent change in volume or unidimensional and bidimensional measurements, with the CT simulation measurements serving as baseline. Results: A total of 22 patients were evaluated. The median thoracic RT dose was 62.4 Gy (range: 50.0 – 69.6), and all patients had a Karnofsky performance status >80. Chemotherapy (mostly carboplatin/paclitaxel) was given to 17 patients. Nineteen patients had Stage III NSCLC; 1 patient was in Stage I, 1 was in Stage IV, and 1 was recurrent. A total of 107 thoracic CT scans (22 pretreatment and 85 follow-up), averaging 4.9 scans per patient, were analyzed. Tumors reached the smallest volume at a median of 11.0 months from RT completion in all patients, 8.5 months in patients who subsequently failed locally (n ⴝ 8), and 11.9 months in those who did not fail locally. Failure rates were as follows: in-field, 36% (8/22); intrathoracic (lung nodules, effusion, pleura), 55% (12/22); and distant, 50% (11/22). Eleven patients are still alive, 4 free of disease. Overall median survival time (MST) is 27.3 months. The median initial tumor volume was 88.0 cc (range: 3.8 –218) for all patients; median BP was 33.0 cm2 (range: 3.1–112.1), and median tumor largest dimension was 7.6 cm (range: 2.2–13.5). The MST of patients with initial tumor volume <63.0 cc (n ⴝ 9) was >53.0 months and of those with tumor volume >63.0 cc was 17.3 months. The MST of patients (n ⴝ 6) with initial bidimensional tumor product <16 cm2 was >53.0 months and of those with tumor product >16 cm2 was 17.3 months. The MST of patients with largest initial dimension <4 cm was >53.1 months and of those with largest dimension >4 cm was 25.0 months. At 24 months, 79% of patients with a tumor volume <124.0 cc (n ⴝ 18) had locally controlled tumors, vs. 0% of patients with tumor volumes >124.0 cc. At the same time point, 93% of patients with BP <40 cm2 were locally controlled, vs. 0% of those with BP >40 cm2; 100% of patients with tumor dimensions <7.5 cm were locally controlled, vs. 40% of those with dimensions >7.5 cm. The partial responses in our series (assessed as the best response obtained during observation period) were as follows: 4 patients assessed based on either dimension only, product only, or volume only; 15 partial responses based on dimension or product; 16 partial responses based on volume alone; 3 cases of no tumor response, based on dimension or product; and 2 cases based on tumor volume alone. That represents good to excellent agreement among all three methods of measurement. Conclusions: (1) The response of locally advanced NSCLC to nonoperative therapy is a slow process, with tumor volumes reaching their nadir several months after treatment. (2) Smaller initial tumor size, as measured by largest tumor dimension, bidimensional product, or tumor volume, is associated with better local control and survival than larger initial measurements. (3) Any of the three tumor measurements (largest dimension,
Reprint requests to: Maria Werner-Wasik, M.D., Department of Radiation Oncology, Kimmel Cancer Center of Jefferson Medical College, 111 South 11th Street, Philadelphia, PA 19107. Tel: (215) 955-7679; E-mail: [email protected] Presented at the RSNA Annual Meeting, Chicago, IL, Novem-
ber 26 –December 30, 2000. Acknowledgment—We gratefully acknowledge the expert technical assistance of Michael Albert. Received Dec 11, 2000, and in revised form Mar 28, 2001. Accepted for publication Apr 17, 2001. 56
Volumetric tumor response in lung cancer
●
M. WERNER-WASIK et al.
57
bidimensional product, or volume) can be used as a reliable tool in assessing lung cancer response to nonoperative therapy. This confirms further the validity of RECIST and does not suggest that tumor volume is significantly superior for response evaluation. © 2001 Elsevier Science Inc. Lung cancer, Tumor volume, Tumor response, Radiation therapy.
INTRODUCTION The recently proposed RECIST (Response Evaluation Criteria In Solid Tumors) (1) raises the question whether a simple unidimensional tumor measurement is equivalent to the more complicated bidimensional measurements with regard to tumor response. Tumor response after nonoperative lung cancer therapy is currently evaluated by bidimensional measurement of maximum tumor diameters on thoracic computed tomography (CT) scans, based on the World Health Organization’s criteria (2). The optimal time interval between completion of therapy and performance of such measurements is not well defined. To our knowledge, no longitudinal study on the pace of lung cancer response after radiotherapy (RT) has been performed. In addition, any potential factors affecting such a response have not been well studied. The converse question is whether bidimensional measurements alone are precise enough to measure tumor response, or whether tumor volume may be a more sensitive tool in doing so. For spherical tumors, tumor product (area) and tumor volume are fully correlated. Lung tumors, however, are highly irregular in shape, and therefore we hypothesize that measurement of tumor volume provides a more precise assessment of response than measurement of bidimensional product.
METHODS AND MATERIALS Patient population Patients with locally advanced non–small-cell lung cancer (NSCLC) who underwent a CT simulation in preparation for thoracic RT at the Department of Radiation Oncology at Thomas Jefferson University Hospital between 1997 and 1999 were identified. The patients were eligible for the analysis if they underwent a CT simulation before RT, had a long follow-up with 2–9 post-treatment thoracic CT scans, all performed at Thomas Jefferson University Hospital, and if their follow-up images were transferable electronically from the Department of Radiology to the Kimmel Cancer Center computer, which has a commercial contouring software. All patients received thoracic RT to a median dose of 62.4 Gy (range: 50.0 – 69.6). Thoracic RT dose had to be at least 50.0 Gy with treatment breaks not exceeding 2 weeks for the patient to be eligible for analysis. Elective nodal irradiation was delivered to the immediately adjacent mediastinum, but the supraclavicular areas were not irradiated electively. No special maneuvers were performed to account for respiratory motion, but patients underwent fluoroscopy to visualize tumor motion and allow wider margins during RT delivery.
Initial tumor measurements The outline of the primary lung tumor and grossly involved hilar/mediastinal lymph nodes was contoured manually on each transaxial image derived from the commercial planning CT simulator. Intravenous contrast was used for both planning CT scans and follow-up scans. Involved lymph nodes were defined as those measuring ⱖ1.5 cm in the largest dimension. Soft-tissue windows were used for defining mediastinal masses, and lung windows were used for lesions bordering/surrounded by lung parenchyma. Involved supraclavicular lymph nodes were omitted from evaluation because of the difficulty in defining their outlines. The largest tumor and lymph node dimension, as well as dimension perpendicular to it, were measured on transaxial CT images using statistics software available in the commercial CT simulator (AcQsim, Picker International). Bidimensional tumor product was calculated as a product of the largest dimension and the dimension perpendicular to it. For each patient, a sum (or products) of the largest dimensions of the primary tumor and all involved lymph nodes was recorded as the initial largest dimension (or bidimensional tumor product). Tumor volume was also measured using statistics software in the CT simulator. Initial tumor volume was defined as the sum of the primary tumor volume and all involved lymph nodes. Tumor measurements after therapy Post-therapy CT scan images of the chest/upper abdomen were transferred electronically from the Department of Radiology at Thomas Jefferson University Hospital to a computer terminal at the Department of Radiation Oncology at Kimmel Cancer Center of Jefferson Medical College. Again, tumor/lymph node outlines were contoured on CT slices with a commercially available contouring/planning software (Focus/CMS). Tumor largest dimensions, perpendicular dimensions, and volumes were measured using the statistics package of the same software. Bidimensional tumor product (BP), the product of its largest perpendicular dimensions, was calculated for each tumor. Again, follow-up tumor largest dimension/product/volume were defined as the following: sum of the largest dimensions/products/volumes of the primary tumor and all involved lymph nodes. Assessment of patient survival and local control Survival time in months, as well as follow-up time in months, was measured from the date of CT simulation performed for RT planning. Of note, patients scheduled to receive induction chemotherapy underwent their CT simulation procedure before chemotherapy. Local failure was defined as occurring within RT fields.
58
I. J. Radiation Oncology
●
Biology
●
Physics
Volume 51, Number 1, 2001
For a local failure to be scored, tumor size had to increase on follow-up scans. In addition, lung failure (within lung parenchyma, pleura, or pleural cavity), as well as distant failure (outside chest), was scored. Statistical methods Time to death and time to local failure were estimated by the product limit method of Kaplan and Meier (3); equality of medians was compared by the log-rank test (4). Computations were performed by SAS 6.12. Agreement among responses for dimension and product, dimension and volume, and product and dimension was measured by kappa (5, 6). Computations were performed by StatXact 4.0.1. To identify initial tumor dimension, product, and volume as the best cut point, with regard to its predictive value for survival and time to local failure, successive values between the minimum and maximum divided patients into small and large groups. The cut point associated with the smallest p value (log-rank test, testing equality between small and large groups) determined the best predictor. Although all three of these differences were statistically significant, they were chosen to be optimal for these data and need to be validated in other data sets. Therefore, p values are not reported.
RESULTS Patient characteristics A total of 65 patients with locally advanced NSCLC (Stage III) underwent CT simulation between January 1997 and December 1999. Twenty-two patients with NSCLC were identified as eligible for the analysis, as defined in “Methods.” Nineteen had Stage III disease, and one each had Stage I, Stage IV, and recurrent disease. Their average age was 66 years (range: 50 – 80), and 10 patients were female. All patients had Karnofsky performance status of at least 80. Seventeen patients received chemotherapy (mostly carboplatin/paclitaxel) in addition to thoracic RT. Chemotherapy was administered as induction only in 4 patients, as concurrent only in 4 patients, and as both induction and concurrent with thoracic RT in 9 patients. Overall survival and failure patterns Median survival time (MST) of all patients was 27.3 months (range: 6.2–53.8⫹). Eleven patients are still alive, 4 without evidence of disease. All failures noted during follow-up time were scored. Failure patterns were as follows: crude local failure, defined as occurring in the RT field, 36% (8/22); lung failure, 55% (12/22); and distant failure, 50% (11/22). Actuarial local control at 24 months was 60% and at 48 months was 40% (Fig. 1). Tumors reached the smallest volume at a median of 11.0 months from RT completion in all patients, a median of 8.5 months in those patients who subsequently failed locally (n ⫽ 8), and a median of 11.9 months in those who did not fail locally (Fig. 2 and Fig. 3).
Fig. 1. Local tumor control in all patients.
Response evaluation and local control by tumor volume, bidimensional product, and largest dimension A total of 107 thoracic CT scans were reviewed in 22 patients. Twenty-two scans, one per patient, were obtained directly before initiation of therapy and the remainder, during follow-up. On average, 4.9 scans per patient were evaluated. A total of 4 (18%) patients achieved a complete radiographic response at some point of their observation; 15 (68%) patients had a partial response (by product as endpoint), and 3 (14%) had no response. Median initial tumor volume was 88.0 cc (range: 3.8 –218.1). While trying to identify an initial tumor volume as best cut point with regard to value in predicting local control, a range of tumor volumes between 63.0 and 125.0 cc seemed predictive, with significant p values. We decided to choose a volume associated with the lowest p value, namely 124.0 cc. Similarly, for BP the best cut point seemed to be 40 cm2, and for largest tumor dimension, 7.5 cm.
Fig. 2. Tumor volume response in locally controlled patients.
Volumetric tumor response in lung cancer
●
M. WERNER-WASIK et al.
59
Table 1. Best tumor response by type of measurement
Largest dimension Bidimensional product Volume Average % of all patients
CR
PR
NR
4 4 4 18%
15 15 16 69%
3 3 2 12%
Abbreviations: CR ⫽ complete response; PR ⫽ partial response; NR ⫽ no response.
Fig. 3. Tumor volume response in patients failing locally.
At 24 months, 79% of patients with tumor volumes ⱕ124.0 cc (n ⫽ 18) had locally controlled tumors, vs. 0% of patients with tumor volumes ⬎124.0 cc. Median initial BP was 33.0 cm2 (range: 3.1–112.1). At 24 months, 93% of patients with BP ⱕ40 cm2 were locally controlled, vs. 0% of those with BP ⬎40 cm2. Median initial largest dimension was 7.6 cm (range: 2.2–13.5). At 24 months, 100% of patients with dimensions ⱕ7.5 cm were locally controlled, vs. 40% of those with dimensions ⬎7.5 cm. The effect of initial tumor volume, initial tumor bidimensional product, and initial largest dimension on patient survival While trying to identify an initial tumor volume as best cut point with regard to its predictive value for patient survival, a range of tumor volumes between 30 cc and 90 cc seemed predictive, with significant p values. We decided to choose a volume associated with the lowest p value, namely 63.0 cc. The MST of patients whose initial tumor volume was ⱕ63.0 cc (n ⫽ 9) was ⬎53.0 months and of those with tumor volume ⬎63.0 cc, 17.3 months. Similarly, the MST of patients (n ⫽ 6) with an initial tumor product of ⱕ16 cm2 was ⬎53.0 months and of those with tumor product ⬎16 cm2, 17.3 months. The MST of patients with an initial largest dimension of ⱕ4 cm was ⬎53.1 months, and of those with a largest dimension of ⬎4 cm was 25.0 months. Agreement among initial tumor measurements by different methods RECIST (2) defines partial response as a decrease of at least 50% in BP, at least 30% in largest dimension, or at least 65% in tumor volume. The partial responses in our series were achieved in 15 patients as measured by largest dimension, in 15 patients as measured by bidimensional product, and in 16 patients as measured by tumor volume (Table 1). Complete responses were achieved in 4 patients
by each of the three measurements. No responses were seen in 3 patients as measured by largest dimension, in 3 patients by bidimensional product, and in 2 patients by volume. That represents good to excellent agreement among all three methods of measurement. Kappa (5, 6) for agreement between volumetric response and BP response was 0.776 (95% CI 0.357–1.0, substantial agreement)(Fig. 4)). It was 0.776 between volumetric response and dimension response (95% CI 0.357–1.0, substantial agreement)(Fig. 5), and it was 1.0 between BP response and dimension response (perfect agreement, data not shown). DISCUSSION The AJCC staging system for lung cancer (7) does not incorporate size into its tumor descriptions, except for the T1 subset, which specifies a tumor of ⱕ3 cm in largest diameter. A T2, T3, or T4 tumor can be of any size, as long as it invades the requisite structures in the chest. Statistics reported for locoregional control in the thorax after definitive RT vary widely, depending on how sophisticated the assessment of the response is and when it takes place. When posterior and lateral chest radiographs are used (RTOG 73-01) (8), and a cross-section of the tumor or the pulmonary shadow is recorded, a complete response occurs in
Fig. 4. Relationship between changes in bidimensional product and tumor volume.
60
I. J. Radiation Oncology
●
Biology
●
Physics
Fig. 5. Relationship between changes in tumor diameter and tumor volume.
24% of patients treated with 60.0 Gy, and a partial response occurs in 32%. Another 35% of patients have stable disease, and only 9% have progressive disease. These two-dimensional data, however, may not reflect the true volumetric responses. A more rigorous definition of complete response is absence of tumor documented by clinical, radiographic, and bronchoscopic examinations and endoscopic biopsy (9, 10). When response is evaluated by these criteria, only 16 –20% of patients have a complete response; 15% have a partial response; 16 –20%, stable disease; and 45–53%, progressive disease at 3 months after the completion of radiation therapy. If this evaluation is repeated every 6 months, local control at 3 years is only 7– 8%. The assessment of how tumors respond to treatment has traditionally been performed by measuring the tumor’s two dimensions (the longest diameter and its perpendicular diameter) in vivo or on radiographic images. The partial response has been defined as 50% or greater decrease in the sum of the products of diameters of all measured lesions persisting for a minimum of 4 weeks. An objective system of evaluating tumor response is mandatory to compare the effectiveness of various treatment modalities. The less prone such a system is to human error, the more valuable it is in clinical trials and daily oncologic practice. It seems that volumetric tumor measurement may be a more objective tool for assessing tumor response and represents an improvement over the traditional bidimensional measurements. However, the use of volume requires that a tumor be outlined on each transaxial CT/MRI slice, rather than measured on one selected slice, as for bidimensional evaluation. This can be accomplished either by manual “digitizing” of tumor margin by physician or, more conveniently, by automatic selection of contrast-enhancing tumor regions on each slice by the specialized software. Such software in general uses tissue segmentation techniques, and its application will
Volume 51, Number 1, 2001
probably be limited for a while to the institutions performing clinical research. A modern radiation therapy planning tool, a CT simulator, is equipped with a statistics package that allows the automatic display of tumor volume and its three largest dimensions after tumor outline has been digitized on each CT slice. However, volume is notoriously difficult to measure on post-therapy images. This difficulty is well illustrated in our study, where only a portion of all patients in our database could have their follow-up CT scans forwarded electronically to a computer with software that measures tumor volume. The newly introduced RECIST (1, 11), which relies on a single largest dimension of tumor rather than on the product of perpendicular diameters, is intended to further simplify the assessment of tumor response. RECIST is based on the assumption that “tumors are spherical and that responding patients have equivalent percentage reductions in the measures of length, width and depth of the tumor, which makes no difference in defining a partial response based on changes in largest dimension or the product of perpendicular diameters” (11). RECIST has been validated in a retrospective analysis of more than 4,000 patients treated on 14 clinical trials run by the National Cancer Institute, National Cancer Institute of Canada, Bristol-Myers Squibb, and Rhone-Poulenc Rorer. A partial tumor response is described as a 50% decrease in bidimensional product, 65% decrease in volume, and 30% decrease in tumor largest dimension (2). Separately, Watanabe et al. (12) concluded, based on evaluation of 99 patients with NSCLC, that unidimensional measurements may be sufficient for evaluating tumor response to chemotherapy, even though bidimensional measurements reflected tumor volume changes better than unidimensional measurement. Our data indicate that all three tumor size evaluation criteria—largest dimension, product of perpendicular dimensions, and volume—are approximately equally accurate when evaluating tumor response to either combined chemotherapy and radiotherapy or radiotherapy alone. That is a very important observation, because product of perpendicular diameters was in use as a surrogate for tumor volume, the presumed “truest” measure of tumor burden. Therefore, demonstrating that all three tumor measurements fare equally well as tools that evaluate the response to treatment relieves us from pursuing a cumbersome volumetric assessment as a “gold standard.” Time to “best” tumor response may likely vary, depending on which treatment modality is used, chemotherapy alone or radiation therapy alone. The pace of tumor response may differ yet again for combined modality approaches. RECIST does not specify how often patients should be imaged when receiving radiotherapy and recommends evaluation after only “every other cycle” of chemotherapy while on study and at “twice-longer intervals” after the end of treatment (2). Knowledge of the average time it takes for the irradiated tumor to shrink maximally after radiotherapy is crucial to the design of clinical trials using
Volumetric tumor response in lung cancer
radiotherapy. It allows for ordering of the imaging studies for the evaluation of “best” tumor response at the appropriate time and justifies repeated imaging of patients after therapy, a practice challenged by some. Our data demonstrate that the time required for the lung tumor to respond maximally after thoracic radiotherapy with or without chemotherapy is quite prolonged, on average 5–11 months. We also found that only smaller tumors can be controlled longterm with good probability after a combined modality regimen. In addition, larger tumor volumes seemed to be predictive of shorter median survival time than smaller volumes. It would be essential to ask this question again in a larger group of patients. A similar observation with regard to prognostic significance of tumor volume in lung cancer has been recently presented by the group from Wu¨rzburg, Germany (13). A total of 784 CT scans of 136 patients who received thoracic radiation therapy for lung cancer were evaluated with regard to pre- and post-therapeutic tumor volume. For tumors of more than 100 cc, the 1-year local control rate did not exceed 42%, irrespective of the RT dose applied (30 –74 Gy). Only 2/48 tumors of more than 100 cc were controlled
●
M. WERNER-WASIK et al.
61
more than 2 years after treatment. Tumor volume did not retain prognostic significance with regard to survival in multivariate analysis, which may possibly be explained by separately evaluating primary tumors and involved lymph nodes. In contrast, in limited-stage small-cell lung cancer, Japanese investigators established a tumor volume of less than 100 cc to be associated in a multivariate analysis of 77 patients with extended survival (31 months) when compared with volumes of more than 100 cc (survival 12 months) (12). It is interesting to note that the cutoff point of a “more favorable” initial tumor volume of lung cancer seems to be in the vicinity of 100 cc, which roughly corresponds to the largest tumor dimension, 6 –7 cm. Assessment of response in irradiated tissue is sometimes fraught with difficulty, mostly due to the treatment-related fibrosis obscuring measured tumor and to displacement of tumor and normal structures caused by scarring, etc. In the lung, anatomic considerations such as pleural effusion or atelectasis may make measurements difficult, if not sometimes impossible. It is hoped that the advent of functional imaging, such as PET scanning, will help resolve those issues.
REFERENCES 1. 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. 2. Miller AB, Hoogstraten B, Staquet M, et al. Reporting results of cancer treatment. Cancer 1981;47:207–214. 3. Kaplan EL, Meier P. Nonparametric estimation from incomplete observations. J Am Stat Assoc 1958;53:457– 481. 4. Peto R, Pike MC. Design and analysis of randomized clinical trials requiring prolonged observation of each patient. II. Analysis and examples. Br J Cancer 1977;35:1–39. 5. Agresti A. Categorical data analysis. New York: John Wiley & Sons; 1990. 6. Landis JR, Koch GG. The measurement of observer agreement for categorical data. Biometrics 1977;33:159 –174. 7. Fleming ID, Cooper JS, Henson DE, et al. AJCC cancer staging manual. 5th edition. Philadelphia: Lippincott-Raven Publishers; 1997. p. 135–137. 8. Perez CA, Stanley K, Rubin P, et al. A prospective randomized study of various irradiation doses and fractionation schedules in the treatment in inoperable non-oat-cell carcinoma of the lung. Cancer 1980;45:2744 –2753.
9. Le Chevalier T, Arriagada R, Quoix E, et al. Radiotherapy alone versus combined chemotherapy and radiotherapy in nonresectable non–small-cell lung cancer: First analysis of a randomized trial in 353 patients. J Natl Cancer Inst 1991;83: 417– 423. 10. Le Chevalier T, Arriagada R, Tarayre M, et al. Significant effect of adjuvant chemotherapy on survival in locally advanced non-small-cell lung carcinoma. J Natl Cancer Inst 1992;84:58. 11. Gehan EA, Tefft MC. Will there be resistance to the RECIST (Response Evaluation Criteria in Solid Tumors)? J Natl Cancer Inst 2000;92:179 –181. 12. Watanabe H, Sekine I, Sawada M, et al. Measuring response of small cell lung cancer to chemotherapy: Comparison of unidimensional and bidimensional measurements. Proc ASCO 2000;19:498a. 13. Willner J, Caragiani E, Tschlammer A, et al. The influence of total dose and tumor volume on local control and survival in non-small cell lung cancer. Int J Radiat Oncol Biol Phys 2000;48:327.