- Email: [email protected]
Implementation of a model for estimating normal tissue complication probability: D0erivation of model parameters from clinical data
240
Radiation
Oncology,
Biology, Physics
Volume 27, Supplement
1
183 VALIDATION AND CLINICAL USEFULNESS OF AN OBJECTIVE PLAN EVALUATION MODEL IN THE THREE-DIMENSIONAL TREATMENT OF NON-SMALL CELL LUNG CANCER Mary V. ~rahamt, Nilesh L. Jait@, Michael G. Kahn2.3, Robert E. Drzymalaf. Michael A. Mackeyt, James A. Purdyl *Radiation Oncology Center, Mallinckrodt Institute of Radiology; *Division of Medical Informatics, Department of Internal Medicine; %epartment of Computer Science; Washington University School of Medicine, St. Louis MO 63110 Purpose/Objective: Evaluation of 3D radiotherapy plans is difficult because it requires the review of vast amounts of data. Selecting the optimal plan from a set of competing plans involves making trade-offs among the doses delivered to the target volumes and normal tissues. Decision theory, which provides a framework to specify and use these trade-offs, was used to develop an objective plan evaluation model. The purpose of this study was to validate the model and test its clinical usefulness in treatment planning. Materials & Methods: Each treatment plan consists of several issues such as non-eradication of tumor and radiation-induced damage to normal tissues. Afigure ofmerir (FOM) is computed for each plan by combining the numerical value (uriliry in decision-theoretic terms) for each issue. The utility is computed from the probability of occurrence of the issue and a physician-specific weight indicating its clinical relevance. The probabilities are computed using existing Tumor Control Probability (TCP) and Normal Tissue Complication Probability (NTCP) models. The FOM is used to rank the competing plans for a patient, and the utility is used to identify issues that need to be improved. Twenty patients with inoperable non-small cell lung cancer were treated with definitive radiotherapy using full 3D techniques for treatment design and implementation. For each patient, the evaluator (the treating radiation oncologist) initially ranked three plans using room view dose surface displays and dose volume histograms, and identified the issues which needed to be improved. The three plans were then ranked by the model. At this stage, the probabilities and weights of the issues were reviewed and changed if deemed necessary, producing a final improved ranking. Based on all these, the evaluator was asked to provide a final ranking of the three plans and the issues which needed to be improved in each plan. These were compared with the final recommendations of the evaluation model. The issues identified in the best treatment plan were then used to further direct the optimization of this plan. The FOM and individual issue urilicy value.s for the best plans were compared across patients. Results: For the twenty patients in the study, the final plan ranking produced by the evaluation model agreed with the ranking provided by the evaluator. The model was also able to correctly identify the issues which needed improvement in each plan. Subsequent replanning confirmed that further plan optimization could be achieved. A comparison of the modified FOM and uriliry values is beginning to have absolute meaning. Early experience suggests that one may correlate them with clinical outcomes such as pneumonitis to give us FOM and utilify values which can be used as thresholds for rejecting a tentative plan during treatment planning. Conclusion: The ranking of issues and the FOM were useful in achieving improved plans. Automated algorithms to generate and optimize radiotherapy plans are needed and are being developed. Any such algorithm needs a way of comparing different tentative plans. Clinically validated objective evaluation models such as the one we have developed can be used for this purpose. Objective evaluation models can also be used in the process of manually optimizing plans as they suggest the direction of plan improvement. Supported in part by: American Cancer Society Clinical Oncology Career Development Award, NLM Training Grant 5-T-15-LM07049, NLM Grant 5-R29-LM05387, NC1 Contract NOl-CM-97564, and US Btosciences Industrial Grant.
184 IMPLEMENTATION OF A MODEL FOR ESTIMATING NORMAL TISSUE COMPLICATION PROBABILITY: DERIVATION OF MODEL PARAMETERS FROM CLINICAL DATA. Andrzej Niemierko, Ph.D.; Michael Goitein, Ph.D.
Department of Radiation Oncology, Massachusetts General Hospital, Boston, MA 02114 Purpose: A model for calculating normal tissue complication probability (NTCP) has been developed. A method for deriving the model parameters from clinically available data will be presented and examples of using it to estimate late effects for the kidney and the spinal cord will be discussed. Materials & Methods: The model for estimating NTCP for inhomogeneously irradiated organ is based on
the linear-quadratic model, the concept of functional subunits (FSU), and the assumption that the NTCP is fully determined by the fraction of surviving FSU’s composing an organ, and the functional reserve of an organ (i.e., the minimal fraction of surviving FSU’s necessary to maintain functionality of the organ). Based on these assumptions it will be shown that the model parameters, that is, surviving fraction at 2Gy, size of an FSU (i.e., the number of clonogens per FSU), functional reserve of an organ, and variability in clonogen sensitivities from patient to patient can be determined from clinical data such as TD50 doses for the whole and partial irradiations, and the slope of the clinically observed dose-response relationship for a given end-point.
241
Proceedings of the 35th Annual ASTRO Meeting
Results: The method for deriving model parameters will be presented and discussed for the example of the human kidney (a “parallel” organ where all FSU’s are assumed to perform the same function in parallel and the output of the organ is the sum of outputs of the FSU’s), with clinical nephritis as the late functional endpoint, and for the human and the rat spinal cords (“serial” organs where damage to one FSU results in a complication for the whole organ), with myelitis as the late functional end-point. A comparison of the results for the human spinal cord and the rat spinal cord will be presented. Conclusions: The NI’CP model describes some radiobiological principles and the architecture of irradiated organs. The parameters of the model can be directly and fairly simply derived from clinical data. This work was suppor(:edby NM grants CA 50628 and CA 21239.
185 COMBINING CYTOSTATIC AGENTS WITH RADIATION THERAPY STRATEGIES FOR IMPROVED LOCOREGIONAL CONTROL
FOR RAPIDLY
Daniel G. Petereit, M.D.‘, Paul M. Harari. M.D.‘, Lorenzo Contreras, Eugene W. Gemer, Ph.D.2, and Timothy J. Kinsella, M.D.~
B.S. l, Michael A. Pickart. B.S.‘,
‘Department of Human Oncology, University of Wisconsin Comprehensive Arizona Cancer Center, Tucson, AZ 95724
DIVIDING
HEAD AND NECK TUMORS:
Ajit K. Verma, Ph.D.‘.
Cancer Center, Madison, WI 53792
, 2University of
Purwse: Locoregbnal control is adversely affected as clonogens from rapidly proliferating tumors repopulate during a course of radiation therapy. The cytostatic agent a-difluoromethylornithine (DFMO) was investigated for its capacity to slow proliferation kinetics in human squamous cell carcinomas (XC) of the head and neck (H&N), with the ultimate objective of improving locoregional control in rapidly dividing tumors treated with radiation therapy. Materials Three human SCC cell lines established from primary H&N tumors were evaluated in vitro (cell culture) and in v&o (SCC tumor xenografts in athymic mice) for the capacity of DFMO to induce growth inhibition. Flow cytometry analysis of SCC tumor growth kinetics and quantitative assessment of polyamine biosynthesis inhibition were performed to verify DFMO activity. DFMO effects on in vitro SCC radiosensitivity using clonogenic survival were also studied. Results: A non-cytotoxic ewsure to DFMO (5mM x 72 hours) induced pronounced growth inhibition in all 3 SCC cell lines (7090% at 7 days), and induced a 2-3 fokf delay in volume doubling time for SCC tumor xenografts when administered orally in the drinking water (1.5%) to athymic mice. Kinetic analysis via flow cytometry confirmed that DFMO produced a lengthening of SCC cell cycle times, but did not alter in vitro radiosensitivity. Inhibition of ornithine decarboxylase (ODC) activity and depletion of endogenous polyamines (putrescine and spermidine), were confirmed in normal tissue (mouse skin) and in human SCC tumor xenografts of athymic mice receiving continuous oral DFMO. Conclusian: These data indicate that antiproliferative agents, such as DFMO, are capable of altering human SCC growth kinetics without alteringintrinsic radbsensftivity. Such kinetic modulation may therefore provide a strategy to reduce the adverse impact of turnOr cell proliferation during a radiotherapy treatment course for rapidly dividing tumors such as those in the H&N.
188 TIME-DOSE
ANALYSIS
OF
CHEMORADIOTHERAPY (work done during ASTRO
LOCAL
CONTROL
ON AN ALTERNATE
FOR
WEEK
ADVANCED
HEAD
AND
NECK
NEOPLASMS
TREATED
WITH
CONCOMITANT
SCHEDULE.
Fellowship year)
W. Wang. D. Haraf. E. Vokes. R. Mick. R. Weichselbaum University of Chicago. Purpose:
TO &yse
es~llation sties
in
Chicago,
IL
the time-dose relationship for local control of disease in patients with advanced head and neck neoplssms enrolled in two sequential phase I dose which concomitant cbemothenpy and radiotherapy were delivered on an alternateweek schedule (i.e. one week of concomitant therapy alternate with
one week of rest). hl~terl&
and Method:
wcrc wlmiitual patient8 wea
From
conwmi~ntly
L986 to 1988. sixty-five
patients (pts) were arolled
in two phase I clinical trials. In uial I. S-flumuurril
with radiation (RT) on an ahemate week schedule (39 pi).
In trial II. &plain
(CDDP)
(FU) and hydroxywea
(HU)
was added to HU and FU (26 ~a). Fifty-nine
ewbble for local control. including 31 patients who had received no prior local therapy (group A) and 2Jl patiettu who had failed plin local therapy and wexe reuated @cntp B). The median follow-up was 58 months. The ranges of dose intensity of FU. HU and CDDP were 638-2387 mg/m*heek, 1286-82.50 mg/week md 8.3-33.3 mg/mVwek, respectively. The mediandose of RT nd the median dumtinnof tbexapywere 70.2 Gy and 14 weeks for group A, and 59.7 Gy and 12 we&s for gmup 8. RT was given in 1.8-2 Gyl fraction. The biological effective dose (BED) of RT was calculated using the equation proposed by J. Fowler:
Radiation
Oncology,
Biology, Physics
Volume 27, Supplement
1
183 VALIDATION AND CLINICAL USEFULNESS OF AN OBJECTIVE PLAN EVALUATION MODEL IN THE THREE-DIMENSIONAL TREATMENT OF NON-SMALL CELL LUNG CANCER Mary V. ~rahamt, Nilesh L. Jait@, Michael G. Kahn2.3, Robert E. Drzymalaf. Michael A. Mackeyt, James A. Purdyl *Radiation Oncology Center, Mallinckrodt Institute of Radiology; *Division of Medical Informatics, Department of Internal Medicine; %epartment of Computer Science; Washington University School of Medicine, St. Louis MO 63110 Purpose/Objective: Evaluation of 3D radiotherapy plans is difficult because it requires the review of vast amounts of data. Selecting the optimal plan from a set of competing plans involves making trade-offs among the doses delivered to the target volumes and normal tissues. Decision theory, which provides a framework to specify and use these trade-offs, was used to develop an objective plan evaluation model. The purpose of this study was to validate the model and test its clinical usefulness in treatment planning. Materials & Methods: Each treatment plan consists of several issues such as non-eradication of tumor and radiation-induced damage to normal tissues. Afigure ofmerir (FOM) is computed for each plan by combining the numerical value (uriliry in decision-theoretic terms) for each issue. The utility is computed from the probability of occurrence of the issue and a physician-specific weight indicating its clinical relevance. The probabilities are computed using existing Tumor Control Probability (TCP) and Normal Tissue Complication Probability (NTCP) models. The FOM is used to rank the competing plans for a patient, and the utility is used to identify issues that need to be improved. Twenty patients with inoperable non-small cell lung cancer were treated with definitive radiotherapy using full 3D techniques for treatment design and implementation. For each patient, the evaluator (the treating radiation oncologist) initially ranked three plans using room view dose surface displays and dose volume histograms, and identified the issues which needed to be improved. The three plans were then ranked by the model. At this stage, the probabilities and weights of the issues were reviewed and changed if deemed necessary, producing a final improved ranking. Based on all these, the evaluator was asked to provide a final ranking of the three plans and the issues which needed to be improved in each plan. These were compared with the final recommendations of the evaluation model. The issues identified in the best treatment plan were then used to further direct the optimization of this plan. The FOM and individual issue urilicy value.s for the best plans were compared across patients. Results: For the twenty patients in the study, the final plan ranking produced by the evaluation model agreed with the ranking provided by the evaluator. The model was also able to correctly identify the issues which needed improvement in each plan. Subsequent replanning confirmed that further plan optimization could be achieved. A comparison of the modified FOM and uriliry values is beginning to have absolute meaning. Early experience suggests that one may correlate them with clinical outcomes such as pneumonitis to give us FOM and utilify values which can be used as thresholds for rejecting a tentative plan during treatment planning. Conclusion: The ranking of issues and the FOM were useful in achieving improved plans. Automated algorithms to generate and optimize radiotherapy plans are needed and are being developed. Any such algorithm needs a way of comparing different tentative plans. Clinically validated objective evaluation models such as the one we have developed can be used for this purpose. Objective evaluation models can also be used in the process of manually optimizing plans as they suggest the direction of plan improvement. Supported in part by: American Cancer Society Clinical Oncology Career Development Award, NLM Training Grant 5-T-15-LM07049, NLM Grant 5-R29-LM05387, NC1 Contract NOl-CM-97564, and US Btosciences Industrial Grant.
184 IMPLEMENTATION OF A MODEL FOR ESTIMATING NORMAL TISSUE COMPLICATION PROBABILITY: DERIVATION OF MODEL PARAMETERS FROM CLINICAL DATA. Andrzej Niemierko, Ph.D.; Michael Goitein, Ph.D.
Department of Radiation Oncology, Massachusetts General Hospital, Boston, MA 02114 Purpose: A model for calculating normal tissue complication probability (NTCP) has been developed. A method for deriving the model parameters from clinically available data will be presented and examples of using it to estimate late effects for the kidney and the spinal cord will be discussed. Materials & Methods: The model for estimating NTCP for inhomogeneously irradiated organ is based on
the linear-quadratic model, the concept of functional subunits (FSU), and the assumption that the NTCP is fully determined by the fraction of surviving FSU’s composing an organ, and the functional reserve of an organ (i.e., the minimal fraction of surviving FSU’s necessary to maintain functionality of the organ). Based on these assumptions it will be shown that the model parameters, that is, surviving fraction at 2Gy, size of an FSU (i.e., the number of clonogens per FSU), functional reserve of an organ, and variability in clonogen sensitivities from patient to patient can be determined from clinical data such as TD50 doses for the whole and partial irradiations, and the slope of the clinically observed dose-response relationship for a given end-point.
241
Proceedings of the 35th Annual ASTRO Meeting
Results: The method for deriving model parameters will be presented and discussed for the example of the human kidney (a “parallel” organ where all FSU’s are assumed to perform the same function in parallel and the output of the organ is the sum of outputs of the FSU’s), with clinical nephritis as the late functional endpoint, and for the human and the rat spinal cords (“serial” organs where damage to one FSU results in a complication for the whole organ), with myelitis as the late functional end-point. A comparison of the results for the human spinal cord and the rat spinal cord will be presented. Conclusions: The NI’CP model describes some radiobiological principles and the architecture of irradiated organs. The parameters of the model can be directly and fairly simply derived from clinical data. This work was suppor(:edby NM grants CA 50628 and CA 21239.
185 COMBINING CYTOSTATIC AGENTS WITH RADIATION THERAPY STRATEGIES FOR IMPROVED LOCOREGIONAL CONTROL
FOR RAPIDLY
Daniel G. Petereit, M.D.‘, Paul M. Harari. M.D.‘, Lorenzo Contreras, Eugene W. Gemer, Ph.D.2, and Timothy J. Kinsella, M.D.~
B.S. l, Michael A. Pickart. B.S.‘,
‘Department of Human Oncology, University of Wisconsin Comprehensive Arizona Cancer Center, Tucson, AZ 95724
DIVIDING
HEAD AND NECK TUMORS:
Ajit K. Verma, Ph.D.‘.
Cancer Center, Madison, WI 53792
, 2University of
Purwse: Locoregbnal control is adversely affected as clonogens from rapidly proliferating tumors repopulate during a course of radiation therapy. The cytostatic agent a-difluoromethylornithine (DFMO) was investigated for its capacity to slow proliferation kinetics in human squamous cell carcinomas (XC) of the head and neck (H&N), with the ultimate objective of improving locoregional control in rapidly dividing tumors treated with radiation therapy. Materials Three human SCC cell lines established from primary H&N tumors were evaluated in vitro (cell culture) and in v&o (SCC tumor xenografts in athymic mice) for the capacity of DFMO to induce growth inhibition. Flow cytometry analysis of SCC tumor growth kinetics and quantitative assessment of polyamine biosynthesis inhibition were performed to verify DFMO activity. DFMO effects on in vitro SCC radiosensitivity using clonogenic survival were also studied. Results: A non-cytotoxic ewsure to DFMO (5mM x 72 hours) induced pronounced growth inhibition in all 3 SCC cell lines (7090% at 7 days), and induced a 2-3 fokf delay in volume doubling time for SCC tumor xenografts when administered orally in the drinking water (1.5%) to athymic mice. Kinetic analysis via flow cytometry confirmed that DFMO produced a lengthening of SCC cell cycle times, but did not alter in vitro radiosensitivity. Inhibition of ornithine decarboxylase (ODC) activity and depletion of endogenous polyamines (putrescine and spermidine), were confirmed in normal tissue (mouse skin) and in human SCC tumor xenografts of athymic mice receiving continuous oral DFMO. Conclusian: These data indicate that antiproliferative agents, such as DFMO, are capable of altering human SCC growth kinetics without alteringintrinsic radbsensftivity. Such kinetic modulation may therefore provide a strategy to reduce the adverse impact of turnOr cell proliferation during a radiotherapy treatment course for rapidly dividing tumors such as those in the H&N.
188 TIME-DOSE
ANALYSIS
OF
CHEMORADIOTHERAPY (work done during ASTRO
LOCAL
CONTROL
ON AN ALTERNATE
FOR
WEEK
ADVANCED
HEAD
AND
NECK
NEOPLASMS
TREATED
WITH
CONCOMITANT
SCHEDULE.
Fellowship year)
W. Wang. D. Haraf. E. Vokes. R. Mick. R. Weichselbaum University of Chicago. Purpose:
TO &yse
es~llation sties
in
Chicago,
IL
the time-dose relationship for local control of disease in patients with advanced head and neck neoplssms enrolled in two sequential phase I dose which concomitant cbemothenpy and radiotherapy were delivered on an alternateweek schedule (i.e. one week of concomitant therapy alternate with
one week of rest). hl~terl&
and Method:
wcrc wlmiitual patient8 wea
From
conwmi~ntly
L986 to 1988. sixty-five
patients (pts) were arolled
in two phase I clinical trials. In uial I. S-flumuurril
with radiation (RT) on an ahemate week schedule (39 pi).
In trial II. &plain
(CDDP)
(FU) and hydroxywea
(HU)
was added to HU and FU (26 ~a). Fifty-nine
ewbble for local control. including 31 patients who had received no prior local therapy (group A) and 2Jl patiettu who had failed plin local therapy and wexe reuated @cntp B). The median follow-up was 58 months. The ranges of dose intensity of FU. HU and CDDP were 638-2387 mg/m*heek, 1286-82.50 mg/week md 8.3-33.3 mg/mVwek, respectively. The mediandose of RT nd the median dumtinnof tbexapywere 70.2 Gy and 14 weeks for group A, and 59.7 Gy and 12 we&s for gmup 8. RT was given in 1.8-2 Gyl fraction. The biological effective dose (BED) of RT was calculated using the equation proposed by J. Fowler: