Professional Documents
Culture Documents
A Free Program For Calculating EUD-based NTCP and TCP in External Beam Radiotherapy
A Free Program For Calculating EUD-based NTCP and TCP in External Beam Radiotherapy
A Free Program For Calculating EUD-based NTCP and TCP in External Beam Radiotherapy
discussions, stats, and author profiles for this publication at: https://www.researchgate.net/publication/6014108
CITATIONS READS
97 514
2 authors, including:
Andrzej Niemierko
Partners HealthCare
253 PUBLICATIONS 5,709 CITATIONS
SEE PROFILE
All in-text references underlined in blue are linked to publications on ResearchGate, Available from: Andrzej Niemierko
letting you access and read them immediately. Retrieved on: 04 May 2016
Physica Medica (2007) 23, 115e125
available at www.sciencedirect.com
TECHNICAL NOTE
a
Department of Radiation Oncology, The Brody School of Medicine at East Carolina University,
600 Moye Boulevard, Greenville, NC 27834, USA
b
Department of Radiation Oncology, Massachusetts General Hospital, Boston, MA, USA
KEYWORDS Abstract Purpose: Provide a simple research tool that may be used to calculate the NCTP or
Normal tissue TCP of a particular treatment plan. Illustrate the implementation of the EUD-based NTCP and
complication TCP models as a research tool.
probability; Methods and materials: A high-level computing language was chosen to implement Niemier-
Tumor control ko’s EUD-based NTCP and TCP mathematical models. The necessary treatment planning
probability; software requirements were clearly defined.
Equivalent uniform Results: The computer code is presented and explained. Six simple examples were created to
dose; quickly troubleshoot the reader’s code implementation. A table of model parameters based on
Doseevolume the Emami data was generated.
histogram; ª 2007 Published by Elsevier Ltd on behalf of Associazione Italiana di Fisica Medica.
EUD;
NTCP;
TCP;
DVH;
Software;
Program;
Radiotherapy;
Radiation
Introduction
1120-1797/$ - see front matter ª 2007 Published by Elsevier Ltd on behalf of Associazione Italiana di Fisica Medica.
doi:10.1016/j.ejmp.2007.07.001
116 H.A. Gay, A. Niemierko
some of these models often alienates clinicians from this that can be used for both tumors and normal tissues, where
area of research. Various NTCP mathematical models based a is a unitless model parameter that is specific to the nor-
on retrospective data that correlate with clinical outcome mal structure or tumor of interest, and vi is unitless and
exist for specific organs at risk like the rectum [17] and represents the i’th partial volume receiving dose Di in Gy.
lung [18]. Many of these models have been used primarily Since the relative volume of the whole structure of interest
as research tools. Unfortunately, the clinical validation corresponds to 1, the sum of all partial volumes vi will equal
and clinical use of most of these mathematical models 1. For normal tissues, the EUD represents the uniform dose
has been disappointing. which leads to the same probability of injury as the exam-
Over the past decade, various software packages ined inhomogeneous dose distribution. To illustrate the EUD
implementing other NTCP and TCP models have been definition, imagine that you have just calculated the brain’s
created such as BIOPLAN written in Visual Basic [4], a series EUD for a particular radiotherapy plan. By looking at the
of Matlab modules by Warkentin et al. [1], and more whole brain volume curve, the solid line in Fig. 1, you could
recently the doseeresponse explorer system (DREES) also determine the corresponding NTCP. On the other hand, for
based on Matlab and freely distributed via the web tumors the EUD represents the uniform dose which leads to
(http://radium.wustl.edu/drees) [19]. the same probability of local control as the actual nonuni-
Our goal is to provide a simple research tool that may be form dose distribution. The Di and vi data pairs are obtained
used to calculate the NCTP or TCP of a particular treatment from the differential, not cumulative, doseevolume histo-
plan, and ignite further interest in this research area. gram from a given radiotherapy plan.
The choice of parameter a will determine the behavior of
Methods and materials the EUD-based model. For example, as a increases to a large
positive number, the EUD approaches the maximal dose; as
a decreases to a large negative number, the EUD approaches
Introduction to the EUD-based NTCP and TCP
the minimal dose; if a equals 1, the EUD becomes the dose
mathematical models average; and if a equals 0, the EUD equals the geometric
mean [21].
The EUD-based mathematical model [7,8] is simple be- The local control of a tumor will likely depend on the
cause it is based mainly on 2 equations, and versatile volume that received the minimum dose, since this is where
because the same model may be used for both TCP and the tumor clonogen survival should be highest. Conse-
NTCP calculations. The model has an excellent ability in quently, the EUD for tumors will be close to the minimal
fitting, for example, the Emami et al. normal tissue toler- dose, and the parameter a should be a large negative
ance values [20] (See Fig. 1). The original definition of the number.
EUD was derived on the basis of a mechanistic formulation In normal tissues with a serial or ‘‘links in a chain’’
using a linearequadratic cell survival model [21]. Subse- architecture like the spinal cord, breaking one of the links
quently, Niemierko has suggested a phenomenological will likely rupture the functional tissue chain. It is the
model of the form: volume or link that receives the maximum dose which will
!1 most likely be responsible for breaking the chain. There-
X! " a
EUDZ a
vi D i ð1Þ fore, the EUD for tissues with a serial architecture will be
iZ1 close to the maximal dose, and parameter a will usually
be a large positive number.
For normal tissues that exhibit a large volume effect
(e.g., liver, parotids, and lungs), the doseeresponse may
be closer to the average dose [21e24]. Consequently, the
EUD for these tissues will be close to the average dose,
and parameter a should be a small and positive number.
In summary, the EUD model can mimic the dosee
response behavior of the volume of interest through
parameter a.
To calculate the EUD-based normal tissue complication
probability (NTCP), Niemierko proposed parametrization of
the doseeresponse characteristics using the logistic func-
tion [8]:
1
NTCPZ # $4g50 ð2Þ
TD50
1þ
EUD
result in the same DVH and consequently in the same DVH- degree of dose inhomogeneity using standard doses
based NTCP or TCP estimates. Nevertheless, at present per fraction. However, if treatment plans do not
there is no solid clinical evidence that these different meet these criteria, as in the case of hypofractionated
spatial distributions with identical DVHs correspond to lung radiosurgery using large doses per fraction the
different clinical outcomes. dose bins should be modified using the linearequadratic
Optimization of intensity-modulated radiotherapy model to calculate the BED in standard fractions prior
plans can be based on the EUD-based models, and has to calculating the objective function. Otherwise, the
been described in detail by Wu et al. [25]. Eqs. (2) and complication or tumor control results could be grossly
(3) in this paper can be used as the IMRT optimization underestimated.
objective functions for the organs at risk and tumors, A simpler and attractive use of a modified version of this
respectively. Optimized IMRT plans using EUD-based software is the dose per fraction optimization of a given
models tend to have a greater degree of dose inhomoge- treatment plan. This mandates the use of the lineare
neity within the tumor target requiring additional dose quadratic model and prospectively obtained a/b ratios for
constraints to minimize this effect [25]. Of note, Wu both tumors and normal tissue. Once the minimum TCP
et al. did not obtain the corresponding BED in 1.8 or 2 and maximum NTCP are defined, the program could gener-
Gy fractions for each dose from the differential DVH prior ate a table of the number of fractions (for example, 1e40),
to calculating the result of the objective function. the dose per fraction that optimally satisfies the criteria,
This may be acceptable for treatment plans with a small and the corresponding NTCP and TCP. The physician could
then select, for example, the treatment that most closely
satisfies the NTCP and TCP criteria in the fewest fractions.
This could be of great benefit in countries with limited ex-
ternal beam radiotherapy facilities. The site best suited for
Table 2 Normal tissue parameters based on the Emami testing this concept would be breast, since alpha/beta
et al. normal tissue tolerance when radiation is given in ratios based on prospective data are available for both
1.8e2 Gy fractions tumor control [26] and various normal tissue late adverse
effects [27].
Structure Endpoint a g50 TD50 One area where validated NTCP and TCP models could
(Gy) be extremely useful is lung hypofractionated lung radio-
Brain Necrosis 5 3 60 surgery. The large doses per fraction employed challenge
Brainstem Necrosis 7 3 65 our current normal tissue complication and tumor control
Optic chiasm Blindness 25 3 65 knowledge based on standard fractions. Clinicians can no
Colon Obstruction/perforation 6 4 55 longer rely on their acquired clinical intuition based on
Ear (mid/ext) Acute serous otitis 31 3 40 standard fractions. Validated models could assist the
Ear (mid/ext) Chronic serous otitis 31 4 65 clinician in choosing safer and more effective treatment
Esophagus Perforation 19 4 68 plans.
Heart Pericarditis 3 3 50
Kidney Nephritis 1 3 28
Lens Cataract 3 1 18
Liver Liver failure 3 3 40 Appendix 1
Lung Pneumonitis 1 2 24.5 Matlab program code
Optic nerve Blindness 25 3 65
Retina Blindness 15 2 65 The sample EUD model parameters included in the program
code are based on Refs. 20, 24, 28e33.
Free program for calculating EUD-based NTCP and TCP 119
120 H.A. Gay, A. Niemierko
Free program for calculating EUD-based NTCP and TCP 121
122 H.A. Gay, A. Niemierko
Free program for calculating EUD-based NTCP and TCP 123
124 H.A. Gay, A. Niemierko
tumour excision: long-term results of a randomised trial. analysis to predict radiation hepatitis. Int J Radiat Oncol
Lancet Oncol 2006;7(6):467e71. Biol Phys 1992;23(4):781e8.
[27] Yarnold J, Ashton A, Bliss J, Homewood J, Harper C, Hanson J, [31] Kwa SL, Theuws JC, Wagenaar A, Damen EM, Boersma LJ,
et al. Fractionation sensitivity and dose response of late ad- Baas P, et al. Evaluation of two doseevolume histogram re-
verse effects in the breast after radiotherapy for early breast duction models for the prediction of radiation pneumonitis.
cancer: long-term results of a randomised trial. Radiother Radiother Oncol 1998;48(1):61e9.
Oncol 2005;75(1):9e17. [32] Chao KS, Deasy JO, Markman J, Haynie J, Perez CA, Purdy JA,
[28] Brenner DJ. Dose, volume, and tumor-control predictions in et al. A prospective study of salivary function sparing in pa-
radiotherapy. Int J Radiat Oncol Biol Phys 1993;26(1):171e9. tients with head-and-neck cancers receiving intensity-modu-
[29] Dawson LA, Normolle D, Balter JM, McGinn CJ, Lawrence TS, lated or three-dimensional radiation therapy: initial results.
Ten Haken RK. Analysis of radiation-induced liver disease Int J Radiat Oncol Biol Phys 2001;49(4):907e16.
using the Lyman NTCP model. Int J Radiat Oncol Biol Phys [33] Powers BE, Thames HD, Gillette SM, Smith C, Beck ER,
2002;53(4):810e21. Gillette EL. Volume effects in the irradiated canine spinal
[30] Lawrence TS, Ten Haken RK, Kessler ML, Robertson JM, cord: do they exist when the probability of injury is low?
Lyman JT, Lavigne ML, et al. The use of 3-D dose volume Radiother Oncol 1998;46(3):297e306.