877 / 1082
S861
ESTRO 36 2017
_______________________________________________________________________________________________
Physics, Aarhus C, Denmark
3
Heidelberg University Hospital, Department of
Radiooncology, Heidelberg, Germany
Purpose or Objective
Dose planning constraints, such as the volume receiving
xGy (V
x
), are often extracted from clinical outcome data.
These data are tainted by uncertainties in dose- and
output recording, large patient heterogeneity, small
sample size and -variability. Our study is dedicated to the
investigation of the fundamental uncertainty with which
dose planning constraints can be extracted from clinical
radiation pneumonitis data and how this relates to patient
number and complication incidence rate.
Material and Methods
In order to measure the reliability of a V
x
logistic
regression model, the dose-response mechanism
generating the complication events needs to be known.
For this reason, we generated cohorts of patients using
real-life dose distributions of patients treated for
advanced lung cancer, combined with a postulated V
x
logistic dose-response model. In each of the 1000 cohorts,
the patients were randomly assigned complication/no-
complication based on the individual risks given by the
postulated model. Thus, “alternative reality” cohorts
comprised of the same patients, but with different
outcomes from the same dose distributions were created.
Each cohort thus represented a possible result of a clinical
study. They were analyzed with a number of logistic V
x
models, and the best fitting model was selected. This was
matched to the postulated model to determine its
recognition rate. The postulated model was varied to
produce low, intermediate and high incidence rates.
Results
For a patient cohort of 100 individuals, a postulated model
with an incidence rate of 15/100 was recognized in 31% of
the cohorts. For a cohort size of 500, the correct-
recognition rates increased to 75%. For a lower incidence
model (7/100), these recognition frequencies dropped to
20% and 56%,
respectively.
To ensure a recognition rate >90%, large cohorts of
between 500 and 2000 patients were required, see Figure
1(a). Figure 1(b) shows that the distribution width for the
15/100 incidence rate model decreased from a standard
deviation of 10Gy for 100 patients to 1Gy for 2000
patients.
Conclusion
For realistic dose distributions and cohort sizes, a state-
of-the-art analysis failed to identify the postulated dose-
response in about 2-in-3 cases for the low incidence of the
large-volume effect complication radiation pneumonitis.
Very large patient cohorts were required to ensure
recognition rates above 90%. This fundamentally low
success rate could explain the persistent difficulties to
derive dose constraints from clinical data for
complications in large-volume effect, “parallel” organs.
EP-1615 Second cancer risk after radiation of localized
prostate cancer with and without flattening filter
M. Treutwein
1
, M. Hipp
2
, R. Loeschel
3
, O. Koelbl
1
, B.
Dobler
1
1
Klinik und Poliklinik für Strahlentherapie- Unive,
Regensburg University Medical Center- Department of
Radiation Oncology, Regensburg, Germany
2
Klinkum St. Marien, Department of Radiotherapy,
Amberg, Germany
3
Ostbayerische Technische Hochschule, Faculty of
Computer Science and Mathematics, Regensburg,
Germany
Purpose or Objective
Radiotherapy is a standard treatment modality with
curative intent for localized prostate cancer. Prostate
cancer is a disease of elderly men. Nevertheless these
patients have a remaining life span of ten years or more.
Radiotherapy compared to surgery may increase the risk
for second cancer. Minimizing this risk can be one criterion
in deciding for a specific technique. Therefore we
compared the organ equivalent dose (OED) and excess
absolute risk (EAR) for second cancer for different
treatment techniques.
Material and Methods
For ten patients four different plans were calculated,
using a seven field intensity modulated radiotherapy
(IMRT) and a single arc volumetric modulated arc therapy
(VMAT) with and without flattening filter. The