Adaptation Time as a Determinant of the Dosimetric Effectiveness of Online Adaptive Radiotherapy for Bladder Cancer

Abstract

Interfraction anatomic deformations decrease the precision of radiotherapy, which can be improved by online adaptive radiation therapy (oART). However, oART takes time, allowing intrafractional deformations. In this study on focal radiotherapy for bladder cancer, we analyzed the time effect of oART on the equivalent uniform dose in the CTV (EUD_CTV) per fraction and for the accumulated dose distribution over a treatment series as measure of effectiveness. A time-dependent digital CTV model was built from deformable image registration (DIR) between pre- and post-adaptation imaging. The model was highly dose fraction-specific. Planning target volume (PTV) margins were varied by shrinking the clinical PTV to obtain the margin-specific CTV. The EUD_CTV per fraction decreased by-4.4 ± 0.9% of prescribed dose per min in treatment series with a steeper than average time dependency of EUD_CTV. The EUD_CTV for DIR-based accumulated dose distributions over a treatment series was significantly dependent on adaptation time and PTV margin (p < 0.0001, Chi2 test for each variable). Increasing adaptation times larger than 10 min by five minutes requires a 1.9 ± 0.24 mm additional margin to maintain EUD_CTV for a treatment series. Adaptation time is an important determinant of the precision of oART for one half of the bladder cancer patients, and it should be aimed at to be minimized.

Keywords: bladder cancer; bladder deformation; intrafractional motion; online adaptive radiotherapy.

Figure 1

Empirical cumulative distribution of the equivalent uniform doses normalized to the prescribed dose for the CTV with a margin of 5 mm in the nine treatment series using the adaptive plan on CBCT1 at t₁ = 0 min adaptation time (blue), on CBCT2 at the clinical adaptation time t₂ (orange) and for the model-based CTV generated for an adaptation time of 10 min (yellow).

Figure 2

Trend analysis of the nEUD values for the CTV on adaptation time for the nine treatment series. nEUD_CTV values were ranked for each series and the rank numbers were scaled to the interval [0, 1]. There was a significant inter-series heterogeneity of slopes (p = 0.0157, F-test). According to the overall mean slope, series were ranked as time-sensitive (blue regression lines), with smaller slops than the overall slopes, or time-insensitive, with larger slops (black regression lines). Ranked nEUD_CTV values from a time-insensitive series were labeled by the same black filled symbol, and values from a time-sensitive series by the same blue open symbol. The 95% confidence limits of mean-predicted values are given as transparent confidence bands.

Figure 3

Empirical cumulative distribution of the nEUD_CTV values from the adaptive plans presenting a comparison between the actual nEUD values and the nEUD values calculated for the CTV at the simulated time points using the digital bladder model. The solid lines represent the observed nEUD_CTV values with a treatment duration belonging to the terciles T1 (red line), T2 (green line) and T3 (blue line). The dashed lines indicate the nEUD values for the model-based time-dependent CTV volumes at the median adaptation times in T1 (orange), T2 (green) and T3 (blue), i.e., 14.14 min, 18.82 min and 26.69 min. The distribution of the nEUD_CTV values on CBCT1 at t = 0 min is indicated for comparison (black).

Figure 4

Residual nEUD_CTV values for the CTV volumes from a mixed linear model adapted to the nEUD values for the CTV volumes on CBCT2 after clinical adaptation times from fraction to fraction and to the interpolated CTV volumes at 4 distinct time points: 10 min, 14.14 min, 18.82 min and 26.69 min. The intrafractional dependence of the nEUD_CTV values on the adaptation time was modeled using the interpolated CTV volumes at the above 4 distinct time points. The nEUD_CTV values were explained by the respective time as a continuous covariate and the respective dose fraction as a mean effect using a repeated measures design. The interfractional dependence of the respective nEUD_CTV values used the clinically observed adaptation times and the respective patient as explanatory variables.

Figure 5

nEUD values for the CTV from the accumulated dose distribution of the adaptive plans on CBCT2 over a treatment series in dependence on PTV margin and adaptation time. Data are shown for the 5 time-sensitive treatment series. Open brown squares represent data modeled at 10 min adaptation time, green rhombus, red triangles and blue circles data at 14.1429 min, 18.877 min and 26.5866 min adaptation time. PTV margins were varied in 1 mm increments and data for the different adaptation times at the same integer PTV margin were separated by changing the horizontal position. This was carried out by varying the horizontal position around the respective PTV margin from the longest to the shortest adaptation time in submillimeter steps. Vertical box plots represent the indicated PTV margin and adaptation time extending from the 25th to the 75th percentile of values; the horizontal line inside the box represents the median of values and the whiskers indicate the whole range of values. Median nEUD_CTV values for the series at neighboring PTV margins were connected by drawn lines for each adaptation time.

Figure 6

Adaptation time dependence of the probability of an nEUD_CTV < 95% for the accumulated dose distribution of a treatment series at different PTV margins. Data from Figure 5 were analyzed with a bivariate logistic model at an nEUD_CTV cut-off value of 95%. Logistic curves were given together with their 95% confidence bands. There was a significant effect of adaptation time and PTV margin on nEUD_CTV (p < 0.0001, Chi2 test for each variable).