Session Item

Using data from an international prospective cohort study of prostate cancer (PCa) patients receiving radical radiotherapy (RT), this analysis aims to estimate α/β ratios for individual urinary toxicity (tox) endpoints: grade 2+ (G2+) urinary frequency, G2+ urinary incontinence and G1+ haematuria. Owing to a previously suggested consequential component of late bladder damage, we also considered the impact of including the total treatment time in NTCP models.

Non-metastitic PCa patients (pts) were enrolled in 8 countries (04-2014/03-2016). RT was prescribed according to local regimens, but centres used standardised data collection (including baseline symptoms). Tox was scored using CTCAE. Pts in the study had full 3D dosimetric information; we used solid bladder DVHs. Data were available for 1009 pts.

We fitted the probability of late urinary tox within 24 months with the Logit-EUD sigmoid model, explicitly including the α/β ratios and also allowing a term for the impact of total treatment time. The general expression for the NTCP model is in fig 2a. Four parameters describe the model (EUD₅₀, k, n, α/β). A fifth parameter γ was needed for RT time correction. We used the Maximum Likelihood method (Optimization Toolbox, Matlab) to obtain the best estimates of the model parameters.

We compared models including the new estimates of α/β with models with α/β=3Gy (conventional value for late tox).

The cohort included 678 conventionally fractionated pts and 331 hypofractionated pts (Fig 1a). Fig 1b shows the incidence of the tox endpoints within the different subcohorts and Fig 1c the distribution of toxicity events as a function of dose and dose/fraction. We found a consequential effect between acute and late tox for all endpoints (fig 2b).

Details on model parameters are in fig 2c. For frequency and incontinence, the late bladder α/β estimates were low (without correction for RTtime), always arriving at the lower allowed boundary, with the inability to converge to a set of parameters. We chose to stop optimisation at α/β =0.3 Gy, which already indicates an extreme effect of fraction size. When including treatment time correction, α/β was 1.4 Gy for frequency and 3 Gy for incontinence, with γ=0.9 and 0.8, respectively. Of note, the volume effect parameter n was 0.06 for incontinence and 0.54 for frequency, pointing at a serial behaviour for the bladder for incontinence and at a more parallel behaviour for frequency. Models with α/β=3Gy had a significantly worse Likelihood and inability to converge in the dose domain (estimates always arriving at the upper allowed boundary). For haematuria, our cohort was unable to discriminate the best α/β ±γ estimate.

We found a greater unexpected impact of hypofractionation on RT induced urinary tox. This could be explained by a bladder α/β<0.3Gy or, radiobiologically more plausible, by introducing a time factor likely to represent a previously hypothesised consequential component of late effect.

This work was funded by REQUITE (EU 7th Framework Programm grant 601826) and RADprecise (ERA-NET ERA PerMed grant ERAPERMED2018-244)