Challenges in radiobiological modeling: can we decide between LQ and LQ-L models based on reviewed clinical NSCLC treatment outcome data?

Alina Santiago, Steffen Barczyk, Urszula Jelen, Rita Engenhart-Cabillic, Andrea Wittig

To study the dose-response of stage I non-small-cell lung cancer (NSCLC) in terms of long-term local tumor control (LC) after conventional and hypofractionated photon radiotherapy, modeled with the linear-quadratic (LQ) and linear-quadratic-linear (LQ-L) approaches and to estimate the clinical α/β ratio within the LQ frame. We identified studies of curative radiotherapy as single treatment through MedLine search reporting 3-year LC as primary outcome of interest. Logistic models coupled with the biologically effective dose (BED) at isocenter and PTV edge according to both the LQ and LQ-L models with α/β = 10 Gy were fitted. Additionally, α/β was estimated from direct LQ fits. Thirty one studies were included reporting outcome of 2319 patients. The LQ-L fit yielded a significant value of 11.0 ± 5.2 Gy for the dose threshold (Dt) for BED10 at the isocenter. The LQ and LQ-L fits did not differ substantially. Concerning the estimation of α/β, the value obtained from the direct LQ fit for the complete fractionation range was 3.9 [68 % CI: 2.2-9.0] Gy (p > 0.05). Both LQ and LQ-L fits can model local tumor control after conventionally and hypofractionated irradiation and are robust methods for predicting clinical effects. The observed dose-effect for local control in NSCLC is weaker at high doses due to data dispersion. For BED10 values of 100-150 Gy in ≥3 fractions, the differences in isoeffects predicted by both models can be neglected.