New Clinical Estimates of the α/β of Cervical Cancer Suggest a Low Value
Abstract
Purpose
Estimating Cervical cancer α/β values from patient local control data and three methods of radiobiological parameter estimation.
Methods
A retrospective analysis of 180 cervical cancer patients was conducted, using local control rates with follow-up periods of one to five years. Employing the linear quadratic model with repopulation, a heterogeneous Poisson tumor control probability (TCP) model, a logistic TCP model, and a Cox Proportional Hazards (CPH) model were fit to the patient data to determine α/β while accounting for the effects of adjuvant treatments such as chemotherapy and surgery.
Results
The heterogeneous Poisson model, when fit to the entire patient dataset predicted an α/β of 7.96 Gy. Fitting it to only the patients that didn’t receive surgery predicted a much lower α/β of 3.53 Gy. Similarly, the logistic TCP when fit to only dosimetric variables, predicted an α/β of 27.96 Gy, whereas the model accounting for parameters such as tumor stage and surgery predicted a lower estimate of 2.62. Finally, the CPH model, accounting for tumor stage, chemotherapy, and surgery gave a similarly low estimate of 3.20 Gy. All three models gave very similar estimates, averaging to 3.12 Gy.
Conclusion
When accounting for variables such as surgery, chemotherapy and tumor stage, all three methods of α/β predication reliably gave similar estimates, averaging to 3.12 Gy, much lower than the assumed 10 Gy for cervical cancer. Accounting for population heterogeneity and non-dosimetric variables such as tumor stage and chemotherapy has a significant effect on the predicted α/β. Based on these results, the α/β of cervical cancer is likely around 3 Gy, far lower than what has been previously assumed.