Comparison of In Vitro and Clinical Alpha/Beta Estimates With Potential Impact of Tumor Heterogeneity

To determine if in vitro α/β estimates agree with clinical values, and if there are discrepancies, how they potentially impact clinical outcomes.

A total of 234 clinical and in vitro estimates of α/β for prostate, breast, and head and neck cancer were collected across 108 studies. Statistical tests and correlation analysis were performed to determine a pattern between clinical and experimental values. Bootstrapped Monte Carlo sampling was applied to determine the underlying distribution of α and α/β values, comparing clinical and in vitro distributions. The Poisson tumor control probability (TCP) model was applied to determine the potential differences between clinical and experimental α and α/β. For prostate cancer, logistic regression was used to fit TCP models to published patient outcome data. Applying a 2-population TCP model, lower-risk cohorts and high-risk subpopulations were used to determine the influence of tumor heterogeneity on α/β.

For prostate cancer, the median in vitro α/β estimate was 3.8 Gy and median clinical estimate was 2.3 Gy. For breast cancer, clinical and in vitro estimates were 3.9 Gy and 3.6 Gy, respectively. Head and neck clinical and in vitro doses were 9 Gy and 14 Gy, respectively. Correlation analysis only found a correlation with α/β and organ type. TCP curves for in vitro and clinical estimates demonstrated significant differences in the shapes of the curves they generated. Applying the 2-population TCP model to prostate cancer outcomes data gave better agreement with in vitro α/β estimates in the case of medium-risk patient groups at a value of about 4 Gy. The high-risk group predicted a larger and more radioresistant subpopulation, with an α/β of 2.3 Gy and a fraction of 0.44.

Despite some prominent claims, no evidence was found supporting an agreement between in vitro and in vivo α/β estimates. Statistically significant differences existed between in vitro and clinical estimates, especially in the estimation of the α parameter. α values tend to be larger for in vitro estimates. Therefore, in vitro estimates of α/β cannot directly predict clinical radiosensitivity estimates. However, tumor heterogeneity may play a significant role in the clinical environment that is not seen in laboratory settings. Accounting for heterogeneity can provide more reasonable fits to patient data, and more complex models may help to more accurately correlate in vitro measurements with clinical values.