Optimization of uncomplicated control for head and neck tumors.

ABSTRACT

Almost 200 patients have been treated for head and neck tumors at two different dose levels. Based on the clinically observed probabilities for tumor control and fatal normal tissue complications at the two dose levels, the dose giving maximum uncomplicated control has retrospectively been calculated and compared with the clinical data. A Poisson statistical model for control and complications has been used including a correlation parameter, delta, to describe the fraction of patients where control and complications are statistically independent. The clinically observed probability of uncomplicated tumor control, P+, is consistent with only a small fraction of the patients treated being statistically independent (delta = 0.2 or 20%). Customarily, 100% of the patients are assumed to be statistically independent with regard to tumor control and normal tissue complications. More precisely, the clinical data are consistent, with almost 20% of the patients being significantly more sensitive to radiation since they gain local tumor control but simultaneously suffer fatal complications. An even larger fraction of the patients (almost 30%) seemed to be more resistant to radiation, showing neither serious treatment complications nor control of the local tumor growth. It is suggested that if these patient groups could be identified by a predictive assay for the radiation sensitivity of their normal tissues and preferably also for their tumors, the uncomplicated tumor control could be increased by about 20%. This figure is based on the actuarial survival of the patients and has been corrected for the inevitable uncertainty in dose delivery. It is also pointed out that about 20% of the patients can never be saved by a predictive assay because of the considerable statistical variance associated with the Poisson process and the eradication of the last clonogenic tumor cell. Finally, note that the possible existence of radiation sensitive and resistant patient groups is consistent with known genetic deficiencies such as ataxia telangiectasia for the sensitive patients and the existence of repair efficient head and neck tumors that are unusually efficient in repairing double strand breaks. If such sensitive and resistant patient groups do exist, it should be sufficient to perform a predictive assay on normal tissues alone avoiding the often impossible task of sampling the most radiation resistant tumor cell line.