Local control of carcinoma of the tonsil by radiation therapy: An analysis of patterns of fractionation in nine institutions

Purpose: To investigate the importance to outcome of treatment, for squamous cell carcinomas of the tonsillar fossa, of dose per fraction, overall treatment duration, and total dose. Methods and Materials: A collaborative retrospective study was undertaken in nine centers that used widely different dose-fractionation patterns for external beam radiation therapy. Results: There were 676 eligible cases treated only with photon beams during the years 1976-1985. The probability of local control (of the tonsillar fossa primary) was influenced by both T-stage and N-stage. Significant treatment parameters were total dose and overall treatment duration, but not dose per fraction. Over the range of about 40 to 90% and for a constant overall treatment duration, local tumor control probability increased by nearly 2% for each 1 Gy increase in total dose. For a constant total dose there was a decrease in the probability of local control associated with prolongation of overall treatment duration, presumed to result from accelerated regrowth of surviving tumor clonongens during the course of treatment. If it is assumed that accelerated regrowth occurred at a constant rate and began within 9 days of the start of treatment, an average of 0.53 Gy extra dose per day's extension of treatment would be required to maintain a constant probability of local control. Correspondingly, the probability of local control from a constant dose would be lowered by an average of at least 1% for each day's extension of treatment duration. However, the data are slightly more consistent with an average delay of as long as 30 days before onset of accelerated repopulation, with a consequent increase to an average of 0.73 Gy per day for the value of the compensatory dose. The α/β ratio for this tumor is high enough that the effect of fraction size on the probability of local control can be ignored; a precise estimate is not possible because the best value for β was close to zero. After accounting for the significant variables studies (treatment time, T-stage, N-stage). the dose-response curves for tumor control were still shallow, suggesting that there are additional causes for heterogeneity of responses among these tumors. Conclusions: Total dose is important to treatment outcome: After accounting for other treatment variables, there is about a 2% per Gy increase in probability of tumor control over the ranges of control commonly achieved. Overall treatment duration is important: There is at least a 1% per day decrease in tumour control probability if delivery of a constant total dose is prolonged, requiring a compensatory increase in dose by 0.5-0.7 Gy per day to achieve a constant rate of tumor control. Fraction size is not, of itself, an important factor in the response of primary carcinoma of the tonsil. If a tumor has demonstrated a capacity for metastatic spread to lymph nodes, a higher total dose should be considered to achieved control rates at the primary site equivalent to those in node negative patients. Even after accounting for variables such as tumor stage, total dose, and overall treatment duration, there is sufficient heterogeneity in other undocumented determinants of tumor control to cause the tumor control probability curve to be a shallow function of dose.