Send to

Choose Destination
Radiother Oncol. 1989 Nov;16(3):169-82.

Clinical radiobiology of malignant melanoma.

Author information

Department of Biomathematics, University of Texas M.D. Anderson Cancer Center, Houston.


Tumor-control probability (TCP) was analyzed in a series of 121 patients having 239 histologically proven recurrent or metastatic malignant melanomas. These were treated with fractionated radiotherapy with various doses per fraction, total doses, and overall times. Cutaneous lesions (127, 53%) were treated with electron beams, and more deeply seated tumors (112, 47%) with 60Co or 4-8 MV X-rays. The fraction size was highly variable, and this permitted determination of the alpha/beta ratio in the multifraction linear-quadratic model, which was estimated at 0.57 Gy with 95% confidence limits [-1.07, 2.5] Gy. Treatment time had no demonstrable influence on TCP. Thus this tumor exhibits the fractionation sensitivity characteristic of a late-responding normal tissue, suggesting that an adequate fractionation schedule for malignant melanomas would be characterized by larger-than-conventional doses per fraction, possibly about 6 Gy per fraction. This is consistent with the conclusions of other authors. Tumor size, evaluated as mean tumor diameter, S, had a major impact on TCP: the number of target cells increased as a power function of S with exponent 0.72 (95% confidence limits [0.49, 0.94]. In fact, a considerable amount of the heterogeneity in the dose-response data could be removed by accounting for size. Thus, the weak or absent dose response became highly significant. When a patient had multiple lesions, the responses of these to radiotherapy tended to be similar, thus implying that results were significantly influenced by a "hidden parameter" (such as inherent radiosensitivity or immunological status). A test of the predictive value of the TCP-model was performed in a different series of 183 cutaneous and lymph node malignant melanomas. The observed dose-response relationship in this data set was in good agreement with the model prediction. A chi-square test for goodness-of-fit showed that the variation between predicted and observed results could be explained by the binomial variation on quantal response data.

[Indexed for MEDLINE]

Supplemental Content

Full text links

Icon for Elsevier Science
Loading ...
Support Center