[en] In this note, maximum-likelihood estimation or 'direct analysis' is proposed for joint analysis of local control data from tumours treated in air or under clamped conditions. The multi-fraction versions of the linear-quadratic (LQ) and multi-target (MT) models are both fitted to a series of fractionated irradiations in a C3H mouse mammary carcinoma. Goodness-of-fit test statistics showed that whereas the deviations from the LQ model were statistically insignificant (p>·25), the MT model could be rejected for this data set (p approx = 0.005). A simulation study comparing direct analysis with a two-step analysis based on TCD₅₀ values showed a greater robustness of the direct approach, especially in the case of data from non-optimal experimental designs. (author)