It is clear from discussions in this paper that phase III testing of hyperthermia in human patients must proceed in a cautious and stepwise fashion. Because of the risks of increasing late effects, either due to direct thermal damage or thermo-radiosensitization of normal tissues, it is not prudent to proceed with such testing in sites where there is a risk of excessive normal tissue heating. The correlations between temperature and prognosis in heated tumours implies that sites and techniques should be chosen where the chance of achieving relatively uniform heating are maximized. Methods of quality assurance are of equal importance and need to be carefully designed. Even then, retrospective analyses with temperature variations used as prognostic covariates are essential. Other factors, such as tumour volume and radiotherapy dose should be carefully controlled in experimental and control groups. Finally, protocol compliance is a real problem which will cause problems in interpretation of results, especially in studies designed to look at hyperthermic time-dose fractionation.