New technology in radiotherapy-we do not need phase iii trials-con

The computer revolution has led to an avalanche of technological innovation in radiotherapy treatment planning and delivery over the last decade. Some of these developments, such as computerised treatment planning and patient management systems, have led to greater efficiencies. Others have resulted in improved accuracy of treatment planning and delivery, for example 4D CT and in-room image guidance. The benefits whether in terms of time saved or improved ability to detect the gross tumour volume seem self evident, and it would be hard to persuade an ethics committee that a randomised trial should be performed to confirm the usefulness of the “efficiency and accuracy” technologies, although it is likely that all could be tested using randomised trial methodology. New technology may also affect patient outcomes either directly through a potentially greater cytotoxic effect or superior dose distribution, for example hadron therapy or IMRT; or indirectly through dose escalation and/or hypofractionation made possible by greater treatment precision and consequent improvement in the therapeutic ratio. The endpoints of interest may be biological (survival, local control, toxicity or quality of life) or they may be economic (cost benefit, cost utility). In the case of these endpoints, confirmation of the superiority of the new over the old requires randomised phase III trials, since it cannot be assumed that the outcomes will inevitably favour the experimental arm. For example, a meta-analysis of 57 randomised trials conducted by the RTOG revealed that innovative treatments were as likely as standard treatments to be successful (odds ratio for survival: 1.01, P=0.5). On the other hand, treatment related mortality was worse on the experimental arms (OR: 1.76, P=0.008).1 It is important to distinguish the “direct effect” technologies from the “indirect”. The former require direct comparison of the technology for example IMRT versus 3D conformal. Although the dose distributions achieved with IMRT may appear superior, there has been unanticipated normal tissue injury in organs receiving low doses. Examples are the deaths due to pneumonitis which were observed following IMRT for mesothelioma2 and helical Tomotherapy for lung cancer.3 Although these toxicities were observed in single arm studies, had they been less frequent, the impact of these iatrogenic deaths on outcome may only have been detected otherwise by randomisation. The proponents of proton beam therapy have argued that the beam's physical characteristics are by themselves sufficient to establish therapeutic superiority over x-rays, and that randomised trials are unnecessary to demonstrate the obvious.4 It is possible that protons are more effective for treating lung cancer (but we don't know since they have not been compared), but given the cost of protons they would need to be very much more effective than x-rays to provide a cost benefit. There is no other way of demonstrating this other than by unbiased randomisation with proper costings. The “indirect effect” technologies (such as hypofractionated stereotactic body radiotherapy - SBRT) can be thought of as applications of new technology (or combinations of technologies) to deliver unconventional treatment schedules which may have greater biological effect than conventional treatment. Here the primary comparison should not be between the various technologies that can be used to deliver SBRT, but between novel and standard treatment schedules. Nevertheless there is a risk that the high levels of local control observed with hypofractionated SBRT may be mistakenly attributed to the technology rather than the treatment schedule, especially within the lay community. Marketing claims of the superiority of one technology over another in terms of improved local control should be viewed with scepticism in the absence of a direct randomised comparison. The observation that outcomes were the same in patients whose brain metastases were treated stereotactically by either a linac or a Gamma Knife in a non-randomised comparison5 make it unlikely there would be any enthusiasm for undertaking a phase III trial. Finally a word of caution regarding the high dose hypofractionated schedules in popular use for the treatment of stage I lung cancer. The rates of local control are much higher than observed in historical controls treated with conventional fractionation. But because there has been no randomised comparison, we do not know if the better local control translates into an improvement in survival. The local control in the phase 2 study RTOG 0236 of SBRT in early stage lung cancer was an impressive at 90% at 3 years but survival was only 56%.6 Further, nine of 55 patients (16%) died of unknown causes. Is this an excess number of deaths, possibly due to undetected treatment related toxicity? Only by randomisation can we answer this question. In summary, computer driven technological advances have improved efficiency and treatment accuracy in lung cancer radiotherapy, and we do not need randomised trials to confirm these benefits. But when it is claimed that treatment outcomes are improved by some inherent property of the technology, then those claims should be tested scientifically by randomised studies as is the case for any other new therapeutic intervention, especially if the toxicity and monetary costs are likely to be high.