Should randomized clinical trials be required for proton radiotherapy

Recently, several articles have been published in the Journal of Clinical Oncology and other journals reviewing and commenting on the record of proton beam therapy, as well as an analysis including some critical commentary by Brada et al. All of these articles make the uncontested point that there are almost no randomized clinical trials (RCTs) comparing proton beam therapy with conventional x-ray therapy. We wish to address the issues of why this is, whether RCTs would be appropriate, and whether they are necessary before proton beam therapy is widely promulgated and reimbursed. In brief, the arguments for the use of protons in radiation therapy are as follows. (1) Owing primarily to their depth dose characteristics (for each proton beam, virtually no dose is administered distal to the target volume and substantially less dose is administered than x-rays proximal to the target volume), the dose distributions that can be achieved with protons are in almost all cases superior to those possible with x-rays (with or without intensity modulation, which can be achieved with both modalities). There is generally between two to three times less energy deposited by protons to the uninvolved normal tissues outside the target volume (variously described as integral dose or the dose bath) as compared with the energy that x-rays deposit. (2) There is virtually no difference in tissue response per unit dose between protons of therapeutic energies as compared with x-rays, so that the only relevant differences are physical. (3) Radiation delivered to normal tissues causes damage to them, just as it does to tumors, and the severity of that damage increases with increasing dose. Item (1) has been documented exhaustively in treatment planning studies. There is a large body of in vivo and in vitro evidence underpinning item (2). Item (3) is corroborated in countless clinical reports over many decades. These points are not contested by any of the authors cited above, nor, to our knowledge, by any critics of proton beam therapy. They are not speculations— they are demonstrated facts. It is therefore hard to imagine how any objective person could avoid the conclusion that there is, at the very least, a high probability that protons can provide superior therapy to that possible with x-rays in almost all circumstances. It is primarily for this reason that the practitioners of proton beam therapy have found it ethically unacceptable to conduct RCTs comparing protons with x-rays. Such a comparison would not meet a central requirement for performing RCTs, namely that there be equipoise between the arms of the trial. Brada et al base their opinions on what they understand to be the requirements of evidence-based medicine. In our opinion, the issue has much more to do with the implications of evidence-based medicine than it does with the clinical effectiveness of protons. In short, does evidence-based medicine require that, under all circumstances, positive RCTs are a precondition for the promulgation and reimbursement of new technologies? If it does, and if one accepted evidence-based medicine, so defined, as the sole basis for making medical decisions, then one would have no choice but to agree with the position taken by Brada et al and others. However, we find it impossible to believe that unethical clinical studies could be considered to be a prerequisite to the adoption of a medical therapy. It must surely be the case that there are circumstances under which even the most dedicated advocate of evidence-based medicine would agree that RCTs would be improper. In deciding whether the arms of a trial meet the equipoise standard, one can only rely on informed judgment. It is our argument that informed judgment leads to the conclusion that proton beam therapy is precisely such a circumstance. Advocates of RCTs are prone to state that, although there may be good arguments for the superiority of one arm, one does “know” that there is an advantage. In addition, to justify the conduct of trials that seem not to be in equipoise, they cite trials in which the outcome was the reverse of what was expected. Taking this argument to its extreme, one would have to conclude that there is effectively no clinical knowledge except that learned from RCTs. But this is an untenable position. Knowledge is not a dichotomous quality. We know things with varying levels of confidence. Even when RCTs are available, rarely do they provide all the information that is needed to care most effectively for the patient. We make informed evaluations of the level of confidence one has in any given judgment (for example, virtually all informed persons judge that the validity of the above-listed points is established with extremely high confidence), and we must base our actions on information about which we have good confidence. Brada et al make the point more than once that the apparent advantage of protons at several tumor sites may be due to patient selection bias. Indeed, selection bias is a serious issue in comparing nonrandomized studies (as are the eligibility criteria in assessing the applicability of the conclusions of an RCT to the treatment of a JOURNAL OF CLINICAL ONCOLOGY COMMENTS AND CONTROVERSIES VOLUME 26 NUMBER 2 JANUARY 1