The critical importance of a systematic review protocol for the conduct of the systematic review has been highlighted previously in this journal.1,2 Aromataris and Pearson emphasized that a true systematic review is a protocol-driven research endeavor and that the details regarding the conduct of the review, including the review objectives, questions and the selection criteria, should be clearly delineated in an a priori protocol.3 Issues related to the selection of outcomes for a review, including the subtle distinction between primary and secondary outcomes, and the critical aspect of the number of outcomes proposed for a review, and the (in)appropriateness of surrogate outcomes as primary outcomes for a review, were aptly explored in an editorial by Aromataris.4 It seems that for some reviewers the terminology used for review outcomes may be challenging and that the essence of the concept of "surrogate" outcomes may prove hard to grasp. It is fundamental that reviewers understand the differences between "surrogate" outcomes and those that are not, and select and justify the outcomes proposed for their review based on sound understanding of their review topic and reasoning. The purpose of this editorial is to touch upon some of these issues by defining "surrogate" outcomes, providing examples, and reflecting on the importance of these types of outcomes for the conduct of systematic reviews.

In clinical research, it has long been recognized that it is useful to identify outcomes that are both sensitive to the treatment effects and clinically relevant; these outcomes are called "true endpoints".5 Examples of true endpoints are survival time in cancer, bone fractures in osteoporosis, myocardial infarction and stroke in cardiovascular diseases, and survival time in HIV infection.5 It has also been realized for a similarly long time that due to various reasons, such as research costs, the extensive time required for observations, the sample size needed or measurement difficulties, it might be difficult to use and assess these true endpoints in clinical research.5 Therefore, it was proposed that true endpoints be replaced with other outcomes measured more easily or timely or for less cost when compared to the true endpoints. These "replacements" or "surrogates" for true outcomes were called "surrogate endpoints".5 Examples of surrogate endpoints are time to progress from one stage to another stage in cancer, tumor response to cancer treatment, bone mineral density in osteoporosis, heart ejection fraction, blood pressure or arrhythmias in cardiovascular diseases, and CD4 counts or viral load in HIV infection.5 When surrogate endpoints are used it is expected that they have been validated or evaluated.5 The issues involved in the evaluation or validation of surrogate endpoints, including the theoretical models or frameworks for surrogate endpoints and statistical discussions around requirements for surrogate endpoints, are not discussed here; however there is methodological literature available on the topic.5-15 For the purposes of this editorial, it is suffice to specify that a surrogate endpoint is "expected to predict clinical benefit (or harm, or lack of benefit or harm) based on epidemiologic, therapeutic, pathophysiologic or other scientific evidence".6^(p.304) The potential problems or hazards related to the use of surrogate endpoints have been debated in scientific literature and are not presented here; however, reviewers are strongly encouraged to familiarize themselves with these issues.16-19 The essence of the problems or hazards related to the use of surrogate endpoints was captured in a brilliant summary included in the conclusions of an article with the suggestive title The idolatry of the surrogate: "studies of hard endpoints [true endpoints] are necessary to practice truly patient centered medicine. In order to fully engage our patients in treatment decisions, we must understand how therapies affect outcomes that are important to them. Surrogate end points will not provide us with these answers".17^(p.3)

What is the importance of surrogate outcomes for the conduct of systematic reviews? It is important to emphasize first that it was acknowledged in scientific literature on clinical trials that "although many would like to avoid surrogate endpoints altogether, sometimes surrogates will be the only reasonable alternative, especially when the true endpoint is rare and/or distant in time".5^(p.4) Thus, in some fields of research, researchers have to use surrogate outcomes and therefore reviewers should be aware of this and present the background information and justification for the necessity of using surrogate outcomes in their review protocol and review report in a transparent way. If surrogate endpoints are used then the scientific evidence about the validation or evaluation of these surrogate endpoints should be explicitly presented and referenced in the review. Peer reviewers, editors and readers or users of the review should be able to check if the surrogate endpoints used in the review are supported or not by existing good quality methodological literature. Moreover, the reviewers should discuss in an explicit and transparent way all potential problems or hazards related to the use of specific surrogate endpoints in terms of misjudgments about the real effect of the treatments on true endpoints, implications for future research and recommendations for practice.

The hard line I propose here is that whenever possible only true endpoints should be used as outcomes for JBI systematic reviews. At the very least, the primary outcomes for the review should be by default true endpoints. I don't think it is justifiable or acceptable to use surrogate outcomes for a review when validated or evaluated true endpoints are available. Whenever surrogate endpoints are used, the reviewers should present information on existing true endpoints (if any) in their review protocol and review report, and provide a clear justification for using the surrogate endpoints rather than true endpoints.

References