Like most research endeavors, systematic reviews are all about looking for answers. When framing a PICO-based question for a quantitative review, the answers generally lie with the "O" - the outcomes. Selecting a priori appropriate indicators of the effectiveness of an intervention, in other words, what you determine in advance is going to be important to provide the answers you are looking for. This is no straightforward task and requires careful consideration on the part of review authors so as to ensure the relevance of their completed review to users.

Some systematic reviews present outcomes labelled as primary or secondary whereas others, just a straight list of pertinent outcomes, without making the distinction for the reader. Irrespective of the approach used, all of the a priori outcomes identified and listed in a systematic review protocol should be addressed in full from the relevant studies that meet the review eligibility criteria. This also includes reporting on outcomes in the review in the case where none of the included studies provide useable data for an outcome. Inconsistency in reporting of outcomes between review protocols and published reviews is relatively common.1 Changing outcomes during the review process, or reporting on a subset of outcomes or outcome measures based on the reporting of the included studies rather than the a priori protocol, may introduce risk of reporting bias and selection bias to the systematic review.1

Some of the terminology surrounding outcomes for a systematic review can create confusion. Similar terms that a review author will encounter can mean slightly different things, depending on the type of research being looked at. Use of the labels "primary" and "secondary" is an example. For a systematic review, the distinction between primary and secondary outcomes is dependent on their importance in answering the question posed and their ability to provide clear conclusions about the effects of an intervention for decision makers.2 For any review, determining the primary outcomes should begin with consideration of those that are directly relevant to the patient - those of clinical importance. A simple example may be a review investigating the effectiveness of antihypertensive agents. Relevant outcomes may include all-cause mortality, cardiovascular mortality, non-fatal cardiovascular morbidity (e.g. myocardial infarction, stroke and angina), combined cardiovascular events, and mean systolic and diastolic blood pressure (change in or endpoint). In this case, the primary clinical outcomes - those that are of relevance to the patient - are clearly those related to their own mortality and that have a direct impact on their quality of life, for example, suffering a stroke or a heart attack. Whilst blood pressure is an obvious measure as it is clearly linked to the cardiovascular complications mentioned and commonly measured in interventional studies, it may be a secondary outcome of interest to the review author. In reality, the patient doesn't care too much about their blood pressure - they just don't want to have a stroke or heart attack!! Investigators conducting and reporting clinical trials are interested in blood pressure and we'll find it measured in all sorts of relevant studies. Why? Selecting a primary outcome for a clinical trial is quite different to selecting primary outcomes for a systematic review. Unlike a systematic review, a clinical trial should only ever have one primary outcome. The primary outcome is the outcome used to determine the "power" of the study, which has a direct bearing on the number of participants recruited. For Phase 1 or 2 trials in particular, this is often a continuous measure and a surrogate endpoint - for example, blood pressure - an outcome that can be measured relatively easily and at predetermined frequencies within the available time frame and funding for the trial. For a systematic review, however, it is rare that a surrogate outcome should be amongst the primary outcomes.

Another distinction authors should be aware of is that a single outcome may have several legitimate outcome measures. Providing some indication of the acceptable measures of an identified outcome is good practice in a review protocol. Outcomes such as "satisfaction" or "quality of life" are classic examples where review authors should provide some indication of valid tools or instruments that will be included to measure these outcomes, dependent on the participant group.

There is no golden rule to the number of outcomes that should be reported on in a systematic review and this will vary with the nature and scope of each question. Most reviews of effects that appear in the JBI Database of Systematic Reviews and Implementation Reports typically list three to seven identified outcomes. To truly inform effectiveness, both benefits and harms should be reported so a decision maker can assess both sides of the coin before making a decision. To accommodate this, the adverse events or side effects of an intervention should ideally also be included amongst the review outcomes, where possible. In my capacity as Editor of the JBI Database of Systematic Reviews and Implementation Reports, I commonly see review protocols that identify only one or two outcomes in total. This immediately raises the question that I often relay to the authors of the review protocol: is there something important to the question or is the full picture of the intervention being missed? Conversely, where 15-20 outcomes are listed for one review question, 50% may appear trivial and may be related to the intervention, but which have little bearing on the question being asked by the review. For example, a review investigating the effectiveness of non-pharmacological interventions to manage cancer-related fatigue will undoubtedly identify exercise amongst the interventions investigated; however having BMI or weight as an outcome a priori for such a question would appear trivial and should not be reported, despite its obvious link with exercise.

This editorial touches on only a few of the issues review authors should consider when selecting outcomes that will answer their review question during the preparation of their protocol. I do hope you read these with interest and, of course, continue to enjoy the range of multidisciplinary reviews and reports of the use of evidence in clinical settings presented in this issue of the JBI Database of Systematic Reviews and Implementation Reports.

Director, Synthesis Science, The Joanna Briggs Institute

Editor, JBI Database of Systematic Reviews and Implementation Reports

References