Review authors should decide in advance whether they will collect information about all outcomes measured in a study, or about only those outcomes of (pre-specified) interest in the review. Because we recommend in Section 7.3.6 that results should only be collected for pre-specified outcomes, we also suggest that only the outcomes listed in the protocol be described in detail. However, a complete list of the names of all outcomes measured allows a more detailed assessment of the risk of bias due to selective outcome reporting (see Chapter 8, Section 8.14).
Information about outcomes that is likely to be important includes:
definition (diagnostic method, name of scale, definition of threshold, type of behaviour);
timing;
unit of measurement (if relevant); and
for scales: upper and lower limits, and whether a high or low score is favourable.
It may be useful to collect details of cited reports associated with scales, since these may contain further information about upper and lower limits, direction of benefit, typical averages and standard deviations, minimally important effect magnitudes, and information about validation.
Further considerations for economics outcomes are discussed in Chapter 15 (Section 15.4.2), and for patient-reported outcomes in Chapter 17.
Collection of adverse effect outcomes can pose particular difficulties, discussed in detail in Chapter 14. Information falling under any of the terms ‘adverse effect’, ‘adverse drug reaction‘, ‘side effect’, ‘toxic effect’, ‘adverse event’ and ‘complication’ may be considered as being potentially suitable for data extraction when evaluating the harmful effects of an intervention. Furthermore, it may be unclear whether an outcome should be classified as an adverse outcome (and the same outcome may be considered to be an adverse effect in some studies and not in others). No mention of adverse effects does not necessarily mean that no adverse effects occurred. It is usually safest to assume that they were not ascertained or not recorded. Quality of life measures are usually general measures that do not look specifically at particular adverse effects of the intervention. While quality of life scales can be used to gauge the overall well-being, they should not be regarded as substitutes for a detailed evaluation of safety and tolerability.
Precise definitions of adverse effect outcomes and their intensity should be recorded, since they may vary between studies. For example, in a review of aspirin and gastrointestinal haemorrhage, some trials simply reported gastrointestinal bleeds, while others reported specific categories of bleeding, such as haematemesis, melaena, and proctorrhagia (Derry 2000). The definition and reporting of severity of the haemorrhages (for example, major, severe, requiring hospital admission) also varied considerably among the trials (Zanchetti 1999). Moreover, a particular adverse effect may be described or measured in different ways among the studies. For example, the terms ‘tiredness’, ‘fatigue’ or ‘lethargy’ might all be used in reporting of adverse effects. Study authors may also use different thresholds for ‘abnormal’ results (for example, hypokalaemia diagnosed at a serum potassium concentration of 3.0 mmol/l or 3.5 mmol/l).