Choosing which approach is appropriate

Although, on balance, IPD reviews are likely to be more costly and time-consuming, the difference in resource requirements for different types of well-conducted systematic reviews may not be as great as is often thought. It is, of course, not always necessary to

Table 11.3 Factors that may influence the systematic review approach. Modified and reproduced with permission from [29]

When IPD may be beneficial

When IPD may not be beneficial

Poor reporting of trials. Information inadequate,

Detailed and clear and unbiased reporting of

selective or ambiguous

trials (CONSORT quality)

Long-term outcomes

Short-term outcomes

Time-to-event outcome measures

Binary outcome measures

Multivariate or other complex analyses

Univariate or simple analyses

Differently defined outcome measures

Outcome measures defined uniformly across


Subgroup analyses of patient-level characteristics

Patient subgroups not important


IPD available for high proportion of trials/individuals

IPD available for only a limited number of trials

High proportion of individuals excluded from Reported analyses include almost all individuals analyses

High proportion of individuals excluded from Reported analyses include almost all individuals analyses go to the lengths of collecting IPD. For example, if we are interested in binary outcomes that are likely to occur relatively quickly, where all relevant trials are published and data presented in a comprehensive and compatible way, then the most straightforward of meta-analyses based on data presented in trial publications, is probably all that is required. However, such an extreme is likely to be rare and usually unpublished trials will need to be assessed and trialists contacted to provide at least some additional summary data.

At the outset of any systematic review, the methodological factors likely to influence the results in that particular situation should be considered together with resource and time constraints so that an active decision can be taken about which approach to adopt. Table 11.3 gives some of the factors likely to influence this decision. The proportion of data available, in each of the formats, may also be a major factor in deciding which approach to take. For example, there would be little point collecting IPD if it were only available for very few trials, nor would it be useful to extract data from publications if only a very small proportion presented usable data.

0 0

Post a comment