Information Technology Reference
In-Depth Information
META-ANALYSIS
“Meta-analysis should be viewed as an observational study of the evidence.
The steps involved are similar to any other research undertaking: formula-
tion of the problem to be addressed, collection and analysis of the data,
and reporting of the results. Researchers should write in advance a detailed
research protocol that clearly states the objectives, the hypotheses to be
tested, the subgroups of interest, and the proposed methods and criteria
for identifying and selecting relevant studies and extracting and analysing
information” (Egger, Smith, and Phillips, 1997).
30
Too many studies end with inconclusive results because of the relatively
small number of observations that were made. The researcher can't quite
reject the null hypothesis, but isn't quite ready to embrace the null
hypothesis, either. As we saw in Chapter 1, a post hoc subgroup analysis
can suggest an additional relationship, but the relationship cannot be
subject to statistical test in the absence of additional data.
Meta-analysis is a set of techniques that allow us to combine the results
of a series of small trials and observational studies. With the appropriate
meta-analysis, we can, in theory, obtain more precise estimates of main
effects, test
a priori
hypotheses about subgroups, and determine the
number of observations needed for large-scale randomized trials.
By putting together all available data, meta-analyses are also better
placed than individual trials to answer questions about whether an overall
study result varies among subgroups—for example, among men and
women, older and younger patients, or subjects with different degrees of
severity of disease.
In performing a meta-analysis, we need to distinguish between observa-
tional studies and randomized trials.
Confounding and selection bias can easily distort the findings from
observational studies. Egger et al. [1998] note, “An important criterion
supporting causality of associations is a dose-response relation. In occupa-
tional epidemiology the quest to show such an association can lead to very
different groups of employees being compared. In a meta-analysis that
examined the link between exposure to formaldehyde and cancer, funeral
directors and embalmers (high exposure) were compared with anatomists
and pathologists (intermediate to high exposure) and with industrial
workers (low to high exposure, depending on job assignment). There is a
striking deficit of deaths from lung cancer among anatomists and patholo-
gists [standardized mortality ratio 33 (95% confidence interval 22 to 47)],
which is most likely to be due to a lower prevalence of smoking among
30
Reprinted with permission from the BMJ Publishing Group.
