| Method |
Strengths |
Limitations |
| Determining when meta‐analyses require updating (Barrowman 2003; Sutton 2006). |
a) Efficient.
b) Easy to use,.compute formula.
c) Reduced type‐I error relative to conventional cumulative meta‐analysis.
d) Test sensitivity and specificity easily modifiable. |
a) Applicability limited to meta‐analyses with statistically non‐significant results.
b) Assumes no secular trend in effect and that the variance of pooled estimate shrinks at a rate inversely proportional to the total number of participants in all studies.
c) Test results sensitive to studies' sizes . |
| Conventional cumulative meta‐analysis (Lau 1992; Lau 1995; Baum 1981; Berkey 1996). |
a) Defines the earliest time at which an intervention can be shown to be efficacious or harmful.
b) Monitors the effect size and direction over time.
c) Timing for each update is known.
d) Ascertains the contribution of individual studies to the cumulatively pooled effect estimate.
e) Allows one to explore heterogeneity and perform sensitivity analysis.
f) Provides up‐to‐date information.
g) Useful in stopping ongoing trials or planning future trials. |
a) Inefficient ‐ an update is conducted every time a new study is available.
b) Inflated type‐I error due to multiple testing.
c) Affected by publication bias. |
| Cumulative meta‐analysis using the cumulative slope as an indicator of stability (Mullen 2001). |
a) ‐ g) of "Conventional cumulative meta‐analysis" above.
h) Explores the stability of the effect size and informs the need for updating. |
a) ‐ c) of "Conventional cumulative meta‐analysis".
d) Judging extent of stability is arbitrary.
e) The variance of the 'cumulative slope' is invalid.
f) The minimum size of a meta‐analytic database for fitting a regression line whose slope would be a valid indicator of (in)stability of effect not specified. |
| Cumulative meta‐analysis using sequential monitoring boundaries (Pogue 1997). |
a) ‐ g) of "Conventional CMA" above.
h) Controls type‐I error by using sequential monitoring boundaries. |
a) and c) of "Conventional cumulative meta‐analysis".
d) Requires prior calculation of the OIS.
e) Does not account for heterogeneity or bias among studies.
f) Requires accumulation of more data compared to conventional cumulative meta‐analysis. |
| Recursive cumulative meta‐analysis
(Ioannidis 1999; Ioannidis 2001). |
a) ‐ g) of "Conventional cumulative meta‐analysis" above.
h) Incorporates results from unpublished studies and follow‐up or more detailed data for studies already included in the cumulative meta‐analysis.
i) Documents the evolution of results as missing, updated, and new data are incorporated in information steps.
j) Evaluates updated follow‐up information, publication bias or lag, and heterogeneity.
k) Treatment effect estimates are based on relatively accurate and complete data. |
a) and b) of "Conventional cumulative meta‐analysis".
c) Unpublished and updated information must be carefully studied and verified to minimize bias.
d) Analysis of updated follow‐up data may sometimes be inappropriate since many post‐study patients will cross over.
e) More costly and resource‐consuming than conventional cumulative meta‐analysis. |
