Trial Sequential Analysis Explained: When and Why to Use It

A standard meta-analysis tells you the pooled effect size and whether it's statistically significant. What it doesn't tell you is whether you actually have enough accumulated evidence to trust that result -- or whether you're looking at a false positive that a few more trials would overturn. Trial sequential analysis (TSA) exists to answer that second question.

The problem TSA is designed to solve

Meta-analyses accumulate evidence the same way an interim analysis in a single large trial does -- studies get added over time, and each addition changes the pooled estimate. A single conventional significance test (p < 0.05) applied to a meta-analysis carries the same risk of a spurious false-positive finding that repeated interim looks at accumulating trial data would carry in a single study, if you don't adjust for it.

TSA borrows methodology from sequential monitoring in clinical trials and applies it to meta-analysis, calculating an "information size" -- essentially, how many patients' worth of data you'd need to reliably detect or rule out a clinically meaningful effect -- and adjusting your significance boundaries accordingly.

What a TSA plot actually shows you

A trial sequential analysis plot displays the cumulative Z-curve (the running significance level as each study is added chronologically) against several boundaries:

- **Conventional boundaries** (the standard p < 0.05 threshold) - **Trial sequential monitoring boundaries**, which are more conservative and account for repeated testing as evidence accumulates - **The required information size**, marking whether your accumulated sample size has actually reached the threshold needed to draw a reliable conclusion

If your cumulative Z-curve crosses the trial sequential monitoring boundary, your result is more likely to be a genuine effect rather than a chance finding from an underpowered body of evidence. If it hasn't reached the required information size and hasn't crossed a boundary, the honest conclusion is that more evidence is needed -- not that there's no effect.

When TSA adds real value to your review

TSA is most useful when:

- Your meta-analysis includes a growing number of relatively small trials - You want to test whether a "significant" pooled result is robust, or fragile enough that one more trial could flip it - Your field is prone to early, underpowered positive findings that later get contradicted by larger studies -- a well-documented pattern in several areas of clinical research

It adds less value when you already have a small number of very large, well-powered trials, since the information-size problem TSA addresses is less of a concern there.

Where reviewers get it wrong

The most common mistake is applying TSA and then only reporting the parts of the output that support a positive conclusion, while omitting the required information size when it hasn't actually been reached. A methods reviewer checking your TSA output will look specifically for whether you addressed information size honestly, not just whether your Z-curve crossed a boundary at some point.

Software like TSA (from the Copenhagen Trial Unit) handles the calculations, but the judgment calls -- what effect size counts as clinically meaningful for your outcome, what alpha-spending function to use -- still require methodological input specific to your review question.

If you're deciding whether TSA strengthens your meta-analysis, or you've run it and want a second opinion on interpreting the output before it goes into your manuscript, that's a natural extension of the statistical consulting we already provide for meta-analyses.

**[Get a quote](/systematic-reviews#meta-analysis-support)** for statistical consulting on your review.