Pre-calculated effects
When the paper already gives the effect and you only need to pool
Use this module when you extracted the already calculated effect from the papers, rather than the raw counts. That is the typical case for hazard ratios (HR) from survival analyses, which rarely come with enough data to recompute.
Follow the three-step flow. Here I cover the three input formats and the field that decides whether your analysis comes out right.
Step 1, the data
There are three submodules, chosen on the previous page. Pick the one that matches what you managed to extract from the papers.
| Format | Required columns | Needs a standard error? |
|---|---|---|
| TE with 95% CI | study, TE, lower, upper |
no |
| TE with standard error | study, TE, seTE |
yes |
| Full set | study, TE, seTE, lower, upper |
yes |
n.e and n.c are optional in all three formats and appear in the forest plot when present. Any other column stays available for subgroup analysis and meta-regression.
TE with 95% CI is the only one where you type no standard error at all: the app derives it from the interval. The other two ask for seTE directly.
The Data check verifies the columns, counts the studies and validates the interval, requiring lower <= TE <= upper. If you swapped two columns, it warns you here, before the analysis runs.
Step 2, the parameters
Outcome settings
Outcome name, If the outcome increases, is that beneficial or harmful?, Experimental group name and Control group name work as in the binary module.
Summary measure (sm) (default HR) offers six options, and here it does more than label the axis: it decides whether your numbers go through a log transformation.
- Ratio measures:
HR,RR,OR. Pooling happens on the log scale and the result returns to the original scale in the plots. - Difference measures:
RD,MD,SMD. No transformation.
The scale of TE
This is where mistakes happen most, and it is worth reading carefully.
For ratio measures, MetaVidence expects the value as reported in the paper, not the logarithm. If the paper says “HR 0.76 (95% CI 0.61 to 0.95)”, you type 0.76, 0.61 and 0.95. Typing the logarithm makes the app reject the spreadsheet, since it requires positive values for ratio measures.
For difference measures, the value is already on the right scale and nothing is transformed.
The sm chosen in Outcome settings decides whether your numbers go through the log transformation. Choosing MD when the data are ratios, or the other way round, produces a result that looks plausible and is wrong. It is the kind of error no validation catches, because every number is valid.
seTE goes on a different scale from TE
For ratio measures the two fields do not share a scale. TE goes in as reported, 0.76. seTE goes in on the log scale, which is how papers report it and how the calculation uses it. For the example above, seTE would be 0.113, not 0.087.
Getting this wrong raises no error at all: you get a plausible confidence interval that is wrong. The way to avoid the decision altogether is the TE with 95% CI format, which needs no seTE.
Statistical settings (advanced)
Analysis model, Tau method, Random CI method and Prediction method work exactly as in the flow.
Prediction interval is off by default (No) in all three submodules.
There is no Method field: the effect already comes calculated, so pooling is always by inverse variance.
Step 3, export and results
Every card from the binary module is here, with the same behavior: Forest plot, Leave-one-out analysis, Funnel plot, Small-study effects with Egger’s test, Subgroup analysis and Meta-regression, plus the two summary tables as XLSX.