Continuous outcomes
Mean difference and standardized mean difference
Use this module when the outcome is a numeric measure: blood pressure, days in hospital, quality of life score. Follow the three-step flow; this page covers what the data has to look like and what changes in step 2.
Step 1, the data
Every study needs study, n.e and n.c. Then, for each group, whatever the paper actually reported.
Mean and standard deviation
| Column | What it is |
|---|---|
n.e, mean.e, sd.e |
size, mean and SD of the experimental group |
n.c, mean.c, sd.c |
the same for the control group |
A study reporting a median is accepted too
| Column | What it is |
|---|---|
n.e, median.e, q1.e, q3.e |
size, median and quartiles of the experimental group |
n.c, median.c, q1.c, q3.c |
the same for the control group |
min.e, max.e, min.c and max.c are optional. They can stand in for the quartiles when a paper reports only a range, and the conversion is then less precise.
The two can be mixed in one file
Fill the columns each study reports and leave the rest empty. The example dataset does exactly that: eight studies with mean and SD, four with median and quartiles. MetaVidence converts the median studies before pooling, using the method of Luo and colleagues for the mean and that of Shi and colleagues for the standard deviation, which are the current defaults of the meta package.
The Data check reports how many studies were converted, so the number is in front of you before anything runs and can go straight into the methods section.
It assumes an approximately symmetric distribution. With strongly skewed data, which is often exactly why the paper reported a median in the first place, the estimated mean may not represent the study well. Report how many studies were converted, and consider a sensitivity analysis without them.
That pair cannot be converted, and MetaVidence refuses the file rather than accept it. Left alone, the meta package would drop those studies from the analysis without a word, and a meta-analysis that is quietly missing studies is worse than an error message.
The Data check otherwise works as in the binary module: it verifies the required columns, counts the studies, lists the columns left for subgroup and meta-regression and flags problems, including quartiles out of order.
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 exactly as in the binary module.
Summary measure (sm) (default MD) decides the scale of the pooled result:
MD, mean difference. The result comes out in the original measurement unit, for example “3.2 fewer days in hospital”. It assumes every study used the same scale.SMD, standardized mean difference. The result comes out in standard deviation units, with no measurement unit, so studies that used different instruments can be combined.
Statistical settings (advanced)
Analysis model, Tau method, Random CI method and Prediction method work exactly as in the flow.
Prediction interval here is on by default (Yes), unlike the binary module. Turn it off if you have fewer than 3 studies, or if the interval is so wide that it gets in the way of reading the figure.
There is no Method field in this module. Mantel-Haenszel, Peto and GLMM are methods for event counts, and do not apply to means.
Step 3, export and results
The cards are the same as in the binary module, with the same behavior:
- Forest plot, with each study effect, the weights and the pooled diamond.
- Leave-one-out analysis, to see whether a single study holds up the result.
- Funnel plot and Small-study effects, the latter with Egger’s test.
- Subgroup analysis, with the column selector inside the card itself.
- Meta-regression, with numeric moderators and a bubble plot.
- Subgroup summary table and Meta-regression summary table, as XLSX.
In continuous outcomes it is common for some studies to report change from baseline and others the final value. Mixing the two without thinking produces a meaningless pooled effect. Standardize before loading the spreadsheet.