← allmeta

Bucher Indirect Comparison

Compare treatments A and B indirectly through a named common comparator C. Supply independently pooled A–C and B–C effects on the same analysis scale; the A–B estimate is d(A,C) − d(B,C), with variance equal to the sum of the two variances.

Assumes transitivity. The indirect A-vs-B estimate is valid only if the A-vs-C and B-vs-C trials are similar in effect modifiers (population, dose, follow-up, risk of bias). If those trial sets differ, the estimate is biased — check transitivity before interpreting.

Input

Target comparisons for CSV or network import

Treatment A Treatment B Common comparator C

CSV effects must be treatment1 minus treatment2; use logged effects for ratios. The importer pools only A–C and B–C, makes direction explicit and reports separately any direct A–B rows excluded from the indirect calculation. Multi-arm/overlapping study IDs are rejected.

A vs C summary

Effect d(AC) SE

B vs C summary

Effect d(BC) SE

Scale

Indirect comparison

Comparison	Estimate	SE	95% CI	p

Forest

Download chart:

Export results:

Bucher HC, Guyatt GH, Griffith LE, Walter SD. J Clin Epidemiol 1997;50:683–691. Adjusted indirect comparison assumes independent evidence on both links, identical estimands/scales and transitivity. A normal/Wald 95% CI uses the supplied or fixed-effect inverse-variance summary SEs; a trial-count t distribution is not automatically justified. Multi-arm trials require covariance-aware network meta-analysis. The historical tests/fixtures/bucher-tiny.R validates a differently oriented B-versus-C contrast and must not be treated as verification of arbitrary CSV treatment selection.