Conducts network meta-analysis (NMA) to compare multiple treatments simultaneously using direct and indirect evidence.

Conducts network meta-analysis (NMA) to compare multiple treatments simultaneously using direct and indirect evidence.

Usage

network_meta(
  data,
  study_var = "study",
  treat1_var = "treat1",
  treat2_var = "treat2",
  effect_var = "effect",
  se_var = "se",
  reference = NULL,
  effect_measure = c("hr", "or", "rr", "rd", "md", "smd"),
  model = c("random", "fixed"),
  method = c("bucher", "graph"),
  conf_level = 0.95
)

Arguments

data

Data frame with study-level data. Required columns: study (study identifier), treat1 (treatment 1), treat2 (treatment 2), effect (effect estimate), se (standard error)

study_var

Character. Study identifier column. Default: "study"

treat1_var

Character. Treatment 1 column. Default: "treat1"

treat2_var

Character. Treatment 2 column. Default: "treat2"

effect_var

Character. Effect estimate column. Default: "effect"

se_var

Character. Standard error column. Default: "se"

reference

Character. Reference treatment. Default: first alphabetically

effect_measure

Character. Effect type: "hr", "or", "rr", "rd", "md", "smd"

model

Character. "fixed" or "random". Default: "random"

method

Character. NMA method: "bucher" (simple), "graph" (not yet implemented, defaults to bucher)

conf_level

Numeric. Confidence level. Default: 0.95

Value

List with relative effects, rankings, and network structure

Details

For ratio measures ("hr", "or", "rr"), effect estimates must be on the log scale. For difference measures ("rd", "md", "smd"), effect estimates must be on the raw scale.

Indirect effects are estimated using a single-step Bucher chain via one intermediate treatment (chosen to minimize the indirect standard error). Multiple indirect paths are not combined.

Examples

# Network meta-analysis of 4 studies
nma_data <- data.frame(
  study = c("S1", "S2", "S3", "S4"),
  treat1 = c("A", "B", "A", "B"),
  treat2 = c("B", "C", "C", "D"),
  effect = log(c(0.75, 0.90, 0.80, 0.85)),
  se = c(0.12, 0.15, 0.18, 0.14)
)
result <- network_meta(nma_data, effect_measure = "hr")
result@comparisons
#> # A tibble: 4 × 9
#>   treatment vs    estimate ci_lower ci_upper    se n_studies evidence   rank
#>   <chr>     <chr>    <dbl>    <dbl>    <dbl> <dbl>     <int> <chr>     <dbl>
#> 1 A         A        1        1        1     0            NA reference    NA
#> 2 B         A        0.75     0.593    0.949 0.12          1 direct        2
#> 3 C         A        0.8      0.562    1.14  0.18          1 direct        3
#> 4 D         A        0.638    0.444    0.915 0.184         2 indirect      1
result@network$treatments
#> [1] "A" "B" "C" "D"