```
library(multinma)
options(mc.cores = parallel::detectCores())
```

This vignette describes the analysis of 19 trials comparing statins to placebo or usual care (Dias et al. 2011). The data are available in this package as `statins`

:

```
head(statins)
#> studyn studyc trtn trtc prevention r n
#> 1 1 4S 1 Placebo Secondary 256 2223
#> 2 1 4S 2 Statin Secondary 182 2221
#> 3 2 Bestehorn 1 Placebo Secondary 4 125
#> 4 2 Bestehorn 2 Statin Secondary 1 129
#> 5 3 Brown 1 Placebo Secondary 0 52
#> 6 3 Brown 2 Statin Secondary 1 94
```

Dias et al. (2011) used these data to demonstrate meta-regression models adjusting for the binary covariate `prevention`

(primary or secondary prevention), which we recreate here.

We have data giving the number of deaths (`r`

) out of the total (`n`

) in each arm, so we use the function `set_agd_arm()`

to set up the network. We set placebo as the network reference treatment.

```
<- set_agd_arm(statins,
statin_net study = studyc,
trt = trtc,
r = r,
n = n,
trt_ref = "Placebo")
statin_net#> A network with 19 AgD studies (arm-based).
#>
#> ------------------------------------------------------- AgD studies (arm-based) ----
#> Study Treatment arms
#> 4S 2: Placebo | Statin
#> Bestehorn 2: Placebo | Statin
#> Brown 2: Placebo | Statin
#> CCAIT 2: Placebo | Statin
#> Downs 2: Placebo | Statin
#> EXCEL 2: Placebo | Statin
#> Furberg 2: Placebo | Statin
#> Haskell 2: Placebo | Statin
#> Jones 2: Placebo | Statin
#> KAPS 2: Placebo | Statin
#> ... plus 9 more studies
#>
#> Outcome type: count
#> ------------------------------------------------------------------------------------
#> Total number of treatments: 2
#> Total number of studies: 19
#> Reference treatment is: Placebo
#> Network is connected
```

The `prevention`

variable in the `statins`

data frame will automatically be available to use in a meta-regression model.

We fit fixed effect (FE) and random effects (RE) models, with a meta-regression on the binary covariate `prevention`

.

We start by fitting a FE model. We use \(\mathrm{N}(0, 100^2)\) prior distributions for the treatment effect \(d_\mathrm{Statin}\), study-specific intercepts \(\mu_j\), and regression coefficient \(\beta\). We can examine the range of parameter values implied by these prior distributions with the `summary()`

method:

```
summary(normal(scale = 100))
#> A Normal prior distribution: location = 0, scale = 100.
#> 50% of the prior density lies between -67.45 and 67.45.
#> 95% of the prior density lies between -196 and 196.
```

The model is fitted with the `nma()`

function, with a fixed effect model specified by `trt_effects = "fixed"`

. The `regression`

formula `~ .trt:prevention`

means that interaction of primary/secondary prevention with treatment will be included; the `.trt`

special variable indicates treatment, and `prevention`

is in the original data set.

```
<- nma(statin_net,
statin_fit_FE trt_effects = "fixed",
regression = ~.trt:prevention,
prior_intercept = normal(scale = 100),
prior_trt = normal(scale = 100),
prior_reg = normal(scale = 100))
#> Note: No treatment classes specified in network, any interactions in `regression` formula will be separate (independent) for each treatment.
#> Use set_*() argument `trt_class` and nma() argument `class_interactions` to change this.
```

Basic parameter summaries are given by the `print()`

method:

```
statin_fit_FE#> A fixed effects NMA with a binomial likelihood (logit link).
#> Regression model: ~.trt:prevention.
#> Inference for Stan model: binomial_1par.
#> 4 chains, each with iter=2000; warmup=1000; thin=1;
#> post-warmup draws per chain=1000, total post-warmup draws=4000.
#>
#> mean se_mean sd 2.5% 25% 50% 75%
#> beta[.trtStatin:preventionSecondary] -0.21 0.00 0.11 -0.43 -0.28 -0.21 -0.14
#> d[Statin] -0.10 0.00 0.10 -0.29 -0.17 -0.10 -0.04
#> lp__ -7246.59 0.08 3.28 -7253.97 -7248.57 -7246.24 -7244.23
#> 97.5% n_eff Rhat
#> beta[.trtStatin:preventionSecondary] 0.00 2212 1
#> d[Statin] 0.08 2164 1
#> lp__ -7241.06 1738 1
#>
#> Samples were drawn using NUTS(diag_e) at Thu Feb 24 09:07:41 2022.
#> For each parameter, n_eff is a crude measure of effective sample size,
#> and Rhat is the potential scale reduction factor on split chains (at
#> convergence, Rhat=1).
```

By default, summaries of the study-specific intercepts \(\mu_j\) are hidden, but could be examined by changing the `pars`

argument:

```
# Not run
print(statin_fit_FE, pars = c("d", "beta", "mu"))
```

The prior and posterior distributions can be compared visually using the `plot_prior_posterior()`

function:

`plot_prior_posterior(statin_fit_FE, prior = c("trt", "reg"))`

We now fit a RE model. We use \(\mathrm{N}(0, 100^2)\) prior distributions for the treatment effect \(d_\mathrm{Statin}\), study-specific intercepts \(\mu_j\), and regression coefficient \(\beta\). We use a \(\textrm{half-N}(0, 5^2)\) prior distribution for the heterogeneity standard deviation \(\tau\). We can examine the range of parameter values implied by these prior distributions with the `summary()`

method:

```
summary(normal(scale = 100))
#> A Normal prior distribution: location = 0, scale = 100.
#> 50% of the prior density lies between -67.45 and 67.45.
#> 95% of the prior density lies between -196 and 196.
summary(half_normal(scale = 5))
#> A half-Normal prior distribution: location = 0, scale = 5.
#> 50% of the prior density lies between 0 and 3.37.
#> 95% of the prior density lies between 0 and 9.8.
```

Again, the model is fitted with the `nma()`

function, now with `trt_effects = "random"`

. We increase `adapt_delta`

to 0.99 to remove a small number of divergent transition errors (the default for RE models is set to 0.95).

```
<- nma(statin_net,
statin_fit_RE trt_effects = "random",
regression = ~.trt:prevention,
prior_intercept = normal(scale = 100),
prior_trt = normal(scale = 100),
prior_reg = normal(scale = 100),
prior_het = half_normal(scale = 5),
adapt_delta = 0.99)
#> Note: No treatment classes specified in network, any interactions in `regression` formula will be separate (independent) for each treatment.
#> Use set_*() argument `trt_class` and nma() argument `class_interactions` to change this.
```

Basic parameter summaries are given by the `print()`

method:

```
statin_fit_RE#> A random effects NMA with a binomial likelihood (logit link).
#> Regression model: ~.trt:prevention.
#> Inference for Stan model: binomial_1par.
#> 4 chains, each with iter=2000; warmup=1000; thin=1;
#> post-warmup draws per chain=1000, total post-warmup draws=4000.
#>
#> mean se_mean sd 2.5% 25% 50% 75%
#> beta[.trtStatin:preventionSecondary] -0.30 0.01 0.27 -0.90 -0.43 -0.28 -0.14
#> d[Statin] -0.07 0.01 0.21 -0.48 -0.18 -0.08 0.04
#> lp__ -7255.76 0.16 5.25 -7266.54 -7259.18 -7255.58 -7252.10
#> tau 0.25 0.01 0.21 0.01 0.09 0.20 0.34
#> 97.5% n_eff Rhat
#> beta[.trtStatin:preventionSecondary] 0.22 1499 1
#> d[Statin] 0.39 1452 1
#> lp__ -7246.28 1032 1
#> tau 0.79 737 1
#>
#> Samples were drawn using NUTS(diag_e) at Thu Feb 24 09:08:02 2022.
#> For each parameter, n_eff is a crude measure of effective sample size,
#> and Rhat is the potential scale reduction factor on split chains (at
#> convergence, Rhat=1).
```

By default, summaries of the study-specific intercepts \(\mu_j\) and study-specific relative effects \(\delta_{jk}\) are hidden, but could be examined by changing the `pars`

argument:

```
# Not run
print(statin_fit_RE, pars = c("d", "beta", "mu", "delta"))
```

The prior and posterior distributions can be compared visually using the `plot_prior_posterior()`

function:

`plot_prior_posterior(statin_fit_RE, prior = c("trt", "reg", "het"))`

Model fit can be checked using the `dic()`

function:

```
<- dic(statin_fit_FE))
(statin_dic_FE #> Residual deviance: 45.8 (on 38 data points)
#> pD: 21.5
#> DIC: 67.3
```

```
<- dic(statin_fit_RE))
(statin_dic_RE #> Residual deviance: 42.5 (on 38 data points)
#> pD: 25.1
#> DIC: 67.6
```

The DIC is very similar between FE and RE models, so we might choose the FE model based on parsimony. The residual deviance statistics are larger than the number of data points, suggesting that some data points are not fit well.

We can also examine the residual deviance contributions with the corresponding `plot()`

method.

`plot(statin_dic_FE)`

`plot(statin_dic_RE)`

There are a number of studies which are not fit well under either model, having posterior mean residual deviance contributions greater than 1, and should be investigated to see if there are further substantive differences between studies.

We can produce estimates of the relative effect of statins vs. placebo for either primary or secondary prevention, using the `relative_effects()`

function. The `newdata`

argument specifies a data frame containing the levels of the covariate `prevention`

that we are interested in, and the `study`

argument is used to specify a column of `newdata`

for an informative label.

```
<- relative_effects(statin_fit_FE,
statin_releff_FE newdata = data.frame(prevention = c("Primary", "Secondary")),
study = prevention)
statin_releff_FE#> ---------------------------------------------------------------- Study: Primary ----
#>
#> Covariate values:
#> prevention
#> Primary
#>
#> mean sd 2.5% 25% 50% 75% 97.5% Bulk_ESS Tail_ESS Rhat
#> d[Primary: Statin] -0.1 0.1 -0.29 -0.17 -0.1 -0.04 0.08 2200 2354 1
#>
#> -------------------------------------------------------------- Study: Secondary ----
#>
#> Covariate values:
#> prevention
#> Secondary
#>
#> mean sd 2.5% 25% 50% 75% 97.5% Bulk_ESS Tail_ESS Rhat
#> d[Secondary: Statin] -0.31 0.05 -0.42 -0.35 -0.31 -0.28 -0.21 4047 3614 1
```

The `plot()`

method may be used to visually compare these estimates:

```
plot(statin_releff_FE,
ref_line = 0)
```

Model parameters may be plotted with the corresponding `plot()`

method:

```
plot(statin_fit_FE,
pars = "beta",
ref_line = 0,
stat = "halfeye")
```

Whilst the 95% Credible Interval includes zero, there is a suggestion that statins are more effective for secondary prevention.

Dias, S., A. J. Sutton, N. J. Welton, and A. E. Ades. 2011. “NICE DSU Technical Support Document 3: Heterogeneity: Subgroups, Meta-Regression, Bias and Bias-Adjustment.” National Institute for Health and Care Excellence. https://nicedsu.sites.sheffield.ac.uk.