dat.hasselblad1998 {metadat} | R Documentation |
Studies on the Effectiveness of Counseling for Smoking Cessation
Description
Results from 24 studies on the effectiveness of various counseling types for smoking cessation.
Usage
dat.hasselblad1998
Format
The data frame contains the following columns:
id | numeric | id number for each treatment arm |
study | numeric | study id number |
authors | character | study author(s) |
year | numeric | publication year |
trt | character | intervention group |
xi | numeric | number of individuals abstinent |
ni | numeric | number of individuals in group |
Details
The dataset includes the results from 24 studies on the effectiveness of various counseling types for smoking cessation (i.e., self-help, individual counseling, group counseling, and no contact). The dataset indicates the total number of individuals within each study arm and the number that were abstinent from 6 to 12 months. The majority of the studies compared two interventions types against each other, while 2 studies compared three types against each other simultaneously.
The data can be used for a ‘network meta-analysis’ (also called a ‘mixed treatment comparison’). The code below shows how such an analysis can be conducted using an arm-based and a contrast-based model (see Salanti et al., 2008, for more details).
Concepts
medicine, psychology, smoking, odds ratios, network meta-analysis
Author(s)
Wolfgang Viechtbauer, wvb@metafor-project.org, https://www.metafor-project.org
Source
Hasselblad, V. (1998). Meta-analysis of multitreatment studies. Medical Decision Making, 18(1), 37–43. https://doi.org/10.1177/0272989X9801800110
References
Gleser, L. J., & Olkin, I. (2009). Stochastically dependent effect sizes. In H. Cooper, L. V. Hedges, & J. C. Valentine (Eds.), The handbook of research synthesis and meta-analysis (2nd ed., pp. 357–376). New York: Russell Sage Foundation.
Law, M., Jackson, D., Turner, R., Rhodes, K., & Viechtbauer, W. (2016). Two new methods to fit models for network meta-analysis with random inconsistency effects. BMC Medical Research Methodology, 16, 87. https://doi.org/10.1186/s12874-016-0184-5
Salanti, G., Higgins, J. P. T., Ades, A. E., & Ioannidis, J. P. A. (2008). Evaluation of networks of randomized trials. Statistical Methods in Medical Research, 17(3), 279–301. https://doi.org/10.1177/0962280207080643
Examples
### copy data into 'dat' and examine data
dat <- dat.hasselblad1998
dat
## Not run:
### load metafor package
library(metafor)
### create network graph ('igraph' package must be installed)
library(igraph, warn.conflicts=FALSE)
pairs <- data.frame(do.call(rbind,
sapply(split(dat$trt, dat$study), function(x) t(combn(x,2)))), stringsAsFactors=FALSE)
lvls <- c("no_contact", "self_help", "ind_counseling", "grp_counseling")
pairs$X1 <- factor(pairs$X1, levels=lvls)
pairs$X2 <- factor(pairs$X2, levels=lvls)
tab <- table(pairs[,1], pairs[,2])
tab # adjacency matrix
g <- graph_from_adjacency_matrix(tab, mode = "plus", weighted=TRUE, diag=FALSE)
vertex_attr(g, "name") <- c("No Contact", "Self-Help",
"Individual\nCounseling", "Group\nCounseling")
plot(g, edge.curved=FALSE, edge.width=E(g)$weight, layout=layout_on_grid,
vertex.size=45, vertex.color="lightgray", vertex.label.color="black", vertex.label.font=2)
### calculate log odds for each study arm
dat <- escalc(measure="PLO", xi=xi, ni=ni, add=1/2, to="all", data=dat)
dat
### convert trt variable to factor with desired ordering of levels
dat$trt <- factor(dat$trt, levels=c("no_contact", "self_help", "ind_counseling", "grp_counseling"))
### add a space before each level (this makes the output a bit more legible)
levels(dat$trt) <- paste0(" ", levels(dat$trt))
### network meta-analysis using an arm-based model with fixed study effects
### by setting rho=1/2, tau^2 reflects the amount of heterogeneity for all treatment comparisons
res <- rma.mv(yi, vi, mods = ~ factor(study) + trt - 1,
random = ~ trt | study, rho=1/2, data=dat, btt="trt")
res
### all pairwise odds ratios of interventions versus no contact
predict(res, newmods=cbind(matrix(0, nrow=3, ncol=24), diag(3)),
intercept=FALSE, transf=exp, digits=2)
### all pairwise odds ratios comparing interventions (ic vs sh, gc vs sh, and gc vs ic)
predict(res, newmods=cbind(matrix(0, nrow=3, ncol=24), rbind(c(-1,1,0), c(-1,0,1), c(0,-1,1))),
intercept=FALSE, transf=exp, digits=2)
### forest plot of ORs of interventions versus no contact
dev.new(width=7, height=4)
par(mar=c(5,4,1,2))
forest(c(0,res$beta[25:27]), sei=c(0,res$se[25:27]), psize=1, xlim=c(-3,4), digits=c(2,1), efac=2,
slab=c("No Contact", "Self-Help", "Individual Counseling", "Group Counseling"),
atransf=exp, at=log(c(.5, 1, 2, 4, 8)), xlab="Odds Ratio for Intervention vs. No Contact",
header=c("Intervention", "Odds Ratio [95% CI]"))
############################################################################
### restructure dataset to a contrast-based format
dat <- to.wide(dat.hasselblad1998, study="study", grp="trt", ref="no_contact", grpvars=6:7)
### calculate log odds ratios for each treatment comparison
dat <- escalc(measure="OR", ai=xi.1, n1i=ni.1,
ci=xi.2, n2i=ni.2, add=1/2, to="all", data=dat)
dat
### calculate the variance-covariance matrix of the log odds ratios for multitreatment studies
### see Gleser & Olkin (2009), equation (19.11), for the covariance equation
calc.v <- function(x) {
v <- matrix(1/(x$xi.2[1] + 1/2) + 1/(x$ni.2[1] - x$xi.2[1] + 1/2), nrow=nrow(x), ncol=nrow(x))
diag(v) <- x$vi
v
}
V <- bldiag(lapply(split(dat, dat$study), calc.v))
### add contrast matrix to dataset
dat <- contrmat(dat, grp1="trt.1", grp2="trt.2")
dat
### network meta-analysis using a contrast-based random-effects model
### by setting rho=1/2, tau^2 reflects the amount of heterogeneity for all treatment comparisons
res <- rma.mv(yi, V, mods = ~ self_help + ind_counseling + grp_counseling - 1,
random = ~ comp | study, rho=1/2, data=dat)
res
### predicted odds ratios of interventions versus no contact
predict(res, newmods=diag(3), transf=exp, digits=2)
### fit random inconsistency effects model (see Law et al., 2016)
res <- rma.mv(yi, V, mods = ~ self_help + ind_counseling + grp_counseling - 1,
random = list(~ comp | study, ~ comp | design), rho=1/2, phi=1/2, data=dat)
res
## End(Not run)