bayesmeta (version 3.4)

NicholasEtAl2019: Multiple sclerosis disability progression example data

Description

Proportions of patients with disability progression in the placebo groups of 28 studies.

Usage

data("NicholasEtAl2019")

Format

The data frame contains the following columns:

study	character	publication identifier (first author and publication year)
year	numeric	publication year
patients	numeric	number of placebo patients
prog.percent	numeric	percentage of patients with disability progression

Details

A systematic literature review investigated the characteristics of randomized placebo-controlled trials in multiple sclerosis published between 1988 and 2018 (Nicholas et al., 2019). A number of trends were observed in the trial characteristics over the investigated period; one of these was a decline in the proportion of placebo patients experiencing disability progression within 24 months during the course of a study. The data set contains the placebo groups' sizes along with the percentages of progressing patients within that group for 28 studies. The data were originally extracted from tables or Kaplan-Meier curves.

References

C. Roever, T. Friede. Using the bayesmeta R package for Bayesian random-effects meta-regression. Computer Methods and Programs in Biomedicine, 299:107303, 2023. tools:::Rd_expr_doi("10.1016/j.cmpb.2022.107303").

Examples

# load data:
data("NicholasEtAl2019")

# show data:
head(NicholasEtAl2019)

if (FALSE) {
# compute effect sizes (logarithmic odds) from count data
# (note: effect of potential drop-outs is ignored here):
es <- escalc(measure="PLO",
             xi=patients*(prog.percent/100), ni=patients,
             slab=study, data=NicholasEtAl2019)

# illustrate estimates (log-odds):
forestplot(es, zero=NA, xlab="log(odds)", title="Nicholas et al. (2019) data")

# set up regressor matrix
# (note: "year" variable is re-scaled so that the intercept
# corresponds to the log-odds at year=2000):
X <- cbind("intercept2000" = 1, "year" = (es$year-2000))

# perform analysis:
bmr01 <- bmr(es, X=X)

# show results:
print(bmr01)
plot(bmr01)

# illustrate the data and time trend;
# first derive predictions from the model
# and specify corresponding "regressor matrix":
newx <- cbind(1, (1989:2019)-2000)

# compute credible intervals for the mean:
pred <- cbind("median"=bmr01$qpred(0.5, x=newx),
              bmr01$pred.interval(x=newx))

# compute prediction intervals:
map <- cbind("median"=bmr01$qpred(0.5, x=newx, mean=FALSE),
              bmr01$pred.interval(x=newx, mean=FALSE))

# draw empty plot:
plot(range(newx[,2]), range(map), type="n",
     xlab="publication year - 2000", ylab="log(odds)")

# show the 26 studies' estimates (and 95 percent CIs):
matlines(rbind(es$year, es$year)-2000,
         rbind(es$yi-qnorm(0.975)*sqrt(es$vi), es$yi+qnorm(0.975)*sqrt(es$vi)),
         col=1, lty=1)
points(es$year-2000, es$yi)

# show trend lines (and 95 percent intervals):
matlines(newx[,2], map, col="blue", lty=c(1,2,2))
matlines(newx[,2], pred, col="red", lty=c(1,2,2))
legend("topright", pch=15, col=c("red","blue"), c("mean","prediction"))
}