Forest Plot - performing Meta-analysis in subgroups. Using the Meta package in the R programming language.
# https://www.rdocumentation.org/packages/grid/versions/3.6.2
# https://cran.r-project.org/web/packages/readxl/index.html
# https://cran.r-project.org/web/packages/meta/index.html
library(readxl)
library(grid)
library(meta)# Importe table
dat <- read_xlsx("LP.xlsx")
dat# A tibble: 16 x 8
# study Measure nd meand sdd n mean sd
# <chr> <chr> <dbl> <dbl> <dbl> <dbl> <dbl> <dbl>
# 1 Study 1 CT 6 129.4 5.14 6 145. 11.2
# 2 Study 2 CT 6 98.4 3.28 6 135. 1.96
# 3 Study 3 CT 10 54.3 2.8 10 56.1 2.6
# 4 Study 4 CT 6 90.5 1.87 6 130. 3.02
# 5 Study 5 CT 8 72.7 7.4 8 103 5.5
# 6 Study 6 CT 6 90.3 2.72 6 107. 3.09
# 7 Study 7 CT 6 92.3 2.71 6 106. 3.05
# 8 Study 8 TG 6 128. 6.93 6 269. 13.6
# 9 Study 9 TG 6 110. 2.48 6 25.7 1.16
#10 Study 10 TG 10 60.6 7.5 10 85.5 13.3
#11 Study 11 TG 6 93.3 1.51 6 153 3.35
#12 Study 12 TG 6 88.5 3.23 6 105. 2.8
#13 Study 13 TG 6 88.5 3.23 6 75.4 3.15
#14 Study 14 HDL 6 34.6 2.39 6 22.7 2.58
#15 Study 15 HDL 6 42.5 2.98 6 25.7 1.16
#16 Study 16 HDL 6 31.4 0.82 6 25.2 0.98dat.frame_2 <- data.frame(dat)| study | Measure | nd | meand | sdd | n | mean | sd |
|---|---|---|---|---|---|---|---|
| Study 1 | CT | 6 | 129,46 | 5,14 | 6 | 144,79 | 11,25 |
| Study 2 | CT | 6 | 98,42 | 3,28 | 6 | 134,83 | 1,96 |
| Study 3 | CT | 10 | 54,3 | 2,8 | 10 | 56,1 | 2,6 |
| Study 4 | CT | 6 | 90,5 | 1,87 | 6 | 130,5 | 3,02 |
| Study 5 | CT | 8 | 72,7 | 7,4 | 8 | 103 | 5,5 |
| Study 6 | CT | 6 | 90,3 | 2,72 | 6 | 106,8 | 3,09 |
| Study 7 | CT | 6 | 92,3 | 2,71 | 6 | 106,5 | 3,05 |
| Study 8 | TG | 6 | 127,82 | 6,93 | 6 | 269,3 | 13,59 |
| Study 9 | TG | 6 | 110,22 | 2,48 | 6 | 25,67 | 1,16 |
| Study 10 | TG | 10 | 60,6 | 7,5 | 10 | 85,5 | 13,3 |
| Study 11 | TG | 6 | 93,33 | 1,51 | 6 | 153 | 3,35 |
| Study 12 | TG | 6 | 88,5 | 3,23 | 6 | 105,1 | 2,8 |
| Study 13 | TG | 6 | 88,5 | 3,23 | 6 | 75,41 | 3,15 |
| Study 14 | HDL | 6 | 34,64 | 2,39 | 6 | 22,69 | 2,58 |
| Study 15 | HDL | 6 | 42,46 | 2,98 | 6 | 25,67 | 1,16 |
| Study 16 | HDL | 6 | 31,37 | 0,82 | 6 | 25,16 | 0,98 |
meta_2 <- metacont(dat$n,
dat$meand,
dat$sdd,
dat$nd,
dat$mean,
dat$sd,
dat$study,
data = dat.frame_2,
byvar = Measure,
comb.fixed = TRUE,
sm="SMD"
)> meta_2
SMD 95%-CI %W(fixed) %W(random) Measure
Study 1 -1.6180 [ -2.9980; -0.2380] 13.8 7.9 CT
Study 2 -12.4393 [-18.6163; -6.2624] 0.7 4.9 CT
Study 3 -0.6381 [ -1.5419; 0.2658] 32.2 8.1 CT
Study 4 -14.7004 [-21.9653; -7.4355] 0.5 4.2 CT
Study 5 -4.3940 [ -6.4029; -2.3852] 6.5 7.7 CT
Study 6 -5.2324 [ -8.0260; -2.4387] 3.4 7.2 CT
Study 7 -4.5434 [ -7.0333; -2.0535] 4.2 7.4 CT
Study 8 -12.1070 [-18.1246; -6.0895] 0.7 5.0 TG
Study 9 40.3136 [ 20.6015; 60.0258] 0.1 1.0 TG
Study 10 -2.2088 [ -3.3715; -1.0461] 19.5 8.0 TG
Study 11 -21.1983 [-31.6082; -10.7884] 0.2 2.8 TG
Study 12 -5.0694 [ -7.7906; -2.3483] 3.6 7.3 TG
Study 13 3.7875 [ 1.6198; 5.9552] 5.6 7.6 TG
Study 14 4.4357 [ 1.9925; 6.8789] 4.4 7.4 HDL
Study 15 6.8541 [ 3.3220; 10.3862] 2.1 6.7 HDL
Study 16 6.3442 [ 3.0469; 9.6415] 2.4 6.9 HDL
Number of studies combined: k = 16
SMD 95%-CI z p-value
Fixed effect model -1.2611 [-1.7742; -0.7480] -4.82 < 0.0001
Random effects model -2.2628 [-4.3930; -0.1327] -2.08 0.0373
Quantifying heterogeneity:
tau^2 = 14.3627 [20.6661; 124.3353]; tau = 3.7898 [4.5460; 11.1506]
I^2 = 92.0% [88.6%; 94.4%]; H = 3.54 [2.96; 4.22]
Test of heterogeneity:
Q d.f. p-value
187.76 15 < 0.0001
Results for subgroups (fixed effect model):
k SMD 95%-CI Q I^2
CT 7 -2.0276 [-2.6825; -1.3727] 46.33 87.0%
TG 6 -1.7203 [-2.6622; -0.7783] 73.66 93.2%
HDL 3 5.5232 [ 3.8074; 7.2391] 1.54 0.0%
Results for subgroups (random effects model):
k SMD 95%-CI tau^2 tau
CT 7 -4.6427 [-6.9204; -2.3651] 6.8810 2.6232
TG 6 -3.2292 [-8.3922; 1.9338] 30.5857 5.5304
HDL 3 5.5232 [ 3.8074; 7.2391] 0 0
Plot forest and read the comments in the code about the functions used.
#RE.res <- rma(n,meand,sdd,mean,sd, data=dat.frame_2, slab=paste(Measure))
#RE.res
png(file = 'LP - Forestplot.png', # Save plot as PNG
width=780,
height=620)
forest(meta_2,
order=order(dat$study), # Where you should initially order the data.
leftlabs = c("Lipid Profile
Author", "Total","Mean","SD","Total","Mean","SD"),
comb.random=FALSE, # Do not plot the random effect.
xlim = c(-70,70),
xlab="Standarized Mean Difference (95% CI)",
mlab="RE Model for All Studies",
subgroup = TRUE,
print.byvar = FALSE, # Don't print meta_2$$Measure.
)
dev.off()
-
R Core Team. (2017). R: A language and environment for statistical computing. R Foundation for Statistical Computing, Vienna, Austria. Retrieved from https://www.R-project.org/
-
Veroniki AA, Jackson D, Viechtbauer W, Bender R, Bowden J, Knapp G, et al. (2016): Methods to estimate the between-study variance and its uncertainty in meta-analysis. Research Synthesis Methods, 7, 55–79