Meta-analysis reperfusion
Niels Pacheco
2024-08-25
# Creating the dataframe
meta_repu_or <- data.frame(
Study = c("Richardt, 2024", "Béjot, 2023", "Dicpinigaitis, 2021"),
ich_c = c(8, 4, 25),
ich_c_tot = c(85, 354, 180),
ich_i = c(3, 0, 20),
ich_i_tot = c(12, 28, 180),
sich_c = c(1, NA, NA),
sich_c_tot = c(85, NA, NA),
sich_i = c(0, NA, NA),
sich_i_tot = c(12, NA, NA),
death_c = c(4, 11, NA),
death_c_tot = c(85, 354, NA),
death_i = c(1, 0, NA),
death_i_tot = c(12, 28, NA),
venous_c = c(NA, 2, 0),
venous_c_tot = c(NA, 354, 180),
venous_i = c(NA, 0, 30),
venous_i_tot = c(NA, 28, 180),
mrs_c = c(64, NA, 130),
mrs_c_tot = c(85, NA, 180),
mrs_i = c(8, NA, 90),
mrs_i_tot = c(12, NA, 180)
)
# Display the dataframe
print(meta_repu_or)## Study ich_c ich_c_tot ich_i ich_i_tot sich_c sich_c_tot sich_i
## 1 Richardt, 2024 8 85 3 12 1 85 0
## 2 Béjot, 2023 4 354 0 28 NA NA NA
## 3 Dicpinigaitis, 2021 25 180 20 180 NA NA NA
## sich_i_tot death_c death_c_tot death_i death_i_tot venous_c venous_c_tot
## 1 12 4 85 1 12 NA NA
## 2 NA 11 354 0 28 2 354
## 3 NA NA NA NA NA 0 180
## venous_i venous_i_tot mrs_c mrs_c_tot mrs_i mrs_i_tot
## 1 NA NA 64 85 8 12
## 2 0 28 NA NA NA NA
## 3 30 180 130 180 90 180library(meta)## Loading required package: metadat## Loading 'meta' package (version 7.0-0).
## Type 'help(meta)' for a brief overview.
## Readers of 'Meta-Analysis with R (Use R!)' should install
## older version of 'meta' package: https://tinyurl.com/dt4y5drs# Meta-analysis of odds ratios
meta_analysis_ich <- metabin(
event.e = meta_repu_or$ich_i, # Number of events in intervention group (ich_i)
n.e = meta_repu_or$ich_i_tot, # Total number in intervention group (ich_i_tot)
event.c = meta_repu_or$ich_c, # Number of events in control group (ich_c)
n.c = meta_repu_or$ich_c_tot, # Total number in control group (ich_c_tot)
data = meta_repu_or,
studlab = paste(meta_repu_or$Study),
sm = "OR", # Specify summary measure as Odds Ratio
method.tau = "DL", # DerSimonian-Laird estimator for tau^2
comb.fixed = FALSE, # Random effects model
comb.random = TRUE, # Include random effects
prediction = FALSE # No prediction interval by default
)
# Summary of the meta-analysis
summary(meta_analysis_ich)## OR 95%-CI %W(random)
## Richardt, 2024 3.2083 [0.7189; 14.3183] 27.7
## Béjot, 2023 1.3665 [0.0718; 26.0186] 9.4
## Dicpinigaitis, 2021 0.7750 [0.4135; 1.4524] 62.9
##
## Number of studies: k = 3
## Number of observations: o = 839 (o.e = 220, o.c = 619)
## Number of events: e = 60
##
## OR 95%-CI z p-value
## Random effects model 1.2124 [0.4657; 3.1566] 0.39 0.6932
##
## Quantifying heterogeneity:
## tau^2 = 0.2765; tau = 0.5258; I^2 = 33.6% [0.0%; 78.1%]; H = 1.23 [1.00; 2.14]
##
## Test of heterogeneity:
## Q d.f. p-value
## 3.01 2 0.2217
##
## Details on meta-analytical method:
## - Inverse variance method
## - DerSimonian-Laird estimator for tau^2
## - Mantel-Haenszel estimator used in calculation of Q and tau^2 (like RevMan 5)
## - Continuity correction of 0.5 in studies with zero cell frequenciesMETA-ANALYSIS OF ODDS RATIOS OF ICH (OR>1 favors intervention)
# To visualize the results, you can plot a forest plot
meta::forest(meta_analysis_ich, layout = "JAMA")
Insuficient data for the sich
META-ANALYSIS OF DEATH
# Load the necessary library
library(meta)
# Filter out rows with NA values in relevant columns
meta_repu_or_filtered <- subset(meta_repu_or, !is.na(death_c) & !is.na(death_c_tot) & !is.na(death_i) & !is.na(death_i_tot))
# Meta-analysis of odds ratios
meta_analysis_death <- metabin(
event.e = meta_repu_or_filtered$death_i, # Number of events in intervention group (death_i)
n.e = meta_repu_or_filtered$death_i_tot, # Total number in intervention group (death_i_tot)
event.c = meta_repu_or_filtered$death_c, # Number of events in control group (death_c)
n.c = meta_repu_or_filtered$death_c_tot, # Total number in control group (death_c_tot)
data = meta_repu_or_filtered,
studlab = paste(meta_repu_or_filtered$Study),
sm = "OR", # Specify summary measure as Odds Ratio
method.tau = "DL", # DerSimonian-Laird estimator for tau^2
comb.fixed = FALSE, # Random effects model
comb.random = TRUE, # Include random effects
prediction = FALSE # No prediction interval by default
)
# Summary of the meta-analysis
summary(meta_analysis_death)## OR 95%-CI %W(random)
## Richardt, 2024 1.8409 [0.1883; 17.9983] 61.1
## Béjot, 2023 0.5240 [0.0301; 9.1237] 38.9
##
## Number of studies: k = 2
## Number of observations: o = 479 (o.e = 40, o.c = 439)
## Number of events: e = 16
##
## OR 95%-CI z p-value
## Random effects model 1.1291 [0.1900; 6.7095] 0.13 0.8938
##
## Quantifying heterogeneity:
## tau^2 = 0; tau = 0; I^2 = 0.0%; H = 1.00
##
## Test of heterogeneity:
## Q d.f. p-value
## 0.48 1 0.4894
##
## Details on meta-analytical method:
## - Inverse variance method
## - DerSimonian-Laird estimator for tau^2
## - Mantel-Haenszel estimator used in calculation of Q and tau^2 (like RevMan 5)
## - Continuity correction of 0.5 in studies with zero cell frequencies# To visualize the results, you can plot a forest plot
meta::forest(meta_analysis_death, layout = "JAMA")
META-ANALYSIS Venous thromboembolism
# Load the necessary library
library(meta)
# Filter out rows with NA values in relevant columns
meta_repu_or_filtered_venous <- subset(meta_repu_or, !is.na(venous_c) & !is.na(venous_c_tot) & !is.na(venous_i) & !is.na(venous_i_tot))
# Meta-analysis of odds ratios
meta_analysis_venous <- metabin(
event.e = meta_repu_or_filtered_venous$venous_i, # Number of events in intervention group (venous_i)
n.e = meta_repu_or_filtered_venous$venous_i_tot, # Total number in intervention group (venous_i_tot)
event.c = meta_repu_or_filtered_venous$venous_c, # Number of events in control group (venous_c)
n.c = meta_repu_or_filtered_venous$venous_c_tot, # Total number in control group (venous_c_tot)
data = meta_repu_or_filtered_venous,
studlab = paste(meta_repu_or_filtered_venous$Study),
sm = "OR", # Specify summary measure as Odds Ratio
method.tau = "DL", # DerSimonian-Laird estimator for tau^2
comb.fixed = FALSE, # Random effects model
comb.random = TRUE, # Include random effects
prediction = FALSE # No prediction interval by default
)
# Summary of the meta-analysis
summary(meta_analysis_venous)## OR 95%-CI %W(random)
## Béjot, 2023 2.4737 [0.1160; 52.7713] 48.7
## Dicpinigaitis, 2021 73.1595 [4.4364; 1206.4520] 51.3
##
## Number of studies: k = 2
## Number of observations: o = 742 (o.e = 208, o.c = 534)
## Number of events: e = 32
##
## OR 95%-CI z p-value
## Random effects model 14.0625 [0.3161; 625.6522] 1.37 0.1722
##
## Quantifying heterogeneity:
## tau^2 = 5.2630; tau = 2.2941; I^2 = 70.1% [0.0%; 93.3%]; H = 1.83 [1.00; 3.86]
##
## Test of heterogeneity:
## Q d.f. p-value
## 3.35 1 0.0673
##
## Details on meta-analytical method:
## - Inverse variance method
## - DerSimonian-Laird estimator for tau^2
## - Mantel-Haenszel estimator used in calculation of Q and tau^2 (like RevMan 5)
## - Continuity correction of 0.5 in studies with zero cell frequencies# To visualize the results, you can plot a forest plot
meta::forest(meta_analysis_venous, layout = "JAMA")
META-ANALYSIS OF Good functional outcome (mRS 0–2) at 3 months
# Load the necessary library
library(meta)
# Filter out rows with NA values in relevant columns
meta_repu_or_filtered_mrs <- subset(meta_repu_or, !is.na(mrs_c) & !is.na(mrs_c_tot) & !is.na(mrs_i) & !is.na(mrs_i_tot))
# Meta-analysis of odds ratios
meta_analysis_mrs <- metabin(
event.e = meta_repu_or_filtered_mrs$mrs_i, # Number of events in intervention group (mrs_i)
n.e = meta_repu_or_filtered_mrs$mrs_i_tot, # Total number in intervention group (mrs_i_tot)
event.c = meta_repu_or_filtered_mrs$mrs_c, # Number of events in control group (mrs_c)
n.c = meta_repu_or_filtered_mrs$mrs_c_tot, # Total number in control group (mrs_c_tot)
data = meta_repu_or_filtered_mrs,
studlab = paste(meta_repu_or_filtered_mrs$Study),
sm = "OR", # Specify summary measure as Odds Ratio
method.tau = "DL", # DerSimonian-Laird estimator for tau^2
comb.fixed = FALSE, # Random effects model
comb.random = TRUE, # Include random effects
prediction = FALSE # No prediction interval by default
)
# Summary of the meta-analysis
summary(meta_analysis_mrs)## OR 95%-CI %W(random)
## Richardt, 2024 0.6562 [0.1793; 2.4020] 10.2
## Dicpinigaitis, 2021 0.3846 [0.2482; 0.5959] 89.8
##
## Number of studies: k = 2
## Number of observations: o = 457 (o.e = 192, o.c = 265)
## Number of events: e = 292
##
## OR 95%-CI z p-value
## Random effects model 0.4062 [0.2683; 0.6151] -4.26 < 0.0001
##
## Quantifying heterogeneity:
## tau^2 = 0; tau = 0; I^2 = 0.0%; H = 1.00
##
## Test of heterogeneity:
## Q d.f. p-value
## 0.58 1 0.4444
##
## Details on meta-analytical method:
## - Inverse variance method
## - DerSimonian-Laird estimator for tau^2
## - Mantel-Haenszel estimator used in calculation of Q and tau^2 (like RevMan 5)# To visualize the results, you can plot a forest plot
meta::forest(meta_analysis_mrs, layout = "JAMA")
Look at the paper Dicpinigatis, it looks weird that the mrs is better in
the control. Check this in the data extraction form
Now let’s do the meta-analysis of proportions for all the outcomes with analysis of subgroup
# Creating the dataframe
meta_repu_details <- data.frame(
Study = c("Richardt, 2024 i", "Richardt, 2024 c", "Béjot, 2023 i", "Béjot, 2023 c", "Dicpinigaitis, 2021 i", "Dicpinigaitis, 2021 c", "Leffert, 2016", "Limaye, 2019", "Murugappan, 2006"),
sample_ich = c(12, 85, 28, 354, 180, 180, 7, 8, 8),
ich = c(3, 8, 0, 4, 20, 25, 1, 1, 1),
ich_sub = c("Intervention", "Control", "Intervention", "Control", "Intervention", "Control", "Intervention", "Intervention", "Intervention"),
sample_sich = c(12, 85, 28, 354, NA, NA, 40, 7, 8),
sich = c(0, 4, 0, 11, NA, NA, 3, 0, 0),
sich_sub = c("Intervention", "Control", "Intervention", "Control", NA, NA, "Intervention", "Intervention", "Intervention"),
Sample_death = c(12, 85, 28, 354, NA, 338, NA, NA, NA),
death = c(1, 4, 0, 11, NA, 7, NA, NA, NA),
death_sub = c("Intervention", "Control", "Intervention", "Control", NA, "Intervention", NA, NA, NA),
sample_mrs = c(12, 85, NA, NA, 180, 180, 338, 7, NA),
mrs = c(8, 3, NA, NA, 90, 130, 221, 7, NA),
mrs_sub = c("Intervention", "Control", NA, NA, "Intervention", "Control", "Intervention", "Intervention", NA)
)
# Display the dataframe
print(meta_repu_details)## Study sample_ich ich ich_sub sample_sich sich
## 1 Richardt, 2024 i 12 3 Intervention 12 0
## 2 Richardt, 2024 c 85 8 Control 85 4
## 3 Béjot, 2023 i 28 0 Intervention 28 0
## 4 Béjot, 2023 c 354 4 Control 354 11
## 5 Dicpinigaitis, 2021 i 180 20 Intervention NA NA
## 6 Dicpinigaitis, 2021 c 180 25 Control NA NA
## 7 Leffert, 2016 7 1 Intervention 40 3
## 8 Limaye, 2019 8 1 Intervention 7 0
## 9 Murugappan, 2006 8 1 Intervention 8 0
## sich_sub Sample_death death death_sub sample_mrs mrs mrs_sub
## 1 Intervention 12 1 Intervention 12 8 Intervention
## 2 Control 85 4 Control 85 3 Control
## 3 Intervention 28 0 Intervention NA NA <NA>
## 4 Control 354 11 Control NA NA <NA>
## 5 <NA> NA NA <NA> 180 90 Intervention
## 6 <NA> 338 7 Intervention 180 130 Control
## 7 Intervention NA NA <NA> 338 221 Intervention
## 8 Intervention NA NA <NA> 7 7 Intervention
## 9 Intervention NA NA <NA> NA NA <NA># Load the necessary library
library(meta)
# Calculate the pooled proportion using metaprop function
meta_analysis_ich_2 <- metaprop(
event = meta_repu_details$ich, # Number of events (ich)
n = meta_repu_details$sample_ich, # Total number (sample_ich)
data = meta_repu_details, # Data source
sm = "PLO", # Specify summary measure as Proportion (Logit transformed)
method.tau = "DL", # DerSimonian-Laird estimator for tau^2
prediction = FALSE, # No prediction interval by default
comb.fixed = FALSE, # Random effects model
comb.random = TRUE, # Include random effects
studlab = meta_repu_details$Study, # Label each study by its name
byvar = meta_repu_details$ich_sub # Subgroup analysis by ich_sub
)
# Summary of the meta-analysis
summary(meta_analysis_ich_2)## proportion 95%-CI %W(random) ich_sub
## Richardt, 2024 i 0.2500 [0.0549; 0.5719] 10.8 Intervention
## Richardt, 2024 c 0.0941 [0.0415; 0.1771] 16.0 Control
## Béjot, 2023 i 0.0000 [0.0000; 0.1234] 4.0 Intervention
## Béjot, 2023 c 0.0113 [0.0031; 0.0287] 13.6 Control
## Dicpinigaitis, 2021 i 0.1111 [0.0692; 0.1664] 18.4 Intervention
## Dicpinigaitis, 2021 c 0.1389 [0.0919; 0.1982] 18.7 Control
## Leffert, 2016 0.1429 [0.0036; 0.5787] 6.1 Intervention
## Limaye, 2019 0.1250 [0.0032; 0.5265] 6.2 Intervention
## Murugappan, 2006 0.1250 [0.0032; 0.5265] 6.2 Intervention
##
## Number of studies: k = 9
## Number of observations: o = 862
## Number of events: e = 63
##
## proportion 95%-CI
## Random effects model 0.0891 [0.0497; 0.1546]
##
## Quantifying heterogeneity:
## tau^2 = 0.4984 [0.0274; 3.6248]; tau = 0.7060 [0.1657; 1.9039]
## I^2 = 71.6% [44.1%; 85.6%]; H = 1.88 [1.34; 2.64]
##
## Test of heterogeneity:
## Q d.f. p-value
## 28.21 8 0.0004
##
## Results for subgroups (random effects model):
## k proportion 95%-CI tau^2 tau Q
## ich_sub = Intervention 6 0.1187 [0.0822; 0.1683] 0 0 4.03
## ich_sub = Control 3 0.0579 [0.0156; 0.1927] 1.2989 1.1397 23.42
## I^2
## ich_sub = Intervention 0.0%
## ich_sub = Control 91.5%
##
## Test for subgroup differences (random effects model):
## Q d.f. p-value
## Between groups 1.18 1 0.2777
##
## Details on meta-analytical method:
## - Inverse variance method
## - DerSimonian-Laird estimator for tau^2
## - Jackson method for confidence interval of tau^2 and tau
## - Logit transformation
## - Clopper-Pearson confidence interval for individual studies
## - Continuity correction of 0.5 in studies with zero cell frequenciesMETA ANALYSIS OF PROP for Any ICH
# To visualize the results, you can plot a forest plot
meta::forest(meta_analysis_ich_2, layout = "JAMA")
# Load the necessary library
library(meta)
# Filter out rows with NA values in sich_sub and sich/sample_sich columns
meta_repu_details_filtered_sich <- subset(meta_repu_details, !is.na(sich_sub) & !is.na(sich) & !is.na(sample_sich))
# Calculate the pooled proportion using metaprop function
meta_analysis_sich_2 <- metaprop(
event = meta_repu_details_filtered_sich$sich, # Number of events (sich)
n = meta_repu_details_filtered_sich$sample_sich, # Total number (sample_sich)
data = meta_repu_details_filtered_sich, # Data source
sm = "PLO", # Specify summary measure as Proportion (Logit transformed)
method.tau = "DL", # DerSimonian-Laird estimator for tau^2
prediction = FALSE, # No prediction interval by default
comb.fixed = FALSE, # Random effects model
comb.random = TRUE, # Include random effects
studlab = meta_repu_details_filtered_sich$Study, # Label each study by its name
byvar = meta_repu_details_filtered_sich$sich_sub # Subgroup analysis by sich_sub
)
# Summary of the meta-analysis
summary(meta_analysis_sich_2)## proportion 95%-CI %W(random) sich_sub
## Richardt, 2024 i 0.0000 [0.0000; 0.2646] 2.5 Intervention
## Richardt, 2024 c 0.0471 [0.0130; 0.1161] 19.9 Control
## Béjot, 2023 i 0.0000 [0.0000; 0.1234] 2.6 Intervention
## Béjot, 2023 c 0.0311 [0.0156; 0.0549] 55.6 Control
## Leffert, 2016 0.0750 [0.0157; 0.2039] 14.5 Intervention
## Limaye, 2019 0.0000 [0.0000; 0.4096] 2.4 Intervention
## Murugappan, 2006 0.0000 [0.0000; 0.3694] 2.5 Intervention
##
## Number of studies: k = 7
## Number of observations: o = 534
## Number of events: e = 18
##
## proportion 95%-CI
## Random effects model 0.0393 [0.0255; 0.0602]
##
## Quantifying heterogeneity:
## tau^2 = 0 [0.0000; 0.2769]; tau = 0 [0.0000; 0.5263]
## I^2 = 0.0% [0.0%; 70.8%]; H = 1.00 [1.00; 1.85]
##
## Test of heterogeneity:
## Q d.f. p-value
## 2.59 6 0.8580
##
## Results for subgroups (random effects model):
## k proportion 95%-CI tau^2 tau Q I^2
## sich_sub = Intervention 5 0.0575 [0.0241; 0.1310] 0 0 1.08 0.0%
## sich_sub = Control 2 0.0347 [0.0210; 0.0567] 0 0 0.52 0.0%
##
## Test for subgroup differences (random effects model):
## Q d.f. p-value
## Between groups 0.99 1 0.3195
##
## Details on meta-analytical method:
## - Inverse variance method
## - DerSimonian-Laird estimator for tau^2
## - Jackson method for confidence interval of tau^2 and tau
## - Logit transformation
## - Clopper-Pearson confidence interval for individual studies
## - Continuity correction of 0.5 in studies with zero cell frequenciesMETA ANALYSIS OF PROP for sICH
# To visualize the results, you can plot a forest plot
meta::forest(meta_analysis_sich_2, layout = "JAMA")
Death
# Load the necessary library
library(meta)
# Filter out rows with NA values in death_sub and death/sample_death columns
meta_repu_details_filtered_death <- subset(meta_repu_details, !is.na(death_sub) & !is.na(death) & !is.na(Sample_death))
# Calculate the pooled proportion using metaprop function
meta_analysis_death_2 <- metaprop(
event = meta_repu_details_filtered_death$death, # Number of events (death)
n = meta_repu_details_filtered_death$Sample_death, # Total number (Sample_death)
data = meta_repu_details_filtered_death, # Data source
sm = "PLO", # Specify summary measure as Proportion (Logit transformed)
method.tau = "DL", # DerSimonian-Laird estimator for tau^2
prediction = FALSE, # No prediction interval by default
comb.fixed = FALSE, # Random effects model
comb.random = TRUE, # Include random effects
studlab = meta_repu_details_filtered_death$Study, # Label each study by its name
byvar = meta_repu_details_filtered_death$death_sub # Subgroup analysis by death_sub
)
# Summary of the meta-analysis
summary(meta_analysis_death_2)## proportion 95%-CI %W(random) death_sub
## Richardt, 2024 i 0.0833 [0.0021; 0.3848] 4.0 Intervention
## Richardt, 2024 c 0.0471 [0.0130; 0.1161] 16.8 Control
## Béjot, 2023 i 0.0000 [0.0000; 0.1234] 2.2 Intervention
## Béjot, 2023 c 0.0311 [0.0156; 0.0549] 46.9 Control
## Dicpinigaitis, 2021 c 0.0207 [0.0084; 0.0422] 30.2 Intervention
##
## Number of studies: k = 5
## Number of observations: o = 817
## Number of events: e = 23
##
## proportion 95%-CI
## Random effects model 0.0304 [0.0203; 0.0451]
##
## Quantifying heterogeneity:
## tau^2 = 0 [0.0000; 2.1150]; tau = 0 [0.0000; 1.4543]
## I^2 = 0.0% [0.0%; 79.2%]; H = 1.00 [1.00; 2.19]
##
## Test of heterogeneity:
## Q d.f. p-value
## 3.06 4 0.5479
##
## Results for subgroups (random effects model):
## k proportion 95%-CI tau^2 tau Q I^2
## death_sub = Intervention 3 0.0240 [0.0123; 0.0464] 0 0 1.78 0.0%
## death_sub = Control 2 0.0347 [0.0210; 0.0567] 0 0 0.52 0.0%
##
## Test for subgroup differences (random effects model):
## Q d.f. p-value
## Between groups 0.76 1 0.3842
##
## Details on meta-analytical method:
## - Inverse variance method
## - DerSimonian-Laird estimator for tau^2
## - Jackson method for confidence interval of tau^2 and tau
## - Logit transformation
## - Clopper-Pearson confidence interval for individual studies
## - Continuity correction of 0.5 in studies with zero cell frequencies# To visualize the results, you can plot a forest plot
meta::forest(meta_analysis_death_2, layout = "JAMA")
Proportions Good functional outcome (mRS 0–2) at 3 months
# Load the necessary library
library(meta)
# Filter out rows with NA values in mrs_sub and mrs/sample_mrs columns
meta_repu_details_filtered_mrs <- subset(meta_repu_details, !is.na(mrs_sub) & !is.na(mrs) & !is.na(sample_mrs))
# Calculate the pooled proportion using metaprop function
meta_analysis_mrs_3 <- metaprop(
event = meta_repu_details_filtered_mrs$mrs, # Number of events (mrs)
n = meta_repu_details_filtered_mrs$sample_mrs, # Total number (sample_mrs)
data = meta_repu_details_filtered_mrs, # Data source
sm = "PLO", # Specify summary measure as Proportion (Logit transformed)
method.tau = "DL", # DerSimonian-Laird estimator for tau^2
prediction = FALSE, # No prediction interval by default
comb.fixed = FALSE, # Random effects model
comb.random = TRUE, # Include random effects
studlab = meta_repu_details_filtered_mrs$Study, # Label each study by its name
byvar = meta_repu_details_filtered_mrs$mrs_sub # Subgroup analysis by mrs_sub
)
# Summary of the meta-analysis
summary(meta_analysis_mrs_3)## proportion 95%-CI %W(random) mrs_sub
## Richardt, 2024 i 0.6667 [0.3489; 0.9008] 13.8 Intervention
## Richardt, 2024 c 0.0353 [0.0073; 0.0997] 14.3 Control
## Dicpinigaitis, 2021 i 0.5000 [0.4247; 0.5753] 22.3 Intervention
## Dicpinigaitis, 2021 c 0.7222 [0.6507; 0.7863] 22.1 Control
## Leffert, 2016 0.6538 [0.6005; 0.7045] 22.7 Intervention
## Limaye, 2019 1.0000 [0.5904; 1.0000] 4.8 Intervention
##
## Number of studies: k = 6
## Number of observations: o = 802
## Number of events: e = 459
##
## proportion 95%-CI
## Random effects model 0.5270 [0.3558; 0.6921]
##
## Quantifying heterogeneity:
## tau^2 = 0.5524 [0.4385; 12.3038]; tau = 0.7433 [0.6622; 3.5077]
## I^2 = 92.2% [85.8%; 95.7%]; H = 3.59 [2.66; 4.84]
##
## Test of heterogeneity:
## Q d.f. p-value
## 64.35 5 < 0.0001
##
## Results for subgroups (random effects model):
## k proportion 95%-CI tau^2 tau Q
## mrs_sub = Intervention 4 0.6138 [0.4759; 0.7355] 0.1818 0.4264 14.16
## mrs_sub = Control 2 0.2425 [0.0049; 0.9543] 8.9026 2.9837 48.71
## I^2
## mrs_sub = Intervention 78.8%
## mrs_sub = Control 97.9%
##
## Test for subgroup differences (random effects model):
## Q d.f. p-value
## Between groups 0.56 1 0.4563
##
## Details on meta-analytical method:
## - Inverse variance method
## - DerSimonian-Laird estimator for tau^2
## - Jackson method for confidence interval of tau^2 and tau
## - Logit transformation
## - Clopper-Pearson confidence interval for individual studies
## - Continuity correction of 0.5 in studies with zero cell frequencies# To visualize the results, you can plot a forest plot
meta::forest(meta_analysis_mrs_3, layout = "JAMA")
As you can see, Dicpinigaitis doesn’t make sense. Please check that! :)