Meta-Analysis on Postoperative Atelectasis • dcv

library(metafor)

Input data

dat.atelectasis <- escalc(measure="RR",
              ai=c( 3,  2,  4,  1,  0),
              bi=c(15, 33, 40, 44, 15),
              ci=c( 4,  6,  6,  2,  1),
              di=c(15, 29, 36, 43, 14),
              slab=c("Boules et al, 2011",
                     "Mahmoud et al, 2017",
                     "Li et al (Standard), 2020",
                     "Li et al (OLA), 2020",
                     "Ammar et al, 2021")
              )

Primary analysis: Paul-Mandel without Hartung-Knapp-Sidik-Jonkman (HKSJ) modification

pm.atelectasis <- rma(yi, vi, data=dat.atelectasis, method="PM") 
forest(pm.atelectasis,
       atransf=exp,
       at=log(c(0.01, 0.1, 0.5, 1, 4, 10)))

Before Li et al. correction, these results agreed with the paper (Fig 2, bottom).

Sensitivity analysis

Sensitivity analysis 1: FE model

fe.atelectasis <- rma(yi, vi, data=dat.atelectasis, method="FE") 
forest(fe.atelectasis, 
       atransf=exp,
       at=log(c(0.01, 0.1, 0.5, 1, 4, 10)))

Note: The standard RE model without HKSJ modification is identical (not shown):

re.atelectasis <- rma(yi, vi, data=dat.atelectasis, method="REML")

Sensitivity analysis 2: Paule-Mandel with HKSJ modification

pmhk.atelectasis <- rma(yi, vi, data=dat.atelectasis, method="PM", test="knha") 
forest(pmhk.atelectasis,
       atransf=exp,
       at=log(c(0.01, 0.1, 0.5, 1, 4, 10)))

Sensitivity analysis 3: Odds ratios (Paule-Mandel without HKSJ modification)

dat.atelectasis.OR <- escalc(measure="OR",
              ai=c(3, 2, 1, 1, 0),
              bi=c(15, 33, 43, 44, 15),
              ci=c(4, 6, 4, 2, 1),
              di=c(15, 29, 38, 43, 14),
              slab=c("Boules et al, 2011",
                     "Mahmoud et al, 2017",
                     "Li et al (Standard), 2020",
                     "Li et al (OLA), 2020",
                     "Ammar et al, 2021")
              )


pm.OR.atelectasis <- rma(yi, vi, data=dat.atelectasis.OR, method="PM") 
forest(pm.OR.atelectasis,
       atransf=exp,
       at=log(c(0.01, 0.1, 0.5, 1, 4, 10)))

Results are similar to primary analysis results.

Sensitivity analysis 4: leave-one-out analysis

Atelectasis

leave1.atel <- leave1out(pm.atelectasis)
forest.default(x=leave1.atel$estimate, ci.lb=leave1.atel$ci.lb, ci.ub=leave1.atel$ci.ub,
               slab=paste("Omitting ", leave1.atel$slab, sep=""),atransf=exp,
               at=log(c(0.15, 0.25, exp(pm.atelectasis$b), 1, 4)), xlim=log(c(0.01, 4.5)),
               refline=pm.atelectasis$b
              )
lines(x=log(c(1, 1)), y=c(0, 6))

Results are almost identical to results in paper (again before Li et al. correction).

Funnel plot

funnel(pm.atelectasis, main="Standard Error")

regtest(pm.atelectasis)
#> 
#> Regression Test for Funnel Plot Asymmetry
#> 
#> Model:     mixed-effects meta-regression model
#> Predictor: standard error
#> 
#> Test for Funnel Plot Asymmetry: z = -0.4689, p = 0.6392
#> Limit Estimate (as sei -> 0):   b = -0.0647 (CI: -2.3644, 2.2349)