Probiotics do not significantly reduce nosocomial pneumonia 英文参考文献.docVIP

Silvestrietal.CriticalCare2012,16:453 /content/16/6/453 LETTER OpenAccess Probioticsdonotsignificantlyreducenosocomial pneumonia LucianoSilvestri1*,HendrickKFvanSaene2andDarioGregori3 SeerelatedresearchbyLiuetal. ,/content/16/3/R109 Liuandcolleagues, intheir recent meta-analysis, con- Silvestriandcolleaguesnotedthatamoreconservative cludedthatprobioticswereassociatedwithastatistically random-effects model should be used to analyze the significant reduction in the incidence of nosocomial pooled incidence of NP. The reason why we used a pneumonia (NP) (odds ratio = 0.75; 95% confidence fixed-effect model in our analysis can be explained by interval=0.57to0.97,P=0.03)[1].Theseresultswere the lack of statistical power. The confidence intervals obtainedusingthefixed-effectmodel,ignoringamoder- tend to be wider especially when implementing a ran- ate level ofheterogeneity (I2 =46%) [2]. However, we dom-effects model, and the statistical significance is believethattheinclusioncriteria,subjects,andinterven- morelikelytobelost.Regardingthepotentialmisinter- tions were different among the included studies, and pretation among the special patient population in our might have impacted theresults. Good practice would reviewthatisdifferentfromotherstudies[4,6],wepre- thereforebetochooseamoreconservativeanalysis;that defined heterogeneity as low, moderate orhigh with I2 is, the random-effects model. Using this model, the valuesabove25%,50%,and75%,respectively,asrecom- claimedreductioninNPwouldhavebeennotsignificant mended by some authors [7]. We calculated all pooled (oddsratio=0.70,95%confidenceinterval=0.46to1.05, oddsratios based onthis criterion toincrease thecon- P=0.085) (Figure 1). Wehave already addressed this sistencyofourmeta-analysis. issuewithreferencetoapreviousmeta-analysisonpro- biotics[3,4]. Silvestri and colleagues are also concerned about the switchbetweendifferentmodels.However,ourpopula- Remarkably, the authors use