Clostridium difficile-Associated Diarrhea and PPI Therapy
Clostridium difficile-Associated Diarrhea and PPI Therapy
We included in this meta-analysis the studies reporting different risk estimates: case–control studies (ORs), and cohort studies (RRs). In practice, these measures of effect would be expected to yield very similar RR estimates, given that the absolute risk of CDAD is low. However, all ORs in this study were converted to RRs using the formula RR=OR/(1−OR)+(Po×OR), where Po is the risk of the outcome/event in the control group.
All studies were initially analyzed together and were subsequently grouped on the basis of study design (cohort vs. case–control). This was done to examine consistency of results, and in an attempt to explain the presence of heterogeneity between all the studies. We used the DerSimonian-Laird method, which assumes a random-effects model to calculate the pooled effect estimate. In the presence of heterogeneity, the random-effects model is recommended by the Cochrane collaboration, because its assumptions account for the presence of variability among studies. Therefore, reported estimates are from a random-effects model. We used Forest plots to summarize the results. Publication bias was assessed using the Begg's, and Egger's regression asymmetry test. Furthermore, Duval and Tweedie's nonparametric "trim-and-fill" rank-based technique, which formalizes the use of funnel plots and estimates, and adjusts for the number of missing studies, as well as the outcomes of the missing studies, was performed. A comprehensive meta-analysis was done to depict the effect of large and small sample–sized studies. Funnel plots were also visually inspected.
To assess statistical heterogeneity between studies, we used the Cochrane's Q-statistic and the I-statistic. An I value of >50% or a P value <0.05 for the Q-statistic was taken to indicate significant heterogeneity. All statistical tests were two-tailed, and a probability level of <0.05 was considered significant. Results are presented in accordance with the guidelines proposed by MOOSE. All analyses were done using STATA 11.1 (STATA, College Station, TX) statistical software and Comprehensive meta-analysis software version 2 (Biostat, Englewood, NJ).
Statistical Analysis
We included in this meta-analysis the studies reporting different risk estimates: case–control studies (ORs), and cohort studies (RRs). In practice, these measures of effect would be expected to yield very similar RR estimates, given that the absolute risk of CDAD is low. However, all ORs in this study were converted to RRs using the formula RR=OR/(1−OR)+(Po×OR), where Po is the risk of the outcome/event in the control group.
All studies were initially analyzed together and were subsequently grouped on the basis of study design (cohort vs. case–control). This was done to examine consistency of results, and in an attempt to explain the presence of heterogeneity between all the studies. We used the DerSimonian-Laird method, which assumes a random-effects model to calculate the pooled effect estimate. In the presence of heterogeneity, the random-effects model is recommended by the Cochrane collaboration, because its assumptions account for the presence of variability among studies. Therefore, reported estimates are from a random-effects model. We used Forest plots to summarize the results. Publication bias was assessed using the Begg's, and Egger's regression asymmetry test. Furthermore, Duval and Tweedie's nonparametric "trim-and-fill" rank-based technique, which formalizes the use of funnel plots and estimates, and adjusts for the number of missing studies, as well as the outcomes of the missing studies, was performed. A comprehensive meta-analysis was done to depict the effect of large and small sample–sized studies. Funnel plots were also visually inspected.
To assess statistical heterogeneity between studies, we used the Cochrane's Q-statistic and the I-statistic. An I value of >50% or a P value <0.05 for the Q-statistic was taken to indicate significant heterogeneity. All statistical tests were two-tailed, and a probability level of <0.05 was considered significant. Results are presented in accordance with the guidelines proposed by MOOSE. All analyses were done using STATA 11.1 (STATA, College Station, TX) statistical software and Comprehensive meta-analysis software version 2 (Biostat, Englewood, NJ).
Source...