Posted on 24-07-2012
Filed Under (ED Treatment) by admin

If relevant data (e.g. standard deviations) were not reported adequately, we attempted to calculate the needed parameters. Trials that did not report complete numerical information for relevant efficacy/harms outcomes (i.e., arm-specific mean endpoint or change in score, standard deviation, or standard error, proportion of patients with an outcome at followup) could not be incorporated in the meta-analyses. Trial reports presenting measures of variability (e.g. standard deviation) only graphically (i.e., no numerical data were available) were not pooled. Online Mexican Pharmacy

We calculated standard deviations from standard errors or 95 percent confidence intervals.

For continuous outcomes (e.g. mean endpoint/change in the total score of IIEF), the absolute difference between treatment-specific means and corresponding standard deviations were ascertained for each individual study. A generic inverse variance method was used to calculate the response outcomes and corresponding 95 percent confidence intervals for the combined treatment groups.

For dichotomous outcomes (e.g. improvement in erection GAQ), studies were grouped by type of treatment and dose to minimize clinical heterogeneity. The intent-to-treat group or number enrolled at the time of study was used for analyses and, when this information was unavailable, we used the number provided in the report.

Pooled relative risks with corresponding 95 percent confidence intervals were generated.

The DerSimonian and Laird random-effects model was used to obtain combined estimates across the studies. The degree of statistical heterogeneity was evaluated by using a chi-square test and the I2 statistic. An I2 of less than 25 percent is consistent with low heterogeneity; 25 to 50 percent with moderate heterogeneity; and over 50 percent with high heterogeneity. When statistically significant heterogeneity was identified, it was explored through subgroup and sensitivity analyses when appropriate. Sources of heterogeneity include reporting and methodological quality (e.g. methods for randomization, adequacy of allocation concealment, blinding, washout period for crossover trials, data analysis) as well as clinical heterogeneity (e.g. study population, dosing of therapeutic agent, duration of followup). Estimates from the heterogeneous groups must be interpreted with caution, especially when small numbers of trials are included.

We also performed a series of subgroup analyses to explore the consistency of the results.

The meta-analyses are presented as forest plots. Publication bias was explored through funnel plots (Figures D1-16, Appendix D) by plotting the relative measures of effect (relative risk) versus a measure of precision of the estimate (1/standard error). The visual asymmetry in funnel plots maybe be suggestive of publication bias, although other potential causes for asymmetry exist. The degree of funnel plot asymmetry was measured using the Egger regression test.

(0) Comments   

Comments are closed.