Page 28 - Journal of Structural Heart Disease Volume 4, Issue 5

223
Original Scientific Article
Methods
We searched PubMed and Cochrane Central Reg- ister of Controlled Trials for trials comparing PFO closure to medical therapy from inception through October, 2017. Only studies in the English language or studies with an English translation were included. Citations were screened at the title/abstract level and relevant citations were retrieved as full reports. Refer- ences from the included studies were also manually searched for relevant studies.
Studies were eligible for inclusion if they were randomized controlled trials that compared PFO clo- sure to medical therapy in patients with cryptogen- ic stroke and PFO. If the medical therapy arm in any study included patients on antiplatelets and/or oral anticoagulation, the study was included only if recur- rent stroke was reported separately for each group of patients. Studies were excluded if they were non randomized trials or if outcomes of patients on anti- platelets and patients on oral anticoagulation were not reported separately. Moreover, patients who re- ceived PFO closure plus anticoagulation and patients who did not receive any antithrombotic therapy in any of the included studies were excluded from the final analysis.
The outcome of the present study was recurrent strokes at the longest follow up period reported in each study. In the CLOSURE I trial, [6] the outcome included was recurrent strokes or transient ischemic attacks. Data were independently extracted from the
Table 1. Characteristics of included trials.
CLOSE [5] 44 42 CLOSURE I [6] 46 48
REDUCE [7] 45 40 RESPECT [8] 46 45
included trials by the first and second authors (G.M. and D.S.) on a pre-specified data sheet. Any discrep- ancy was discussed until there was complete agree- ment on all the results in the final data sheet.
Network meta-analysis was performed using a Bayesian Markov chain Monte-Carlo model. [10] Di- chotomous outcome variables were compared with odds ratios (OR) and 95% credible intervals (CrI). The more conservative random effect model was adopted for final interpretation of the results. Vague (non-in- formative) priors for between-study heterogeneity were applied to the random effects analyses. Analy- ses using the fixed effect model was also performed and was only shown in the forest plot diagram. Three chains with different starting variables were used. To achieve convergence, a burn-in phase of 10,000 simulations was performed then 20,000 simulations were performed for the final analyses. Convergence was confirmed by assessing whether the Monte Car- lo error is less than 5% of the standard deviation of the effect estimates or between study variance and by visual inspection of Gelman Rubin graphs. [11, 12] The heterogeneity between trials was determined from the median between-trial variance τ2. A τ2 es- timate of 0.40 was interpreted as a high degree of heterogeneity. [13] Consistency between direct and indirect evidence was assessed by plotting the pos- terior mean deviance of the individual data points in the inconsistency model against their posterior mean deviance in the consistency model. Consistency was suggested when each data point had a posterior
Trial name
Mean age (years)
Female (%)
PFO closure device
Medical therapy
Follow up duration
Antiplatelets
Oral Anticoagulation
All available devices
STARFlex Septal Closure System
HELEX and Car- dioform Septal Occluders
Amplatzer PFO Occluder
aspirin, clopido- grel or aspirin/ dipyridamole
Aspirin
aspirin, clopido- grel or aspirin/ dipyridamole
aspirin, clopido- grel or aspirin/ dipyridamole
Coumadin or direct oral anti- coagulants
Coumadin N/A
Coumadin
5.4 years
2 years 3.2 years
5.9 years
Mina G. et al.
Meta-Analysis of PFO Closure for Stroke Prevention