Data extraction and quality assessment
All publication pieces of data were extracted: (1) publication information (the first author’s name, publication year), (2) cohort demographics (sample size, gender constituent ratio, mean age, comorbidity), (3) atrial fibrillation (AF) duration, (4) body mass index (BMI), (5) echocardiographic parameters (left ventricular ejection fraction (LVEF), (6) left atrial dimension (LAD), (7) procedural data, including cryoballoon generation (first generation: CB-1, second generation (advanced): CB-2, third generation (short tip): CB-ST), balloon dimension (23 or 28mm), balloon location (right superior pulmonary vein right [RSPV] or inferior pulmonary vein [RIPV]), time and nadir temperature to rPNI, diaphragmatic protection (fluoroscopy, palpation, stimulation, including cycle, strength of impulse, positioning of pacing electrode and CMAP), procedural and delayed complications of phrenic nerve palsy. Whenever the data of interest were incomplete, the investigators did not use the study in the meta-analysis. In some publications, only a selected group was used for analysis once the same data were found in other publications (example Aryana 2014 [34], 2015 [6], 2016 [44]; 2014: CB-1/CB-2 2014: only CB-1 taken from a database; 2015: CB-2 2012-2015.02 CB-2 all taken from database; 2016: CB-2/CB-ST from 2015.3: all data taken from database).