All but one included study [67] were published in peer-reviewed journals. Study designs consisted of prospective cohort (k = 5, 38%) [58,61,66,68,70], before–after (k = 3, 23%) [60,62,64], randomized control trial (k = 2, 15%) [63,67], non-comparative (k = 2, 15%) [59,69] and cross-sectional (k = 1, 8%) [65]. The studies were of variable methodological quality, ranging in overall score from 5 to 19 (maximum = 28) with a mean of 15.8 [standard deviation (SD) = 3.6] (Table 2). Studies with the lowest and highest scores demonstrated 18% and 68% in quality, respectively. Two studies were categorized as low quality [59,69]. The specific methodological problems common across almost all studies were adverse events not measured or reported (k = 13), no blinding of assessors (k = 13), no blinding of participants (k = 12), power calculation not performed or reported (k = 12) and unclear representativeness of study participants (k = 12).
