Risk of bias within studies
In this review, we included 76 cohort studies, of which 64 were prospective 2, 4, 6, 7, 20, 22, 24– 27, 29, 31, 32, 34– 38, 40– 45, 48, 50– 54, 56, 58, 60, 63, 65, 66, 68– 70, 72, 73, 77, 80, 81, 83, 85, 88, 90, 92, 94, 97, 98, 101– 104, 107, 108, 110, 112, 115, 117, 118 and 13 were retrospective 9, 18, 23, 39, 46, 47, 61, 71, 74, 86, 87, 100, 105. Bias was assessed using The Newcastle-Ottawa-Scale 14. Using this scale, studies were given zero to nine stars. A high number of stars equals low risk of bias and vice versa. The studies in this review had a median value of 5 stars, with a range of 2–8. No studies received the highest possible value of nine stars. Very few studies had a comparison group that did not receive DMSO, and often the occurrence of adverse reactions was poorly described. There were 24 randomized controlled trials ( Figure 2). Many studies received an unclear risk of bias because often it was vaguely described how adverse reactions were reported.
Figure 2.  Risk of bias in randomized controlled trials. Overall, there was a high risk of bias when assessing the description of adverse reactions. Some studies were not assessed for bias due to being case-reports, preliminary trials, or because they included more than one study design 17, 19, 28, 62, 84, 91, 99, 109, 111, 113.