Methods

Protocol and eligibility criteria
Our study-protocol is registered at PROSPERO (Registration number: CRD42018096117). The systematic review was performed according to PRISMA-harms (Preferred Reporting Items for Systematic Reviews and Meta-Analyses) guidelines 13.
No limitations were set on the date of publication. The language was restricted to English, Danish, Swedish, Norwegian, and Russian. We included all original studies that administered DMSO to humans and included five or more participants. There was no gender or age restriction. For a study to be included, the authors had to suspect that an observed adverse reaction could be caused by DMSO.

Primary outcome
The primary outcome was any adverse reaction seen in relation to the use of DMSO in humans.

Literature search
The search was performed in PubMed (1966-present), EMBASE (1980-present), and the Cochrane Library. The databases were last searched on February 23, 2018. Our search strategy was formulated with the help of a medical research librarian.
The search string used in PubMed was: ((dimethyl sulfoxide) OR DMSO) AND ((((((administration and dosage) OR adverse reactions) OR alternate effects) OR secondary response) OR toxicology) OR side effects)). The search was restricted to humans. The search string was adapted to EMBASE and Cochrane Library using the same search-words as abovementioned.
The search string used in EMBASE was: ((dmso or dimethyl sulfoxide) and ((side effect or toxicology or secondary response or alternate effects or alternate reactions or (administration and dosage)) and (dmso or dimethyl sulfoxide))).mp. The search was restricted to humans, articles and Medline journals were excluded.
The search string used in Cochrane was: (adverse drug events and dimethyl sulfoxide). The search was restricted to trials.

Study selection and data extraction
Two authors (B.K.M. and D.Z.) independently screened title and abstract according to the eligibility criteria using www.covidence.org. Discrepancies were resolved by discussion. One author screened the full-text articles (B.K.M.). Russian articles were screened by an author fluent in Russian (M.H.). If M.H. was in doubt regarding inclusion of a study the results were presented to B.K.M. and then discussed until a mutual decision was made.
After the screening process was finished, all included studies were imported to an Excel sheet (Microsoft Excel 2016). Data extraction was performed by two authors (M.H. extracted from the Russian articles and B.K.M. extracted from the rest). Data extracted were: author, publication year, country, study characteristics (study design, sample size, size of comparison group if present, time to follow-up), use of DMSO (reason for use, treatment duration, administration route, dose of DMSO), and adverse reactions observed (number of persons experiencing an adverse reaction, method of assessing, and duration of adverse reaction).

Analysis
The Newcastle-Ottawa-Scale was used to assess the risk of bias in non-randomized observational studies 14. Risk of bias in randomized controlled trials was assessed using the Cochrane Handbook “Risk of Bias” assessment tool 15. Risk of bias was assessed at the outcome level.
The primary summary measure was percentage of persons experiencing an adverse reaction, as well as the range in which a reaction occurred in the studies included. No meta-analysis and further summery measures were planned due to the expected large heterogeneity of the studies.