DMSO: miracle cure or alternative-medicine scam?

A separate query: 219 papers on DMSO safety in humans (toxicity, adverse reactions, poisoning, deaths, haemolysis) – read for this paragraph, not counted in YES/NO.

Acute toxicity is low – but "low" is not "none". On the skin: burning, erythema, scaling, urticaria; in the scleroderma trial one in four people in the 70% group withdrew because of their skin. Everyone gets garlic breath (the metabolite, dimethyl sulfide). Intravenously: dose-dependent haemolysis (red blood cells breaking down) – haemoglobin in the urine, with no kidney damage in the JAMA study (1980), but in two elderly people infused for arthritis – anaemia and raised liver enzymes (Lancet 1980). Most serious events are described with infusions of stem cells frozen in DMSO: arrhythmias, seizures, strokes, transient amnesia, isolated deaths – which is why transplant centres lower the concentration and wash DMSO out before infusion. Drinking it on your own: two reports of brain haemorrhage (2004 – death of a patient with metastases after a single dose, 2021 – recurrent, in a man treating his own low mood) – DMSO inhibits platelets. It passes into breast milk (2026). In animals it potentiated bleomycin toxicity; in humans it blocked the activation of sulindac (an anti-inflammatory drug) and there is a case of severe neuropathy when the two were combined. The lens damage that stopped trials in the 1960s (dogs, rabbits) was not found in humans in three trials with full eye examinations. And the most important rule: DMSO carries through the skin anything you mix with it or that is lying on it – "technical grade" DMSO with impurities is not the pharmacy drug.

How to read it: bar length = number of papers, colour = level of evidence from sage (strongest: randomised clinical trials in humans) to brown (weakest: cells in a dish). An answer with a lot of sage and blue is backed by human studies; an answer with only ochre and grey rests on animals, cells and reviews of other people's work.

How to read it: each dot is one paper, the further right the higher the journal ranks in SJR (SCImago). SJR counts not only how many citations a journal receives but also where they come from: a citation from Cell weighs more than one from a small journal, so sheer volume of papers does not inflate it (which is why megajournals that rank high on the h-index drop here). Log scale: the ticks are 0, 1, 3, 10. The vertical line = the group median. Hover over a dot to see the title. Papers in journals outside Scopus (e.g. Cureus, preprints) have no SJR and are not shown.

How to read it: each dot is one paper. RCR (Relative Citation Ratio, NIH iCite) says how many times more often a paper is cited than a typical paper in its field, per year: 1 = average, 2 = twice as often, 0.5 = half as often. It evens out the paper's age and differences between fields, so an old paper does not win just because it is old. Log scale, line = group median. Papers from the last ~2 years have only a provisional RCR or none – they are not on the chart, and provisional values are marked "~" in the table. Raw citation counts are in the table and on hover.

How to read it: the same question as with SJR, an older and simpler measure: each dot is one paper, the further right the more established the journal (OpenAlex h-index: how many of the journal's articles have at least that many citations). The vertical line = the group median, i.e. the typical journal for that answer. Hover over a dot to see the title. Caveat: the h-index favours big, old, broad titles (Science has 1750, a good specialist quarterly 60–250), and a commentary in Science is still just a commentary – the strength of evidence is shown by the first chart, not this one.

How to read it: one bar = one publication year, split into YES, NO and MIXED ("not applicable" papers left out). It shows whether the topic is accelerating and whether new papers pull in one direction or the dispute is sharpening. Hover over a segment to see the count.

Sorted by: stance → level of evidence → citations. "Not applicable" papers at the end. Links go to PubMed, DOI and, where available, the open-access full text.