Most people in the West can't turn soy into equol, the compound that may be doing soy's best work. Twelve trials tried to change that with soy, probiotics, bran and seaweed. None of them managed it

Soy has a strange record in research. In studies that simply watch what people eat, mostly in East Asia, soy eaters tend to do well on hot flushes, heart health and memory. In randomised trials in Europe and the United States, where people are handed soy or a dummy, the results have been all over the place. One explanation has been around for more than twenty years. Soy contains a compound called daidzein, and certain gut bacteria turn it into another one, equol, which acts more strongly on the body's oestrogen receptors and is absorbed more easily. Not everyone has the right bacteria doing the right work. According to the figures the reviewers quote, up to 70% of East Asians produce equol, against 20 to 30% of people in Western countries. If equol is where much of the benefit comes from, most Westerners in a soy trial were never going to get it.

So the obvious next question is whether a non-producer can be turned into a producer. A team led from the University of Pittsburgh, with colleagues in Japan and elsewhere in the US, went looking for every randomised trial that measured people's equol status before and after trying something. Their review came out in Food Science & Nutrition on 27 September 2026. From 503 records they ended up with 12 trials: seven from the US, two from Australia and one each from the UK, Japan and Brazil. The review counts 730 people enrolled, 624 of them women. Six trials took only women past the menopause.

The trials tried most of the sensible ideas. Three simply gave soy isoflavones, for ten weeks, six months and two and a half years. Five added probiotic capsules or yoghurt, or a prebiotic fibre meant to feed helpful bacteria. Two used soy that had already been fermented. One added wheat bran, one seaweed. The shortest was a single serving of soy biscuits followed by 48 hours of urine collection. The answer was the same every time: no intervention turned non-producers into producers to a degree that statistics could tell from chance. The nearest thing to a success was a two-week Australian trial of fermented soy milk in which two of eight women became producers, and that wasn't enough to count either. A few trials saw people who already made equol make somewhat more of it, on probiotics or seaweed, but that is a different thing from switching it on.

The reviewers' reading of the probiotic trials is the most useful paragraph in the paper. The capsules were mostly Lactobacillus acidophilus and Bifidobacterium longum, the ordinary shop-shelf strains. Those bacteria can do the first step, snipping a sugar off soy's isoflavones so that daidzein is released. They cannot do the next steps, the ones that turn daidzein into equol. That job belongs to a small group of oxygen-shy specialists with names like Slackia isoflavoniconvertens and Adlercreutzia equolifaciens, and no trial gave anybody those. Even having them may not be enough: studies the review cites have found equol-making bacteria in producers and non-producers alike, and stable producers tend to have a more varied gut community all round. The reviewers think the wider neighbourhood decides whether the specialists get to work.

They are also frank about how weak the twelve trials are. None was rated good quality; seven were fair and five poor. All but one skipped the calculation of how many people would be needed to see an effect. Five lost more than a fifth of their volunteers, and one lost 55%. Only four used the standard test of feeding everyone soy for three days before measuring. So the authors refuse to call the question closed. 'These negative findings should not be interpreted as evidence that conversion is biologically impossible,' they write, and they point to studies without a control group in which it seemed to happen, including one where eight of 20 non-producers were making equol after 16 weeks of drinking a litre of soy milk a day.

The rest is our reading. The best finding in the review cuts against that hopeful note. In the longest trial, 350 older women took soy isoflavones or a placebo for two and a half years, and some women's equol status flipped back and forth, on the placebo as well. A Japanese study the review cites found that about a quarter of 498 adults did not hold the same status across four years of annual tests. If people drift in and out by themselves, then a study with no comparison group will always find some who 'converted', and the litre-a-day result can't be told apart from that drift. The same goes for two women out of eight. The dose-and-time argument also has a hole in it, since the two longest randomised trials gave soy alone for six months and for two and a half years and saw nothing. And one step further back, the case that equol is the active ingredient rests mostly on comparing producers with non-producers in East Asia. The reviewers themselves note that producers may simply be people with healthier, more varied gut bacteria, in which case equol would be the sign of a good gut and making more of it would not buy the benefit. A US trial that gives equol itself to older adults is under way, and the review's senior author is first author of the paper describing its design. That trial, more than another probiotic study, is the one that could say whether any of this matters. The review was funded by the US National Institutes of Health and the authors declare no conflicts of interest.