Eight randomized trials gave 329 men soy and measured their testosterone. None of it moved — not at low doses, not at high doses, not in any subgroup the researchers checked.
The controversy that fills TikTok and Reddit and the gym floor is not about this hormone. It is about a different one entirely.
Researchers pooled every randomized trial in which healthy men consumed soy and had their testosterone measured. Eight trials. Three hundred and twenty-nine men. Five countries. Eighteen years of data.
Testosterone did not move. The statistical difference was essentially zero, with no sign that soy pulled the number in either direction. The disagreement between the eight independent research teams was also zero — not low, not trending, zero.
Then came the analysis previous reviews never ran: a dose-response curve. Three different mathematical shapes were tested across doses ranging from 49 to 164 milligrams of isoflavones per day. All three came back flat. Free testosterone behaved the same way. There is no dose at which the effect starts — not within anything that has been tested.
Same Shape, Different Lock
If the testosterone answer is that clean, the question shifts: why does half the internet believe the opposite?
The belief rests on a molecular resemblance. Phytoestrogens — the compounds in soy — look similar to estradiol under a microscope. That structural similarity is real. But structure is not function. Isoflavones bind preferentially to estrogen receptor beta, found in the prostate, lungs, and colon — not estrogen receptor alpha, the one active in breast tissue and the uterus.
Same key shape, different lock. The premise that props up the myth has been identifiably wrong since the receptor subtypes were mapped. And the trials confirmed it: the hormone everybody argues about did not respond.
The controversy, it turns out, was never really about testosterone.
The One Number Worth Carrying
Estradiol — the hormone the trials actually flagged — rose with dose. It peaked near 72 milligrams of isoflavones a day — a rise of roughly 6.5 picograms per millilitre above baseline. That peak sits above what traditional soy foods typically deliver: habitual intake in Japan, where soy is a staple rather than a trend, runs 30 to 50 milligrams a day.
The shift itself is small. A normal male estradiol range runs 10 to 40 picograms per millilitre. The measured peak falls within the gap between two men who would both be told their bloodwork is normal.
But this finding rests on only four trials. Pull one out and the direction reverses. Rajaie’s team, who found the curve, then questioned it themselves: most comparison groups drank whey or casein rather than a true placebo, and those men’s estradiol fell on its own.
Part of what reads as soy raising a hormone may be the comparator lowering it.
Three Litres a Day
The scariest version of this story is a man who developed breast tissue after drinking roughly three litres of soy milk every day. The case is real. It is cited in nearly every Reddit thread and TikTok comment section that argues soy is dangerous.
It involved an intake of 310 to 361 milligrams of isoflavones per day — five to nine times what the average Japanese person consumes, and roughly double the highest dose any controlled trial has tested.
The case reports do not contradict the trial evidence. They sit so far beyond any food-based diet that they confirm the dose threshold rather than undermining it. The fear is real at one dose. The evidence is clear at another. They are not the same dose.
The Muscle Question Nobody Tested on Purpose
The question underneath the hormone question was always practical: does soy cost you muscle?
Seventeen head-to-head randomized trials compared soy protein against milk protein for muscle mass. The measured difference was 0.02 — smaller than what the body-composition scanners used to measure it can reliably resolve. No divergence by age. No divergence by sex.
A separate set of nine resistance-training trials — whose lead author discloses a soy-industry affiliation — found the same verdict for lean mass, bench press, and squat.
Two research fields that were never designed to answer each other looked at soy from opposite ends. The hormone panels found nothing. The muscle trials found nothing. Neither field knew it was closing the other’s blind spot.
Choose your protein source on taste, digestion, cost, or ethics. The evidence says the muscle outcome is the same.
One question the muscle evidence opens but does not close: the parity holds when total protein is matched. Matched at what level? That number exists, and it sits in a different analysis entirely.
Where the Evidence Stops
Three boundaries are worth naming plainly. No trial in this evidence base ran longer than twelve weeks. No trial tested above 164 milligrams of isoflavones per day in a controlled setting. And most participants were under fifty.
If you are over fifty or have a hormonal condition, these trials were not designed to include you — the source paper says so directly. That is a conversation for whoever has your bloodwork in front of them.
For everyone else: soy did not lower testosterone in healthy men at any dose the trials tested, the one hormone that moved sat above a dose that traditional soy foods do not reach, and the muscle it was supposed to cost matched milk protein across seventeen head-to-head comparisons.
A block of tofu, a bowl of edamame, a daily soy shake — none of that gets you near the dose where estradiol moved in these trials. Traditional soy foods deliver roughly 3.5 milligrams of isoflavones per gram of soy protein, and habitual intake in Japan — where soy is a staple, not a trend — runs 30 to 50 milligrams a day.
The estradiol threshold sat around 72 milligrams a day. The case reports that fuel the fear were at 310 to 361 — men drinking litres of soy milk daily, not eating tofu with dinner.
If your soy comes from food, you are not in the territory the trials flagged. If it comes from a concentrated isoflavone supplement, the trials tested up to 164 milligrams a day and found testosterone unmoved — but above that, you are off the map. Not in territory the research found safe. In territory it never visited.