Kudzu Cut Binge Drinking 27 Percent. Eight Weeks Earlier, a Second Trial Found It Lost to Placebo.

Two double-blind, placebo-controlled trials of the same herbal extract published eight weeks apart reached conclusions that cannot both be summarized in one sentence.

The first, out of McLean Hospital and Harvard Medical School, ran for eight weeks in nineteen men and women being treated for alcohol use disorder. Participants took 750 milligrams of kudzu isoflavones every day. By the end of the trial, the placebo group had cut its heavy drinking days by 44 percent. The kudzu group had cut its by 30 percent. Placebo won.

The second, published Nov. 4 in the American Journal of Psychiatry by a UC San Francisco and San Francisco Department of Public Health team, enrolled 120 adults with alcohol use disorder who binge drink. They took 2 grams of kudzu extract only on days when they expected to drink. Over twelve weeks of treatment plus weekly counseling, the kudzu arm reported 27 percent fewer binge-drinking days than the placebo arm, an incidence rate ratio of 0.73 with a confidence interval of 0.54 to 0.99. The reduction was still visible three months after the pills stopped.

Same plant. Same design family. Opposite headlines. The variable that separates them is not the herb. It is the clock.

A pill for the nights you plan to drink

The San Francisco trial, led by Glenn-Milo Santos with Phillip Coffin and colleagues from the city health department's Center on Substance Use and Health, tested what the authors call targeted use. Participants were told to take two grams of oral kudzu extract on days when they anticipated binge drinking, rather than on a fixed daily schedule. Everyone also received weekly counseling for twelve weeks.

That combination matters for how the result should be read. Kudzu was not tested against doing nothing. It was tested as an addition to a talking intervention that every participant received, placebo arm included. The primary outcome was the number of binge-drinking days in the past week, self-reported. Prespecified secondary outcomes included the number of binge-drinking weeks, total drinks, and two alcohol biomarkers: ethyl glucuronide, which clears in days, and phosphatidylethanol, which accumulates in red blood cells and reflects drinking over roughly the previous two to four weeks.

Retention was unusually good for an alcohol trial: 88 percent completed, and 86 percent of weekly follow-up visits were completed. That is not a footnote. Attrition is the mechanism by which addiction trials quietly produce false positives and false negatives at the same time, and a trial that keeps nearly nine in ten participants through three months has earned more trust than its sample size alone suggests.

Outcome	Intention-to-treat	As-treated (2.7 or more dosing days per week)
Binge-drinking days, past week	IRR 0.73 (0.54–0.99)	IRR 0.58 (0.38–0.89)
Binge-drinking weeks	RR 0.79 (0.62–0.99)	RR 0.68 (0.47–0.98)
PEth positivity	not significant	RR 0.82 (0.67–0.99)

Adverse events were similar between arms, and no serious event was attributed to kudzu.

The trial that found nothing

The McLean group's paper, published Sept. 4 in Drug and Alcohol Dependence Reports, is the third in a line of kudzu studies from that lab, and it was designed differently on nearly every axis. Nineteen participants, twelve men and seven women aged 27 to 65, all seeking treatment. Daily dosing of 750 milligrams of isoflavones for eight weeks instead of as-needed 2-gram doses. Twice-weekly laboratory visits. Adherence verified with urinary riboflavin rather than self-report. Baseline heavy drinking days were 4.2 per week in the kudzu arm and 4.7 in the placebo arm.

Both arms fell by about 40 percent within four weeks. Across the full eight weeks, the average reduction was 30 percent for kudzu and 44 percent for placebo. Adherence exceeded 96 percent and there were no safety signals, which rules out the two most common explanations for a failed trial in this literature: people not taking the drug, and the drug being unsafe to take.

The authors' own conclusion is more careful than the headlines it invited. Kudzu may help reduce binge drinking, they wrote, but may not provide additional benefit to people already enrolled in an effective alcohol treatment program.

That sentence is the hinge between the two papers. The McLean trial measured kudzu against a backdrop of weekly counseling in a sample that had already sought care. The San Francisco trial measured the same backdrop in a sample recruited for binge drinking specifically, and dosed the drug around the behavior rather than around the calendar.

Why nineteen people and one hundred twenty do not actually disagree

With nineteen participants split roughly nine and ten, the McLean trial had almost no statistical room to detect a moderate effect. A 27 percent reduction in binge days, the size the San Francisco trial found, would have been indistinguishable from noise at that sample size. The 44-percent-versus-30-percent gap it reported is compatible with kudzu working modestly, doing nothing, or helping a subgroup that the trial was never powered to isolate.

What the McLean trial can say is narrower and still useful: in a small treatment-seeking sample on daily dosing, kudzu did not separate from placebo. What the San Francisco trial can say is broader but not final: in a larger sample on targeted dosing, it did.

What kudzu is supposed to be doing

Kudzu root, Pueraria lobata, carries a set of isoflavones — puerarin, daidzin, daidzein — and the most-cited mechanistic story in this literature is enzymatic. Daidzin inhibits aldehyde dehydrogenase, the enzyme that clears acetaldehyde, the first and most unpleasant metabolite of ethanol. Interfere with that clearance and the subjective experience of a drink changes. There is also a body of animal work pointing at the GABA-benzodiazepine receptor complex, which is a different mechanism with a different implication, and the fact that two plausible mechanisms coexist after decades of study is itself informative about how thin the human evidence base is.

Two grams of what, exactly

Virtually every human study before this one tested the same product: NPI-031, a standardized kudzu extract that the McLean group has used across its trials. The San Francisco trial used an oral kudzu extract at roughly triple the isoflavone dose, given episodically. "Two grams of kudzu extract" is not a pharmacological constant. It is a label on a bottle of a dietary supplement, a category the FDA does not review for identity, purity or potency before sale.

The timing hypothesis is the one worth taking seriously, because it is testable and it maps onto how drinking actually happens. If the mechanism involves blunting the response to a drink, a dose taken hours beforehand is doing something a dose taken every morning at breakfast is not. That is the logic of targeted naltrexone, which patients have used as-needed for years, and it is the logic the San Francisco team imported into an herbal context.

Reading the effect size without inflation

The intention-to-treat result is real and it is modest. Twenty-seven percent fewer binge-drinking days, with an upper confidence bound of 0.99, sits a hair away from statistical nothing. A single additional heavy week in the placebo arm could have moved it across the line.

The as-treated analysis is nearly twice as large — 42 percent fewer binge days among people who took the drug at least 2.7 days a week — and it is the number that will get quoted. It should not be. Splitting a trial by how faithfully participants followed instructions breaks randomisation, and the people who take a pill on the days they plan to drink are plausibly different from the people who do not, in ways that correlate with motivation and with drinking itself.

The biomarker result runs the same direction but not identically. Phosphatidylethanol positivity fell significantly only in the as-treated analysis. Ethyl glucuronide was listed as a prespecified measure, and its result does not appear in the published abstract — a detail worth checking in the full tables before anyone describes the trial as biomarker-confirmed.

The most surprising number in the paper is the one nobody will write about: binge-drinking days were still significantly lower in the kudzu arm at the three-month posttreatment visit, with a relative risk of 0.68. A drug that was taken only on anticipated drinking days, then stopped, produced a difference that outlasted the treatment period. Either kudzu changes something durable about drinking patterns, or the participants who volunteered for a kudzu trial changed something about themselves. The paper cannot separate those, and the next trial should try.

Where this would sit next to what already exists

Alcohol use disorder has three approved medications in the United States — naltrexone, acamprosate and disulfiram — and the great majority of people who meet criteria never take any of them. Fewer than one in ten Americans with alcohol use disorder receives any treatment in a given year, a gap documented in this blog's coverage of the screening-to-intervention cascade published last week. A cheap, non-prescription, as-needed option would not be competing with naltrexone in that landscape so much as occupying a space naltrexone never reached.

For anyone weighing it, the honest framing is that alcohol use disorder treatment today rests on behavioral care plus, for some, medication-assisted treatment — and kudzu, on this evidence, is a candidate add-on rather than a replacement. Two trials from the same plant have now produced one null and one modest positive under markedly different dosing schedules. That is not a reason to buy extract at a supplement store. It is a reason for the next trial to be designed around timing, with phosphatidylethanol as the primary endpoint and a sample large enough that a 27 percent effect cannot hide.

The regulatory gap is the part that will not resolve itself. A supplement can be sold, reformulated and relabeled without anyone running a trial, and the two studies above tested products that a patient cannot reliably identify on a shelf. If targeted kudzu dosing turns out to work, the people most likely to try it will be the ones already buying herbal remedies — which means clinicians should be asking what their patients are taking, and researchers should be testing the products people actually buy.