Weight loss supplements and the evidence on benefits and risks

Weight loss supplements can show small average effects. Read the trial quality, uncertainty and safety findings before deciding what to buy.

日本語で読む（Japanese version）

Dinner is cleared away, and an advertisement on your phone offers a simpler route to weight loss: just take a supplement. Finding time to exercise feels difficult; finding time to swallow a capsule seems manageable. The appeal is understandable. Before asking which ingredient works, though, it helps to ask what “works” means. A detectable difference in a study may be much smaller than the change you hope to see in everyday life.

A systematic review of supplements and alternative therapies in adults identified 315 randomized controlled trials. Of these, 52 were classified as having a low risk of bias and enough data to support an assessment of efficacy. Within that group, 16 reported statistically significant differences in weight change between groups. The number of published trials and the number that can support a confident expectation are different things.[1]

This is a way to assess a purchase, without judging anyone who has tried one. The useful questions are how large the average difference was, who was studied, what the comparison involved, and how safety was assessed. Detailed discussions of caffeine and green tea, and of probiotics and gut bacteria, belong in separate articles. Here, the focus is the evidence behind supplements marketed for weight loss, including findings that are less favorable than an advertisement might suggest.

Many studies do not necessarily make a strong recommendation

A systematic review searches for studies according to specified criteria, selects those that qualify, and brings their findings together. A meta-analysis statistically combines results from studies. These labels describe methods; they do not remove the need to examine what was included and what the authors concluded. In the broad review of supplements and alternative therapies, both the risk of bias and the adequacy of the data varied widely.[1]

The review containing 315 trials covered people aged 18 years and older and examined 14 supplements, therapies, or combinations. Its scope matters. It was not a collection limited to supplement tablets, and its results should not be presented as a success rate for every product sold as a diet supplement. Alternative therapies were part of the same evidence base. The authors’ conclusion was that high-quality evidence supporting efficacy for weight loss was limited.[1]

The 52 trials classified as low risk and sufficient to support efficacy represented 16.5% of the total. Of those 52, 16, or 31%, reported significant between-group differences in weight change. The reported differences ranged from 0.3 to 4.93 kg. The largest value is an attractive headline, but it is neither a result everyone can expect nor an average shared by all the ingredients. A range across studies cannot stand in for the evidence on the particular ingredient you are considering.[1]

Keep the denominator in view when reading those figures. The 31% refers to the selected group of 52 trials, rather than to all 315. Likewise, the range describes reported differences, rather than the likely response of a person buying a bottle. Removing either detail gives the figures a different meaning. A careful explanation can preserve the numbers while still acknowledging why a promising result attracts interest.[1]

In a randomized controlled trial, participants are assigned to comparison groups by chance. That design helps researchers compare interventions, but the words “randomized trial” do not guarantee that every study was conducted or reported equally well. The variation in the trials included in this review makes that point concrete. When a label says “clinically tested,” the next questions concern the participants, the comparison group, and the outcome actually measured.[1]

Older evidence also deserves to be read in context. A systematic review published in 2004 required randomized, double-blind trials, in which participants and assessors did not know the allocation. It examined five systematic reviews and meta-analyses plus 25 additional trials. The evidence for most supplements as weight loss aids was not convincing. Ephedra- and ephedrine-containing products showed encouraging efficacy findings, but they were also associated with an increased risk of adverse events.[2]

Putting that review beside the broad 2021 review does not establish that nothing improved between those dates. Their searches and scope differed. What the comparison does show is that a later review still found a limited base of high-quality evidence. Publication date and strength of the conclusion deserve separate attention. A newer paper can contribute useful information without making the entire product category dependable.[1][2]

For a practical reading of an advertisement, keep the weight difference, comparator, and population together. Before-and-after photographs do not reveal how a comparison group changed. Matching an ingredient name does not establish that a commercial formulation or its conditions of use match a trial. A study finding alone is not a personal dosing plan. The evidence can help set expectations while leaving that use decision unresolved.

Statistical significance can coexist with a small weight difference

A 2021 systematic review and meta-analysis collected 67 randomized, placebo-controlled trials of weight loss supplements containing isolated organic compounds. A placebo is a comparison product without the active ingredient. In this setting, the mean difference describes the difference between the supplement and placebo groups after the intervention. It is not simply the amount of weight lost by people taking the supplement.[3]

For chitosan, the mean weight difference was −1.84 kg, with a 95% confidence interval of −2.79 to −0.88 kg. For glucomannan, it was −1.27 kg, with an interval of −2.45 to −0.09 kg. For conjugated linoleic acid, it was −1.08 kg, with an interval of −1.61 to −0.55 kg. All three differences were statistically significant, but none reached the authors’ threshold for clinical significance: a difference of at least 2.5 kg.[3]

The distinction matters because an advertisement can accurately say there was a significant result while leaving its size unclear. Here, the review authors found a statistical difference and still concluded that the evidence was insufficient to recommend these supplements for weight loss. Those statements fit together. They answer different parts of the question: whether a difference was detected, and whether the findings supported a recommendation.[3]

A confidence interval accompanies the average estimate to show its uncertainty. It should not be read as a guarantee that an individual customer’s weight change will fall within that range. For glucomannan, the interval extended to −0.09 kg. Displaying only the central estimate of −1.27 kg leaves out the reported uncertainty. Reading the interval alongside the average preserves the uncertainty rather than turning the estimate into a promise.[3]

The 2.5 kg threshold was the criterion adopted by this review. It is not a universal amount that every person needs to lose, nor should it be presented as an official treatment standard. An average difference below that line also does not mean no participant experienced a change. The appropriate conclusion stays within the review’s scope: there was currently insufficient evidence to recommend the evaluated supplements for weight loss.[3]

The unfavorable findings are part of the same assessment. Fructans did not produce a statistically significant weight difference compared with placebo. Other ingredients and some combinations produced significant differences that were borderline clinically meaningful, but each of those results came from a single trial. The authors called for more evidence before recommending them. A promising example raises a question for further testing; it does not settle whether the result will be reproduced.[3]

For someone shopping after a busy workday, the gap between these findings and the advertising message is easy to miss. “The supplement group differed slightly from the comparison group on average” does not answer “Will this take me to my goal?” To connect the trial with a person’s life, the population, duration, conditions of use, and accompanying measures matter. The review does not establish a common dose or duration for all these ingredients.[3]

It is reasonable to want a simple purchasing decision. The evidence here supports a more limited judgment: take the small average differences seriously, keep the uncertainty attached, and recognize the authors’ reservations. A finding need not be dismissed because it is modest. Equally, a modest finding should not be expanded into a large expected benefit merely because the ingredient is easy to obtain.[3]

Ingredient rankings need their evidence grades beside them

A separate 2023 systematic review and network meta-analysis included 111 randomized trials involving 6,171 adults with overweight or obesity and evaluated 18 ingredients. A network meta-analysis uses the relationships among different interventions and comparators to estimate effects. It is not a single contest in which every ingredient was tested directly against every other ingredient under identical conditions. A ranking therefore needs more interpretation than a shopping list.[16]

In the main analysis incorporating all trials, spirulina had a mean difference of −1.77 kg, with a 95% confidence interval of −2.77 to −0.78 kg. The certainty of that evidence was rated high. Curcumin had a mean difference of −0.82 kg, with an interval of −1.33 to −0.30 kg, and chitosan had a difference of −1.70 kg, with an interval of −2.62 to −0.78 kg. Certainty for curcumin and chitosan was rated moderate.[16]

These estimates should retain their grades when they appear in a comparison. An average difference and an assessment of certainty are separate entries on the evidence sheet. Removing the grade makes the findings look equally dependable, although the authors did not assess them that way. High certainty about a modest difference also does not turn that difference into a large one. Magnitude and confidence in the evidence remain distinct questions.[16]

Psyllium had a mean difference of −3.70 kg, with a 95% confidence interval of −5.18 to −2.22 kg and moderate certainty. That average is larger than the estimates for ingredients discussed in the earlier review. However, values from reviews with different populations, included trials, and methods cannot simply be lined up to identify the strongest product. The overall conclusion of the network analysis was that these supplements could produce small weight loss in adults with overweight or obesity.[3][16]

Glucomannan had a mean difference of −1.36 kg in this analysis, with a 95% confidence interval of −2.17 to −0.54 kg, but certainty was rated low. A result pointing in a similar direction to another review is informative. It still leaves the question of how much confidence to place in the estimate. The numerical difference, its interval, and the certainty grade should all remain visible when deciding what the research says.[16]

The study population is another essential qualifier. These were adults with overweight or obesity. The findings do not establish that everyone wanting a small change in appearance will have the same response. A product description that says “for adults” is broader than the population actually studied. A reader’s wish to change body shape does not, by itself, establish that the trial evidence applies directly to them.[16]

Green tea also appeared in this analysis: the mean difference was −1.25 kg, with a 95% confidence interval of −1.68 to −0.82 kg and low certainty. As with glucomannan, detecting an average difference did not remove the uncertainty in the evidence. Although green tea is a familiar drink in Japan, this analysis concerned supplementation for weight loss. The estimate cannot be assigned directly to the ordinary act of drinking tea.[16]

The useful outcome of this comparison is a clearer way to read results. Ask which population the estimate concerns, how large the difference is, how uncertain it is, and how the evidence was graded. Those questions keep the interesting findings intact while avoiding a stronger recommendation than the authors provided. They also make it harder for a large printed number to obscure a qualification written beside it.[16]

Numbers: the supplement evidence sheet

• 315 trials: randomized controlled trials in an adult review of supplements and alternative therapies, rather than supplement tablets alone.[1]
• 52 trials, or 16.5%: classified as low risk of bias and having sufficient data to support efficacy assessment.[1]
• −1.84 kg: chitosan’s mean difference from placebo; 95% confidence interval, −2.79 to −0.88 kg.[3]
• −1.27 kg: glucomannan’s mean difference from placebo; 95% confidence interval, −2.45 to −0.09 kg.[3]
• −1.77 kg: spirulina’s mean difference in adults with overweight or obesity; 95% confidence interval, −2.77 to −0.78 kg.[16]