Formulate report · September 2026
What 1,450 systematic-review conclusions say about 190 supplements
For each supplement ingredient we read what systematic reviews and meta-analyses concluded, outcome by outcome — did it improve the thing it was tested for, make no difference, or could the reviews not tell? Every conclusion is quoted from a review and linked to it, and the whole dataset is free to download.
What stands out
- Fewer than half of all conclusions (47%) found that an ingredient improved the outcome it was reviewed for.
- About one in 5 (19%) concluded it made no difference.
- 31% could not conclude either way.
- 41 conclusions (3%) reported a harm — listed in full below.
The most-reviewed ingredients
Ranked by how many outcome conclusions the reviews reached.
| Ingredient | Conclusions | Benefit | No effect | Unclear | Harm |
|---|---|---|---|---|---|
| Omega-3 (fish oil, EPA, DHA) | 112 | 50 | 26 | 28 | 8 |
| Vitamin D | 37 | 13 | 13 | 11 | |
| Ashwagandha | 22 | 14 | 2 | 6 | |
| Inositol | 21 | 10 | 4 | 7 | |
| Vitamin B1 (thiamine) | 21 | 8 | 2 | 11 | |
| Vitamin C | 21 | 4 | 7 | 9 | 1 |
| Selenium | 20 | 10 | 6 | 3 | 1 |
| Astaxanthin | 18 | 13 | 4 | 1 | |
| Probiotics | 18 | 6 | 6 | 6 | |
| Catechins | 16 | 10 | 2 | 4 | |
| Taurine | 16 | 9 | 2 | 5 | |
| B Complex | 15 | 6 | 2 | 6 | 1 |
| Beetroot | 15 | 8 | 1 | 6 | |
| Curcumin | 15 | 9 | 3 | 3 | |
| Folate | 15 | 4 | 0 | 7 | 4 |
Most consistently positive
Among ingredients with at least 8 conclusions, the highest share that found a benefit.
| Ingredient | Conclusions | Benefit | No effect | Unclear | Harm |
|---|---|---|---|---|---|
| Chlorogenic Acid | 11 | 10 | 0 | 1 | |
| Artichoke | 10 | 8 | 0 | 2 | |
| Bergamot | 9 | 7 | 0 | 2 | |
| Chicory | 8 | 6 | 0 | 2 | |
| Lactobacillus plantarum | 8 | 6 | 0 | 2 | |
| Lycopene | 8 | 6 | 2 | 0 | |
| Collagen | 11 | 8 | 1 | 2 | |
| Spirulina | 11 | 8 | 2 | 1 | |
| Astaxanthin | 18 | 13 | 4 | 1 | |
| Moringa | 15 | 10 | 2 | 3 |
Least supported
Among ingredients with at least 8 conclusions, the lowest share that found a benefit — mostly no effect, or not enough to say.
| Ingredient | Conclusions | Benefit | No effect | Unclear | Harm |
|---|---|---|---|---|---|
| Vitamin C | 21 | 4 | 7 | 9 | 1 |
| Lutein | 10 | 2 | 3 | 5 | |
| Bilberry | 14 | 3 | 8 | 3 | |
| L-Citrulline | 14 | 3 | 3 | 8 | |
| Beta-Hydroxybutyrate | 13 | 3 | 3 | 7 | |
| Bifidobacterium longum | 12 | 3 | 1 | 8 | |
| Nicotinamide Mononucleotide | 12 | 3 | 3 | 6 | |
| Chondroitin Sulfate | 8 | 2 | 4 | 2 | |
| Folate | 15 | 4 | 0 | 7 | 4 |
| DHEA | 11 | 3 | 4 | 3 | 1 |
Every harm reported
- Acetyl-L-Carnitine — chemotherapy-induced peripheral neuropathy. “Acetyl-l-carnitine was found to be likely ineffective or harmful.” PMID 37654090
- Algal Oil — ldl cholesterol. “DHA supplementation from algal oil, a marine source of (n-3) fatty acids not extracted from fish, may reduce serum TG and increase HDL-C and LDL-C in persons without coronary heart disease.” PMID 22113870
- Alpha Lipoic Acid — allergic contact dermatitis. “Key sensitizers include vitamin derivatives (tocopherol, phytonadione, ascorbic acid), herbal extracts (Ginkgo biloba, turmeric, St. John's Wort), and antioxidants (α-lipoic acid, resveratrol).” PMID 41101984
- Anthocyanins — copd clinical outcomes. “the possible negative effects of anthocyanins warn against ingredient heterogeneity” PMID 40771814
- B Complex — adverse effects. “Oral niacin and NAM were associated with a range of AEs including GI upset, liver dysfunction, flushing/vasodilatory effects, skin rash and fatigue.” PMID 40999226
- Betaine — total cholesterol. “Supplementation with at least 4 g/d of betaine for a minimum of six weeks may moderately increase plasma TC, which might be important in the context of cardiovascular health.” PMID 31809615
- Biotin — immunoassay test results. “Biotin ingested in moderate to high doses can cause immunoassay interference.” PMID 27362288
- Caffeine — adverse side effects. “Caffeine intake in athletes increases the likelihood of side effects in a dose-dependent manner, with tachycardia, insomnia, abdominal discomfort, and anxiety being the most consistent.” PMID 42033594
- Chamomile — bleeding risk with warfarin. “These include Chinese wolfberry, chamomile tea, cannabis, cranberry, chitosan, green tea, Ginkgo biloba, ginger, spinach, St. John's Wort, sushi and smoking tobacco.” PMID 32478963
- DHEA — hdl cholesterol. “Overall, supplementation with DHEA did not change circulating values of TC, LDL-C and TG, whereas it may decrease HDL-C levels.” PMID 32675010
- Folate — gestational diabetes mellitus. “The most notable association was with gestational diabetes mellitus, where elevated intake of folic acid and a high blood folate concentration were found to increase the risk.” PMID 42248048
- Folate — large-for-gestational-age. “Excessive intake of folic acid supplements and a high blood folate concentration were associated with gestational diabetes mellitus, and adverse neurodevelopmental issues in offspring, as well as large-for-gestational-age infants and higher birth weight.” PMID 42248048
- Folate — malaria treatment success. “People who received folic acid supplementation (alone or with iron) and antifolate antimalarial treatment were less likely to clear malaria parasitaemia (day 3), and had increased treatment failure (days 7, 14, 28).” PMID 41705996
- Folate — folic acid intake relative to upper tolerable limit. “Folic acid supplement recommendations and the upper tolerable limit of 1000 mcg set by policy makers warrant careful review in light of potential adverse effects of exceeding the upper tolerable limit on folic acid absorption and metabolism, and subsequent impacts on women's health during their childbearing years.” PMID 35807899
- Ginger — bleeding risk. “healthcare professionals should be aware of the increased risk of bleeding when taking several food and herbs. These include Chinese wolfberry, chamomile tea, cannabis, cranberry, chitosan, green tea, Ginkgo biloba, ginger, spinach, St. John's Wort, sushi and smoking tobacco.” PMID 32478963
- Ginkgo Biloba — allergic skin reaction. “Key sensitizers include vitamin derivatives (tocopherol, phytonadione, ascorbic acid), herbal extracts (Ginkgo biloba, turmeric, St. John's Wort), and antioxidants (α-lipoic acid, resveratrol).” PMID 41101984
- Hyaluronic Acid — adverse events (facial fillers). “The overall prevalence of adverse events was low for botulinum toxin and thread lifts, but high for hyaluronic acid fillers.” PMID 41047572
- Inulin — infant growth. “Conversely, infants on formulas with inulin-enriched oligo-fructose (IOF) experienced reduced WG and HG.” PMID 40215676
- Lactobacillus paracasei — mortality in critical illness. “The largest of these, and the one with the lowest risk of bias, demonstrated that probiotics increased mortality, in part because of the precipitation of ischemic bowel disease (in patients who were also receiving postpyloric enteral nutrition infusions).” PMID 19244148
- Milk Thistle — antiretroviral drug levels. “Calcium carbonate, ferrous fumarate, some forms of ginkgo, some forms of garlic, some forms of milk thistle, St. John's wort, vitamin C, zinc sulfate, and multivitamins were all found to significantly decrease the levels of selected antiretrovirals and should be avoided in patients taking these antiretrovirals.” PMID 27655839
- Niacin — all-cause mortality. “an increased risk was seen with niacin (with statin) for all-cause mortality.” PMID 33509399
- Olive Oil — lean mass. “Lean mass responds negatively to OO the higher the dose (slope = -0.61, 95% CI [-1.01, -0.21], p = 0.003) and time offered (slope = -0.8822, 95% CI [-1.44, -0.33], p = 0.002).” PMID 37278997
- Omega-3 (fish oil, EPA, DHA) — atrial fibrillation risk. “balanced against dose-dependent AF increases at high doses (> 1.5 g/day).” PMID 42144851
- Omega-3 (fish oil, EPA, DHA) — ldl cholesterol. “dietary EPA/DHA supplementation improved the TG and HDL status but increased LDL levels in comparison with ALA” PMID 32175534
- Omega-3 (fish oil, EPA, DHA) — total cholesterol. “total cholesterol (TC) (WMD 0.179 mmol l-1; 95% CI 0.006, 0.352)” PMID 32175534
- Omega-3 (fish oil, EPA, DHA) — atrial fibrillation. “studies of both the EPA/DHA combination and EPA alone showed a significant increase in risk of new-onset atrial fibrillation” PMID 37031750
- Omega-3 (fish oil, EPA, DHA) — large for gestational age. “There was a possibly reduced risk of perinatal death and of neonatal care admission, a reduced risk of LBW babies; and possibly a small increased risk of LGA babies with omega-3 LCPUFA.” PMID 30480773
- Omega-3 (fish oil, EPA, DHA) — ldl cholesterol. “HDL cholesterol and blood pressure benefits are not consistent, fasting glycemia is largely unaffected, and LDL cholesterol may increase, especially in low doses.” PMID 41156531
- Selenium — urinary tract anomalies. “Higher concentrations of Se were associated with increased risk for urinary tract anomalies (n = 2150)” PMID 41683301
- Vitamin A — vitamin a toxicity. “Hypervitaminosis A, particularly arising from non-prescription supplement misuse, mimics various disorders and presents commonly with hypercalcemia and neurological dysfunction.” PMID 42176931
- Vitamin B6 — peripheral neuropathy. “Current scientific evidence supports a neurotoxic role of B6 at high levels.” PMID 37447150
- Vitamin C — kidney and hemolytic adverse effects. “Some of the major risks include oxalate nephropathy and hemolysis in patients with glucose-6-phosphate dehydrogenase deficiency, especially with very large or repeated doses, suggesting pre-screening to avoid these risks.” PMID 41815850
- Watermelon — fasting blood sugar. “In addition, watermelon consumption led to a significant increase in fasting blood sugar (FBS).” PMID 37369281
- Whey protein — kidney and liver function. “The majority of the papers associated the damaging effect with the chronic and abusive use of whey protein, with the kidneys and liver being the main organs affected.” PMID 32702243
By goal
How this was done
For each ingredient, two PubMed searches for systematic reviews and meta-analyses — the newest, and the most relevant — up to 14 reviews in all, keeping the conclusions each review states in its abstract.
An AI model (Anthropic's Claude) read those conclusions and recorded, for every outcome they address, whether the review found a benefit, no effect, harm, or could not tell. Every recorded conclusion must quote a sentence, and that sentence is checked word for word against the abstract of a review it cites; any that cannot be found is dropped rather than kept.
We checked the result against 28 claims a person had read from the literature by hand: the two agreed on 19, differed only in strength on 7, and disagreed outright on 2.
Limits. A conclusion records the direction of an effect, not its size. Reviews inherit the quality and the publication bias of the trials they pool, and study populations vary — the outcome is shown as the review worded it. This is research, not medical advice.
Last refreshed September 2026. Source data version 6a0f47b. Abstracts courtesy of the U.S. National Library of Medicine (PubMed).
Data and citation
Download all 1,450 conclusions (CSV) — ingredient, outcome, direction, PubMed IDs and the quoted sentence. To cite: Formulate, “Supplement systematic-review conclusions,” September 2026, formulate-health.app/reports/supplement-review-conclusions.