Why the evidence for botanicals is thin
Few botanical drugs reach FDA approval, and many botanical trials are small, fragile or inconclusive. FDA data and research on trial methods point to causes that can be measured.
Status in December 2020: 253 of 739 INDs were active, 374 withdrawn or terminated.
Data table: botanical INDs by indication, first FDA action and status
| Indication | INDs | Safe to proceed | Clinical hold | Withdrawn | Other first action | Active, 2020 | On hold | Withdrawn or terminated | Inactive | Other status |
|---|---|---|---|---|---|---|---|---|---|---|
| Oncology | 254 | 175 | 56 | 12 | 11 | 82 | 7 | 137 | 24 | 4 |
| Other indications | 485 | 282 | 111 | 14 | 78 | 171 | 33 | 237 | 31 | 13 |
| All botanical INDs | 739 | 457 | 167 | 26 | 89 | 253 | 40 | 374 | 55 | 17 |
| Proposed phase | INDs | Active, December 2020 | Not active | Active share |
|---|---|---|---|---|
| Phase 1 | 137 | 46 | 91 | 34% |
| Phase 2 | 111 | 33 | 78 | 30% |
| Phase 3 | 6 | 3 | 3 | 50% |
Status. EUTECNIQ Sciences is a company in formation with no trials, data or products of its own. Every result here is published third-party work or the site's own editorial grading; the last section labels EUTECNIQ's role as a proposal.
Related: the four botanical drugs · evidence and pharmacokinetics · proof nobody can own · lessons from large trials
Two routes to market, one of them without efficacy trials
A botanical drug must prove efficacy and batch-to-batch consistency; a botanical supplement reaches the market with no FDA approval at all.
DiMasi JA, Grabowski HG, Hansen RW. J Health Econ 2016;47:20–33 (abstract). Wouters OJ, McKee M, Luyten J. JAMA 2020;323(9):844–853, corrected version (correction notice JAMA 2022;328(11):1111). Bourqui A, et al. Front Pharmacol 2026;17:1766381, opinion article, section 4 (the authors' uva-ursi trial, UK).
The cost of proof lands on one sponsor, while a supplement seller can market the same plant without it; the FDA authors name insufficient raw-material information, the variation and complexity of botanical components and the evidence needed for approval as the main challenges.
Explanation, details and sources
What the 2016 guidance asks. Because a botanical drug is a mixture whose active constituents may be unknown, FDA says “one of the critical issues for botanical drugs is ensuring that the therapeutic effect for marketed drug product batches is consistent.” It supports that consistency with a “totality of the evidence” approach: raw-material control (agricultural practice and collection), chemical tests with manufacturing control and process validation, and a biological assay with clinical data (Section III.B). In Phase 3 it suggests testing several drug-product batches, each made from a different drug-substance batch (Section VI.A), and “highly recommended” randomizing subjects to batches at each site (Section VI.F.1).
Where it is flexible. Early clinical evaluation “does not differ significantly” from that of purified drugs, and products already marketed as supplements “generally would not require typical Phase 1 tolerability studies” when prior human use is shown to be relevant (Section V.E). The Phase 3 and NDA steps carry the extra burden.
What the FDA authors report. Park et al. list “insufficient information on botanical raw materials, which often leads to trial holds or withdrawals in early-phase studies,” “the inherent variation and complexity of botanical components” and “the rigorous evidence-based requirements necessary for drug approval.” Terminations “often occurred when methods, facilities, and controls used for manufacturing the products were inadequate.” They set the share of oncology INDs still active in December 2020, 34% for Phase 1 and 30% for Phase 2, beside published success rates of 66% and 49% for conventional Phase 1 and 2 INDs, and write that the probability of success “seemed lower”; the botanical figure is a share still active, not a count of programs that advanced. Of the oncology INDs, 24% were commercial applications; the authors put research applications from clinical investigators at 76% (data not shown in the paper).
Earlier FDA count. In 2020, FDA's Botanical Review Team reported “over 800” botanical INDs and pre-IND meeting requests before 2018, with two botanical NDAs then approved (Wu et al. 2020). The 739 above exclude pre-IND requests and cover 1984 to 2020.
The supplement route. “Under DSHEA, FDA does not have the authority to approve dietary supplements before they are marketed” (FDA Q&A). A new dietary ingredient that has not been in the food supply in an unaltered form needs a premarket notification at least 75 days before marketing, setting out why it “will reasonably be expected to be safe” (21 CFR 190.6); there is no efficacy review. A structure/function claim needs substantiation held by the firm, a notice to FDA within 30 days of marketing, and the label disclaimer that the statement “has not been evaluated by the Food and Drug Administration” (21 CFR 101.93). Who can and cannot protect the cost of a trial is set out on the food–drug gap page; how often drug programs of any kind reach approval is on the drug pathway.
Cost figures. DiMasi et al. estimated an average out-of-pocket cost of $1,395 million per approved compound, capitalized at a 10.5% real discount rate to $2,558 million (2013 dollars). Wouters et al., in the version corrected in 2022 after an error in a source of clinical-trial success rates, estimated a median of $1,141.7 million (95% CI $888.1–1,480.8 million) and a mean of $1,559.1 million (2018 dollars), after accounting for failed trials; the data came mainly from smaller firms and orphan, first-in-class and accelerated-approval products. The originally published figures were $985.3 million and $1,335.9 million. Neither study is specific to botanicals. In the UK trial Bourqui et al. describe, no GMP uva-ursi product was on the market, so the team had an extract made and encapsulated under GMP, “costing £92,500 to produce enough capsules to treat 382 patients … and taking over 2 years.”
Park et al. 2025 (PMC12446141); Wu C, Lee SL, Taylor C, et al. J Nat Prod 2020;83(2):552–562; FDA Botanical Drug Development guidance (2016); FDA Q&A on Dietary Supplements; eCFR 21 CFR 190.6 and 101.93; DiMasi et al. 2016; Wouters et al. 2020 (corrected 2022); Bourqui et al. 2026 (opinion article).
Seven causes behind “no significant difference”
A null result in a botanical trial can come from the product, the design, the dose reaching the body or the record, before the plant itself is judged.
Product variation, low exposure and small samples make a real effect harder to detect; small-study and country effects push the other way, toward larger effects than large trials report.
Text version: the seven causes and their sources
- The product varies. In 25 U.S. ginseng products, ginsenoside concentrations varied 15-fold in capsules and 36-fold in liquids, and eleutherosides 43- and 200-fold [1]. Of 59 echinacea preparations, 6 (10%) contained no measurable echinacea, and 9 of 21 labeled “standardized” (43%) met their label standard [2]. Of 81 RCTs of five herbs published 2000–2004, 12 (15%) reported testing the content of the product studied [3].
- Trials are small. In 227 Chinese herbal medicine RCTs published in English 2010–2019 (a random 20% of those eligible), the median sample size was 115 participants (range 12 to 3,143) [4]. In 30 herbal RCTs for irritable bowel syndrome with a positive binary outcome, the median fragility index was 1.5, and half the trials would lose significance if one patient's outcome changed [5]. In a Cochrane review of St John's wort, nine larger placebo-controlled trials gave a response ratio of 1.28 and nine smaller ones 1.87 [6]. Across 93 meta-analyses of therapies of all kinds (binary outcomes), trials with fewer than 50 patients showed effects 48% larger, on average, than trials of 1,000 or more [7].
- Reports leave out what matters. Across 206 herbal RCTs, 45% of CONSORT checklist items were reported; allocation concealment, sequence generation and intention-to-treat were incompletely reported, and the percentage of active constituents may be missing [8]. In 2006 the CONSORT group elaborated 9 checklist items for herbal trials; item 4, the herbal intervention, “required the most extensive elaboration” [9].
- Endpoints and populations are unclear. In the 227-trial sample, 40.5% did not clearly specify any primary outcome and 24.7% did not report any adverse event [4]. In the St John's wort review, trials from German-speaking countries reported results more favorable to the extract; the review's plain-language summary says this “could be due to the inclusion of patients with slightly different types of depression, but it cannot be ruled out that some smaller studies from German‐speaking countries were flawed and reported overoptimistic results” [6].
- Too little reaches the blood. After single doses of 0.5 to 8 g of a standardized curcuminoid powder extract, no curcumin (the parent compound) was detected in serum; at 10 and 12 g low levels were found in 1 of 3 volunteers at each dose [10]. Metabolites are another matter, charted with the absorption data on Evidence and pharmacokinetics.
- Positive results are overrepresented. In a review of controlled-trial abstracts to 1995, every acupuncture trial from China, Japan, Hong Kong and Taiwan was positive; among trials of other interventions, 99% of those published in China and 75% of those published in England favored the test treatment, and no trial from China or Russia/USSR found a treatment ineffective [11].
- Few have a reason to pay. FDA has no authority to approve supplements before marketing [12]. In an opinion article, Swiss and UK trial teams write that companies are reluctant to disclose production details partly because herbal food supplements “are difficult to patent,” and that companies have “no reason to obtain GMP certification,” which clinical-trial products need in Switzerland and, for full-scale trials, in the UK [13]. Patent and exclusivity law is set out on the food–drug gap page.
- Harkey MR, Henderson GL, Gershwin ME, Stern JS, Hackman RM. Variability in commercial ginseng products: an analysis of 25 preparations. Am J Clin Nutr 2001;73(6):1101–1106.
- Gilroy CM, Steiner JF, Byers T, Shapiro H, Georgian W. Echinacea and truth in labeling. Arch Intern Med 2003;163(6):699–704.
- Wolsko PM, Solondz DK, Phillips RS, Schachter SC, Eisenberg DM. Lack of herbal supplement characterization in published randomized controlled trials. Am J Med 2005;118(10):1087–1093.
- Hu J, Zhang HN, Feng S, et al. Randomized controlled trials of Chinese herbal medicine published in English from 2010 to 2019: a bibliometrics study. Ann Palliat Med 2021;10(12):12945–12954.
- Luo M, Huang J, Wang Y, et al. How fragile the positive results of Chinese herbal medicine randomized controlled trials on irritable bowel syndrome are? BMC Complement Med Ther 2024;24:300.
- Linde K, Berner MM, Kriston L. St John's wort for major depression. Cochrane Database Syst Rev 2008;(4):CD000448.
- Dechartres A, Trinquart L, Boutron I, Ravaud P. Influence of trial sample size on treatment effect estimates: meta-epidemiological study. BMJ 2013;346:f2304.
- Gagnier JJ, DeMelo J, Boon H, Rochon P, Bombardier C. Quality of reporting of randomized controlled trials of herbal medicine interventions. Am J Med 2006;119(9):800.e1–11.
- Gagnier JJ, Boon H, Rochon P, Moher D, Barnes J, Bombardier C; CONSORT Group. Reporting randomized, controlled trials of herbal interventions: an elaborated CONSORT statement. Ann Intern Med 2006;144(5):364–367.
- Lao CD, Ruffin MT, Normolle D, et al. Dose escalation of a curcuminoid formulation. BMC Complement Altern Med 2006;6:10.
- Vickers A, Goyal N, Harland R, Rees R. Do certain countries produce only positive results? A systematic review of controlled trials. Control Clin Trials 1998;19(2):159–166.
- FDA. Questions and Answers on Dietary Supplements (content current as of August 27, 2026).
- Bourqui A, Csajka C, Heinrich M, et al. Challenges in performing clinical research on herbal products: experiences from Switzerland and the UK. Front Pharmacol 2026;17:1766381 (opinion article).
Why a 30-per-arm trial usually misses a small effect
Power is the chance that a trial detects an effect that is really there; with 30 people per arm and a small effect (d = 0.2), it is 12.1%, about one in 8.
Interactive calculator Change the inputs and the result updates. An educational simulation with stated inputs.
At 30 per arm, a real effect of d = 0.20 gives a non-significant result 88% of the time.
Method, assumptions and limits
- What it calculates
- The probability that a two-arm randomized trial with a continuous outcome reaches a two-sided significant result, given the true standardized effect, the number per arm and α; the number per arm for 80% power; and the smallest effect that has 80% power at the chosen number.
- Equations
- Power = Φ(d·√(n/2) − z1−α/2) + Φ(−d·√(n/2) − z1−α/2), where Φ is the standard normal distribution function and zq its quantile. Solving for n at power 1 − β (neglecting the second, very small term) gives n = 2(z1−α/2 + z1−β)²/d², rounded up; this is the per-group formula n = (Zα/2 + Zβ)²·2σ²/(μ1 − μ2)² of Sakpal (2010) with d = (μ1 − μ2)/σ. Smallest detectable d = (z1−α/2 + z0.80)·√(2/n).
- Variables and units
- d: true difference in means divided by the common standard deviation (dimensionless), 0.05 to 1.5. n: participants analyzed per arm, 4 to 1,000. α: two-sided significance level.
- Assumptions
- Two equal arms, one continuous primary outcome, normally distributed with equal variances, independent participants, no dropout, one test. The normal approximation slightly overstates power at small n compared with the exact t-test (table below).
- Valid range
- Use n of about 10 or more per arm for the approximation; below that the exact noncentral t calculation gives noticeably lower power.
- What it does not establish
- It does not say what effect a botanical has. The effect sizes are inputs, not estimates for any product, and conventions such as d = 0.2 for “small” are arbitrary and should not be read rigidly (Lakens 2013).
- Verification
- Checked against statsmodels 0.15 NormalIndPower (same formula) and TTestIndPower (exact noncentral t); the calculator is checked against an independent scipy implementation
Power at α = 0.05 (two-sided): this calculator against statsmodels d n per arm Normal approximation (this page) statsmodels NormalIndPower statsmodels TTestIndPower (exact t) 0.2 10 7.3% 7.3% 7.1% 0.2 30 12.1% 12.1% 11.9% 0.2 100 29.3% 29.3% 29.1% 0.2 400 80.7% 80.7% 80.6% 0.5 10 20.1% 20.1% 18.5% 0.5 30 49.1% 49.1% 47.8% 0.5 100 94.2% 94.2% 94.0% 0.5 400 100.0% 100.0% 100.0% 0.8 10 43.2% 43.2% 39.5% 0.8 30 87.3% 87.3% 86.1% 0.8 100 100.0% 100.0% 100.0% 0.8 400 100.0% 100.0% 100.0% - References
- Sakpal TV. Sample size estimation in clinical trial. Perspect Clin Res 2010;1(2):67–69. Lakens D. Calculating and reporting effect sizes to facilitate cumulative science. Front Psychol 2013;4:863 (benchmarks d = 0.2, 0.5, 0.8 after Cohen 1988; power 0.80 as the accepted minimum). Button KS, Ioannidis JPA, Mokrysz C, et al. Power failure. Nat Rev Neurosci 2013;14:365–376.
A non-significant result from an underpowered trial is weak evidence of no effect; low power also inflates the effects that do reach significance (Button et al. 2013).
72.8% of graded findings on this site are C or X
Of 1,071 graded findings on 264 ingredients, 72.8% are graded C or X and 24.8% A or B. Botanicals carry 522 of the rows: 76% C, 3% X and 0.8% A.
Data tables: grades and classes
| Grade | Definition used on this site | Findings | Share |
|---|---|---|---|
| A | Consistent results from several randomized trials or a systematic review of them | 27 | 2.5% |
| B | Some positive randomized trials, but small, short, few or inconsistent | 239 | 22.3% |
| C | Limited human evidence: single small trials, observational data or mixed results | 695 | 64.9% |
| D | Laboratory or animal evidence only; not shown in people | 25 | 2.3% |
| X | Well-designed human trials found no meaningful effect | 85 | 7.9% |
| Class | Definition | Ingredients | Share |
|---|---|---|---|
| Class 1 | Well studied in people | 173 | 65.5% |
| Class 2 | Moderately studied in people | 66 | 25.0% |
| Class 3 | Limited studies in people | 22 | 8.3% |
| Class 4 | Animal studies only | 3 | 1.1% |
Data table: grade by category
| Category | Ingredients | Findings | A | B | C | D | X | C or X |
|---|---|---|---|---|---|---|---|---|
| Botanical | 130 | 522 | 4 (1%) | 87 (17%) | 398 (76%) | 17 (3%) | 16 (3%) | 79% |
| Other | 44 | 188 | 10 (5%) | 53 (28%) | 108 (57%) | 1 (1%) | 16 (9%) | 66% |
| Amino acid and protein | 21 | 85 | 2 (2%) | 36 (42%) | 38 (45%) | 2 (2%) | 7 (8%) | 53% |
| Vitamin | 18 | 70 | 4 (6%) | 9 (13%) | 35 (50%) | 0 (0%) | 22 (31%) | 81% |
| Probiotic | 11 | 59 | 1 (2%) | 22 (37%) | 25 (42%) | 0 (0%) | 11 (19%) | 61% |
| Mineral | 13 | 49 | 4 (8%) | 14 (29%) | 26 (53%) | 0 (0%) | 5 (10%) | 63% |
| Fatty acid | 10 | 37 | 2 (5%) | 10 (27%) | 17 (46%) | 0 (0%) | 8 (22%) | 68% |
| Marine and algae | 10 | 32 | 0 (0%) | 6 (19%) | 24 (75%) | 2 (6%) | 0 (0%) | 75% |
| Mushroom | 7 | 29 | 0 (0%) | 2 (7%) | 24 (83%) | 3 (10%) | 0 (0%) | 83% |
Data table: grade by effect studied
| Effect studied | Ingredients | Findings | A | B | C | D | X | C or X |
|---|---|---|---|---|---|---|---|---|
| Heart, lipids and blood pressure | 105 | 167 | 11 (7%) | 62 (37%) | 77 (46%) | 0 (0%) | 17 (10%) | 56% |
| Blood sugar and weight | 86 | 128 | 2 (2%) | 34 (27%) | 84 (66%) | 3 (2%) | 5 (4%) | 70% |
| Digestion and gut | 38 | 87 | 1 (1%) | 29 (33%) | 51 (59%) | 1 (1%) | 5 (6%) | 64% |
| Energy, exercise and muscle | 45 | 68 | 2 (3%) | 17 (25%) | 43 (63%) | 0 (0%) | 6 (9%) | 72% |
| Inflammation, joints and pain | 37 | 58 | 0 (0%) | 15 (26%) | 41 (71%) | 1 (2%) | 1 (2%) | 72% |
| Memory and cognition | 41 | 53 | 0 (0%) | 4 (8%) | 41 (77%) | 0 (0%) | 8 (15%) | 92% |
| Colds and immune function | 27 | 42 | 2 (5%) | 8 (19%) | 28 (67%) | 0 (0%) | 4 (10%) | 76% |
| Cancer research | 30 | 41 | 0 (0%) | 4 (10%) | 25 (61%) | 1 (2%) | 11 (27%) | 88% |
| Skin, hair and healthy aging | 23 | 31 | 0 (0%) | 3 (10%) | 24 (77%) | 1 (3%) | 3 (10%) | 87% |
| Anxiety, stress and cortisol | 20 | 28 | 1 (4%) | 7 (25%) | 19 (68%) | 0 (0%) | 1 (4%) | 71% |
| Menopause | 18 | 24 | 0 (0%) | 5 (21%) | 18 (75%) | 0 (0%) | 1 (4%) | 79% |
| Mood and depression | 18 | 23 | 1 (4%) | 5 (22%) | 16 (70%) | 0 (0%) | 1 (4%) | 74% |
| Eyes and vision | 15 | 23 | 0 (0%) | 4 (17%) | 13 (57%) | 0 (0%) | 6 (26%) | 83% |
| Fertility and pregnancy | 20 | 22 | 3 (14%) | 7 (32%) | 12 (55%) | 0 (0%) | 0 (0%) | 55% |
| Sleep | 16 | 21 | 0 (0%) | 6 (29%) | 15 (71%) | 0 (0%) | 0 (0%) | 71% |
| Liver and kidney | 15 | 18 | 0 (0%) | 2 (11%) | 14 (78%) | 1 (6%) | 1 (6%) | 83% |
| Urinary tract and prostate | 8 | 18 | 1 (6%) | 3 (17%) | 10 (56%) | 0 (0%) | 4 (22%) | 78% |
| Bone health | 13 | 16 | 0 (0%) | 1 (6%) | 14 (88%) | 0 (0%) | 1 (6%) | 94% |
| Menstrual pain and PMS | 12 | 14 | 0 (0%) | 5 (36%) | 9 (64%) | 0 (0%) | 0 (0%) | 64% |
| Headache and migraine | 7 | 10 | 0 (0%) | 3 (30%) | 7 (70%) | 0 (0%) | 0 (0%) | 70% |
Left out for fewer than 10 findings: Hydration and electrolytes (2) . Findings without an effect tag are counted in the totals above but not in this table.
Data table: grade by evidence class
| Evidence class | Ingredients | Findings | A | B | C | D | X | C or X |
|---|---|---|---|---|---|---|---|---|
| Class 1: Well studied in people | 173 | 730 | 27 (4%) | 212 (29%) | 417 (57%) | 2 (0%) | 72 (10%) | 67% |
| Class 2: Moderately studied in people | 66 | 271 | 0 (0%) | 26 (10%) | 229 (85%) | 4 (1%) | 12 (4%) | 89% |
| Class 3: Limited studies in people | 22 | 63 | 0 (0%) | 1 (2%) | 49 (78%) | 12 (19%) | 1 (2%) | 79% |
| Class 4: Animal studies only | 3 | 7 | 0 (0%) | 0 (0%) | 0 (0%) | 7 (100%) | 0 (0%) | 0% |
An X grade needs well-designed trials that found no effect; the vitamin category has the highest X share (31% of 70 rows). Botanical rows mostly sit at C (398 of 522), limited or mixed human evidence. For Class 1, the best-studied ingredients, 57% of rows are C.
Five practices that answer six of the seven causes
Each practice below is already a published standard or policy; together they reach six of the seven causes, and the seventh, who pays, is a question of law and funding.
Text version: the five practices and their sources
- Characterized, standardized material. NIH's Office of Dietary Supplements states that “reproducible research on natural products (NPs) requires the use of replicable, well-characterized NP preparations,” with authentication of identity and assessment of composition, purity and stability. What “standardized” does and does not mean on a label is on How to judge a supplement.
- Adequately powered trials. The sample size is calculated before the trial from the expected difference, its standard deviation, α and power (Sakpal 2010); 0.80 is the generally accepted minimum power (Lakens 2013). The calculator above shows the scale involved.
- Registration before enrollment. ICMJE recommends that journal editors require registration of clinical trials in a public registry “at or before the time of first patient consent for enrollment” as a condition of consideration for publication; its examples of health-related interventions include dietary interventions. A registered protocol fixes the primary outcome and makes unpublished trials visible.
- CONSORT reporting with the herbal extension. The 2006 extension elaborates 9 CONSORT items; the elaboration of item 4 “outlines specific information required for complete reporting of the herbal medicine intervention” (Gagnier et al., Explore 2006).
- Pharmacokinetics before efficacy. FDA's botanical guidance says that when the major active constituents are known, the sponsor “should attempt to measure the blood levels” of them; that an approach based on a pharmacodynamic or clinical endpoint may be appropriate if no quantifiable active constituents are available (Section V.D); and that a clinical dose-response study may be necessary when the dose is uncertain (Section V.E). Why exposure comes before benefit is set out on Absorption is not efficacy.
NIH ODS, Funding: The Importance of Natural Product Characterization and Integrity for Dietary Supplement Research (no date printed; read October 30, 2026). ICMJE, Clinical Trial Registration (journal policy, not law). Gagnier JJ, et al. Ann Intern Med 2006;144:364–367, Explore 2006;2:143–149 and J Clin Epidemiol 2006;59:1134–1149. FDA Botanical Drug Development guidance, December 2016.
How EUTECNIQ would apply this Proposal
EUTECNIQ proposes to supply trial material with documented identity, a validated marker assay and batch records; to plan human studies with a written power calculation, registration before enrollment and CONSORT-herbal reporting; and to place pharmacokinetic studies before efficacy trials, with clinical work run by qualified external partners. None of this exists yet.
Planning human evidence for a botanical?
Proposal EUTECNIQ proposes evidence plans that start from characterized material and an honest power calculation.
Health information notice
These statements have not been evaluated by the Food and Drug Administration. Nothing on this site is intended to diagnose, treat, cure or prevent any disease. The information is for general education only. It is not medical advice and does not replace advice, diagnosis or treatment from a qualified health professional, or any medication a clinician has prescribed.
Do not start, stop or change any medicine, supplement or herbal product because of something you read here. Talk to your doctor or pharmacist first, especially if you are pregnant or breastfeeding, have a medical condition, take any medicines, are preparing for surgery, or are caring for a child. Research findings, traditional uses and safety data are reported as their sources state them; they are not recommendations or claims about any product. If you think you may have a medical emergency, call your doctor or 911 at once.
EUTECNIQ Sciences sells no supplements or health products today and receives nothing from any product or ingredient mentioned on this site; any commercial interest it later has in an ingredient will be stated on that ingredient’s page.