Section 1

The Anatomy of a Safety Claim

"X herb is dangerous" is not a safety determination — it is an assertion about evidence. Before evaluating the claim, you need to identify what kind of evidence it rests on, whether causality was assessed, and which biases may have inflated the apparent signal. These questions are not skepticism for its own sake; they are the minimum clinical epistemic standard.

What type of evidence is the claim citing?

⚠

Case Report or Cluster

Most common basis for herb safety headlines

›

A case report documents an adverse event in one patient. A cluster is several reports pointing toward a pattern. Case reports are signals, not proofs. They establish temporal association at best — the patient used the herb and then something happened. They cannot establish:

Causation (the herb caused the event)
Incidence (how often this happens per user)
Risk compared to baseline (did exposure actually increase rate?)

Critical question: Was the product authenticated? Was causality formally assessed? Is the denominator known (how many people use this herb without incident)?

📊

Adverse Event Database Count (FAERS / CFSAN)

Raw report counts without denominator

›

Adverse event databases aggregate voluntary reports. A high count means the product has high use and/or high media attention around it — both inflate reporting independent of actual risk. There is no denominator (total users), no causality assessment, and systematic underreporting of routine events.

Critical question: Has reporting rate been adjusted for market exposure? Were any reports formally assessed for causality before inclusion?

🐭

Animal Study

Preclinical signal, not clinical evidence

›

Animal hepatotoxicity studies use gavage (bolus delivery to a fasted animal) at doses that produce suprapharmacological human equivalents by body weight. The route, dose rate, fed/fasted state, and species-specific hepatic metabolism all interact. The preclinical failure rate for pharmaceutical candidates exceeds 90% — "no toxicity in rats" does not equal "safe in humans," and the inverse fails too.

Animal data is not useless. It generates coarse-grained signals worth investigating, and mechanistic data can be illuminating even when direct translation fails. The error is treating it as dispositive rather than preliminary.

Critical question: What was the route and dose? Does the dosing protocol bear any resemblance to human exposure patterns? Is the mechanism data interpretable across species?

👥

Observational Cohort / Epidemiological Study

Stronger signal, confounding remains a problem

›

Cohort studies track exposed vs. unexposed populations over time. They can establish incidence and relative risk, but confounding is still a major challenge in herbal research — users of botanical products often differ from non-users in health behaviors, polypharmacy, and baseline health status. Population selection can inflate the apparent signal for a subgroup whose risk actually belongs to their underlying condition or concomitant medications.

Critical question: Is the exposed population comparable to the unexposed? Were confounders (polypharmacy, baseline condition, preparation form) controlled?

Section 2

Causality Assessment Methods

Causality assessment asks: given this adverse event, how likely is it that this substance was responsible? Three method classes exist — each with different data requirements and tradeoffs. Note: the Naranjo scale is not a "method" — it is one algorithmic implementation, designed for pharmaceuticals and generally inapplicable to botanicals without modification. The class matters more than any specific instrument.

Expert Judgment

Subjective

A trained clinician or panel evaluates all available evidence and reaches a narrative probability judgment. No formal scoring system.

Works whenCase is complex, data are sparse, or no validated tool fits the botanical context.

ProblemHigh inter-observer variability. The same case assigned to two experts may produce different conclusions.

Probabilistic

Bayesian

Uses prior probability data and case details to compute a likelihood score. Produces a number with an inherent measure of uncertainty.

Works whenSufficient population-level data exist for prior probability estimates.

ProblemRequires reliable data volume — precisely what is missing in most botanical adverse event cases. Rarely usable in practice.

Algorithmic

Structured

Structured decision tree with weighted yes/no criteria: temporal relationship, dechallenge, rechallenge, alternative explanation, overdose, prior reports.

Works whenStandardization is needed; reduces inter-observer variability; can run on paper.

ProblemCriterion weighting is often arbitrary. No validated tool exists specifically for complex botanical preparations. Method matters more than implementation.

The hybrid argument: The most defensible approach for botanical cases is often algorithmic triage (structured screening) followed by expert review for ambiguous results. The key: you must define the score threshold at which algorithmic output hands off to clinical expert assessment.

Section 3

Data Hierarchy for Causality Assessment

Not all data elements carry equal weight. Before choosing a method, classify what you have.

Tier 1 — Immediate

Highest weight; essential for any causality judgment

Dechallenge + positive rechallenge is the strongest possible causal signal in a single case. Rechallenge may be unethical depending on event severity — this judgment call is part of the assessment.

Tier 2 — Secondary

Necessary to exclude competing explanations

Tier 3 — Tertiary

Supportive but not sufficient alone

Section 4

Bias Sources in Herbal Pharmacovigilance

Attribution Error

A risk signal originating in a specific preparation form gets attributed to the parent herb. The category absorbs the signal.

GTE hepatotoxicity attributed to "green tea." Kratom overdose attributed to "kratom" rather than to adulterated product or polypharmacy.

Denominator Blindness

Reporting counts are not adjusted for market exposure. A highly used herb accumulates more reports than a rarely used one regardless of actual risk per user.

Decades of daily green tea consumption across billions of drinkers vs. adverse event counts from supplements sold to a much smaller population.

Population Selection Bias

The population using a product differs systematically from the population not using it. The baseline risk difference explains events attributed to the product.

Kratom users self-medicating opioid withdrawal have higher baseline risk for opioid-related adverse events regardless of kratom's pharmacology.

Authentication Failure as Confound

If the implicated product was adulterated, misidentified, or contaminated, the adverse event belongs to the adulterant, not the labeled herb.

Germander substitution in skullcap; fentanyl-adulterated kratom products; solvent residuals in concentrates.

Reporter Bias

Adverse events following media coverage are reported at higher rates. Reports cluster after publication of other reports, inflating apparent incidence.

Ephedra and kava adverse event clustering in the periods immediately following negative FDA bulletins and news coverage.

Section 5

What a Case Report Can and Cannot Establish

Action	Case Reports CAN	Case Reports CANNOT
Causation	✓ Generate a causal hypothesis worth investigating	✗ Prove that the herb caused the event
Incidence	✓ Suggest a pattern worth monitoring	✗ Establish how often this occurs per 1000 users
Mechanism	✓ Generate mechanistic hypotheses (esp. with dechallenge data)	✗ Confirm pharmacological mechanism
Pharmacovigilance	✓ Trigger regulatory review and product surveillance	✗ Characterize the full risk profile of an herb
Market safety	✓ Signal authentication failures and adulteration patterns in the market	✗ Distinguish herb risk from product quality failure
Clinical guidance	✓ Identify specific preparation forms or contexts requiring caution	✗ Support a blanket contraindication for an herb category

Case Study · Green Tea

Green Tea Is Hepatotoxic

This is the claim. Work through the analytical layers below to understand how it is simultaneously grounded in real data and systematically misleading — and what a clinically useful answer actually looks like.

Step through the evidence layers

The Claim

Case reports of hepatotoxicity have accumulated in the literature associated with green tea products. FDA, EMA, and several national agencies have issued safety communications. The phrase "green tea hepatotoxicity" appears across the regulatory record and in clinical reference databases. A student searching for herb-drug interactions will find this signal quickly.

The question is not whether reports exist. They do. The question is: hepatotoxicity from what preparation, at what dose, in what context, with what product quality?

Before you accept or reject this claim, you need to interrogate the evidence base for those four variables. A claim that skips this interrogation is not a safety determination — it is pattern recognition dressed as analysis.

The Attribution Error

The adverse event case reports that anchor "green tea hepatotoxicity" cluster in users of high-dose, concentrated EGCG extract supplements — not in drinkers of aqueous green tea infusion. The category label "green tea" absorbs a risk signal that belongs specifically to one preparation form.

The mechanism: Cases enter the literature and regulatory record tagged as "green tea." Media and secondary summaries pick up the label. The risk profile of concentrated supplements gets attributed to the herb category as a whole — including the aqueous infusion that most consumers are actually using.

This matters because it changes the risk assessment entirely:

Aqueous Infusion (Tea)

Centuries of daily use across billions of people. No credible hepatotoxicity signal at normal consumption levels.

LOW SIGNAL

Concentrated EGCG Extract

High-dose supplement taken fasted. Real but rare hepatotoxicity signal. Preparation-specific and dose-dependent.

REAL SIGNAL — CONTEXT-DEPENDENT

If you cannot name this mechanism, you can identify that the evidence is mixed — but you cannot explain why it looks more alarming than it is. That explanation is what clients need.

Dose and Rate

Two variables interact to determine hepatic exposure: total dose and rate of delivery.

Sipping tea

Distributed over 20–30 minutes. Low EGCG total. Fed state. Slow absorption rate. Peak plasma EGCG concentration: low.

Extract capsule with food

Bolus delivery. Higher EGCG total. Fed state slows absorption. Peak plasma concentration: moderate, dose-dependent.

Extract capsule, fasted

Bolus delivery. High EGCG. Fasted state substantially increases EGCG bioavailability — peak plasma concentration may breach hepatotoxic threshold.

The fasting variable is critical. Most weight-loss supplements using GTE are marketed for use on an empty stomach. This is not coincidental — fasted-state use is a major confound in the hepatotoxicity case clusters. Rate of absorption affects whether you exceed a hepatotoxic threshold at all.

The practical implication: The actionable variable for your client is not whether they use green tea — it is the form, dose, and fed/fasted context. A client taking a GTE-based weight loss supplement before breakfast is in a categorically different risk context than a client drinking green tea with meals.

Animal Data — Signal vs. Conclusion

Animal hepatotoxicity data for green tea extract exists. The NTP gavage protocol is the most cited. Here is what it actually tells you:

RouteGavage — bolus delivery directly to stomach of a fasted animal

DoseUp to 1,000 mg/kg — allometric scaling produces absurd human equivalents

Fed stateAnimals typically fasted pre-dosing — maximizes bioavailability

TranslationSpecies-specific hepatic metabolism, route differences, dose rate differences all compound

The correct use of this data: Animal gavage studies are not useless — they confirm the bolus-dose, fasted-state risk hypothesis and identify dose thresholds worth investigating. They cannot establish human incidence. Applying them to a client who drinks tea is a translation error. Applying them to a client taking high-dose GTE fasted is mechanistically more relevant, though still indirect.

Apply the same interrogation to animal data that you apply to human case reports: What was the dose? What was the route? What was the preparation form? Does this exposure scenario map onto any realistic human use pattern?

Authentication: The IPPSC Requirement

Authentication is not a regulatory checkbox. It is a logical precondition for interpreting any adverse event signal: if you do not know what was in the product, the event is unattributable.

Click each element to see what it requires:

Identity

Is this Camellia sinensis extract, confirmed by voucher specimen and/or chemical fingerprinting? Is the preparation form aqueous, or a concentrated lipid-soluble extract? Form determines what testing is even valid.

Purity

No substitutions, contaminants, or residual solvents from the extraction process. Germander has been found in products labeled as skullcap; similar substitutions cannot be assumed absent in GTE products.

Potency

Is the amount of active constituent in this product sufficient to produce the reported effect? Label claims for standardization must be verified against actual analytical data.

Strength

What percentage EGCG? A "green tea extract" standardized to 90% EGCG is not the same product as one at 45%. Batch-to-batch variability is common without 21 CFR Part 111 cGMP enforcement.

Composition

Total catechin profile, not just EGCG. Other tea polyphenols, extraction solvents, and carrier ingredients all affect the hepatic exposure profile. A paper reporting only EGCG content has composition data but is missing the other four IPPSC elements.

Key distinction: Purity and Identity are not the same thing. Purity means free from adulteration. Identity means it is actually what it claims to be — correct plant, correct preparation form. Both must be established before any toxicity signal is interpretable. A paper reporting DNA testing as identity confirmation, then noting the plant DNA was absent, is not an authentication failure — extraction methods denature DNA. Absence is expected and uninformative. Form determines what testing is even valid.

For Your Client

Academic writing represents the literature accurately; clinical communication calibrates a client's decision under uncertainty. These are different tasks. "The evidence is complex and further research is needed" is accurate as a peer-review statement. It is a failure mode in a clinical encounter.

Clinical Translation Tool — Green Tea

What is your client actually using?

Select a preparation form above

Your clinical translation will appear here.

The three-part portable answer:
Aqueous green tea infusion at normal consumption: no credible hepatotoxicity signal.
Concentrated EGCG supplements, high dose, fasted state: real but rare signal, preparation and dose dependent.
Form of preparation is the actionable variable.

Case Study · Kratom

Kratom Overdose Deaths

Case reports of kratom-associated fatalities and overdose events have accumulated in the adverse event database and clinical literature. The same analytical framework that dismantled "green tea is hepatotoxic" applies here — with some important structural differences that make this case more complex.

Step through the evidence layers

The Claim

FDA issued a public health advisory. The CDC database contains hundreds of kratom-associated reports. Several high-profile media cycles have presented kratom overdose deaths as evidence that kratom itself is acutely lethal. The phrase "kratom overdose" appears as if it is a pharmacological category equivalent to "opioid overdose."

The claim requires the same interrogation: Deaths associated with kratom use in what context? With what concomitant substances? From what authenticated product? In what population?

A systematic review of the kratom overdose literature (Stanciu et al., 2023) found that the overwhelming majority of fatal cases attributed to kratom involved concomitant substances — predominantly opioids, benzodiazepines, or both. The question of what kratom itself does pharmacologically is distinct from the question of what happens when it is used alongside these agents in a vulnerable population.

Polypharmacy: The Dominant Confound

This is structurally different from the green tea case. For green tea, the primary attribution error was preparation form. For kratom, the primary confound in the fatal case cluster is polypharmacy.

From Stanciu et al. (2023): The systematic review of kratom overdose risk found that kratom-only fatalities — cases in which kratom was the sole substance implicated — are rare. The majority of reported deaths involve multiple substances, with opioids and benzodiazepines appearing most frequently as co-ingestants.

When applying the causality framework:

Dechallenge and rechallenge are generally unavailable in fatal cases
Alternative explanations (the co-ingestants) are frequently present and uncontrolled
The temporal relationship is often ambiguous when multiple substances are co-ingested
Toxicological confirmation of other substances was not always conducted or reported

This is not exoneration. Kratom's partial opioid agonist activity at MOR is real pharmacology, and its interaction profile with opioids and benzodiazepines is a legitimate clinical concern. The point is not that kratom is safe — it is that "kratom killed this person" requires the same evidentiary standard as any other causality claim, and polypharmacy is a massive uncontrolled confound in the case cluster that has driven regulatory and media attention.

Authentication and Adulteration

The authentication problem for kratom is structurally identical to green tea — but with a higher-stakes adulterant profile.

Identity

Mitragyna speciosa from Southeast Asia — but strain nomenclature ("Maeng Da," "Red Bali") is largely marketing language with no standardized botanical definition. The IPPSC standard requires chemical fingerprinting of alkaloid profile, not just species confirmation.

Purity

Critical: Kratom products have been found adulterated with synthetic opioids including fentanyl and hydrocodone. A "kratom overdose" death in a product adulterated with fentanyl is a fentanyl death. This is not a hypothetical — it has been documented analytically.

Potency

Mitragynine and 7-hydroxymitragynine are the primary pharmacologically active alkaloids. Products vary enormously in alkaloid content depending on harvest conditions, processing, and storage. Label claims are not validated under current dietary supplement regulation.

Strength

"Kratom extract 15x" or "45x" describes extraction concentration, not alkaloid content. A consumer transitioning from plain leaf to extract may multiply their dose by 10–50x without intending to. This is a dose escalation problem, not a toxicology problem per se.

Composition

Over 40 alkaloids have been identified in M. speciosa. Reporting mitragynine only is like reporting EGCG only for green tea — it captures one constituent of a complex preparation. The pharmacological interaction profile of the full alkaloid suite is not well characterized.

When a case report of kratom overdose involves a product that was never authenticated, the signal belongs to whatever was actually in the product — not to kratom pharmacology. A toxicological finding of "kratom alkaloids present" does not rule out co-adulteration with pharmacologically active synthetic agents.

Population Selection Bias

The population using kratom in the United States is not representative of the general population, and this matters enormously for interpreting adverse event data.

A substantial proportion of kratom users are self-managing opioid withdrawal or seeking to reduce opioid consumption. This population has a higher baseline risk for opioid-related adverse events, co-ingestion of opioids and benzodiazepines, and death — regardless of kratom's pharmacology.

This is the same structure as the green tea case — but with a higher-stakes version of the confound. If a person using kratom to manage heroin withdrawal dies with kratom alkaloids in their system, the causal question requires knowing:

Were other opioids or CNS depressants also present?
Was there relapse during the period of kratom use?
Was the kratom product authenticated?
Was the death mechanism consistent with kratom's pharmacology or with an opioid/benzodiazepine toxidrome?

Reporter bias applies here too: Users who die are more likely to be in crisis and have complex substance histories. Users who successfully use kratom for harm reduction and do not die are systematically absent from the adverse event record. The denominator — total kratom users — is poorly characterized, which makes incidence estimates unreliable.

Dose and Preparation Form

Traditional kratom use in Southeast Asia involves brewing the fresh or dried leaf as a tea, consumed at culturally defined doses within a social and ritual context. This is pharmacologically different from the dominant US use patterns:

Traditional Leaf Tea

Aqueous extraction. Cultural dose constraints. Slow alkaloid absorption. Analogous to green tea infusion — low preparation risk profile relative to extract products.

LOW PREPARATION RISK

Powder (Toss-and-Wash)

Variable dose. No social constraint. Higher bioavailability than tea. User-determined dosing with no standardization. Dose escalation common with tolerance.

MODERATE — DOSE-DEPENDENT

Concentrated Extract Product

High alkaloid concentration. "X-factor" labeling is not standardized. A user transitioning from leaf to extract may increase actual alkaloid dose by orders of magnitude.

HIGH SIGNAL — EXTRACT CONTEXT

The parallel to green tea is direct: the adverse event cluster concentrates in extract and high-dose powder users, not in users of traditional leaf preparations. "Kratom is dangerous" attributes a preparation-specific and context-specific signal to the herb category as a whole.

For Your Client

Kratom exists in a complicated regulatory and clinical space. It is not a scheduled substance federally (as of 2025), but several states have restricted it. The adverse event signal is real — it is simply not distributed where the media coverage implies it is.

Clinical Translation Tool — Kratom

What is your client's use context?

Select a use context above

Your clinical translation will appear here.

Parallel to green tea — different stakes:
Traditional/low-dose leaf preparation, no CNS depressant co-use: limited credible acute fatality signal from kratom pharmacology alone.
Extract products or polypharmacy context: real and serious risk, preparation/context dependent.
Adulterated product: the risk is the adulterant — authentication is non-negotiable before any risk estimate.
Form, context, and product quality are the actionable variables.

Parallel Analysis

Green Tea vs. Kratom: Structural Parallels

The same analytical architecture applies to both cases. The variables differ; the framework is identical.

Variable	Green Tea	Kratom
Primary attribution error	Preparation form: EGCG extract risk attributed to "green tea" category	Polypharmacy: co-ingestant risk attributed to "kratom" category
Authentication confound	Adulteration, solvent residuals, IPPSC failure in case-report products	Synthetic opioid adulteration (including fentanyl) in kratom products
Dose-form gradient	Tea → extract with food → extract fasted (increasing risk)	Leaf tea → powder → extract (increasing concentration and dose uncertainty)
Population confound	Weight-loss supplement users have different health profiles from tea drinkers	Opioid withdrawal self-medication context; baseline risk is elevated independent of kratom
Animal data issue	Gavage, fasted, suprapharmacological doses; poor translation to human infusion	Rodent alkaloid studies; route and dose problems; MOR binding confirmed but overdose mechanism in humans involves polypharmacy context
Denominator problem	Billions of daily tea drinkers; supplement users much smaller but poorly characterized	Total US kratom users poorly characterized; fatal cases not adjusted for exposure prevalence
Signal assessment	Real signal in EGCG extract + fasted context; near-zero signal for aqueous infusion	Real signal in extract + polypharmacy context; signal in traditional use context is limited and poorly separated from confounders
Clinical bottom line	Preparation form is the actionable variable	Preparation form + co-ingestion context are both actionable variables

Quick Reference

Portable Checklist

Apply these eight questions to any claim that an herb is dangerous. They are not a shortcut — they are the minimum analytical standard. Click each item as you work through it.

✓
1. What type of evidence is the claim citing?
Case report? AE database count? Animal study? Observational cohort? Media coverage of another study? Each has a different evidentiary weight and a different set of validity threats.
✓
2. Was causality formally assessed?
Which method class? What data were available? Was dechallenge documented? Was rechallenge attempted (or was this clinically unacceptable)? Was the product authenticated?
✓
3. What preparation form does the evidence implicate?
Aqueous infusion vs. concentrated extract vs. standardized isolate are not interchangeable. A risk signal in one preparation cannot be assumed to transfer to another. Preparation form is frequently the hidden variable.
✓
4. Was the product authenticated to IPPSC standard?
Identity, Purity, Potency, Strength, Composition — all five elements. A paper that demonstrates only one is not authentication. Without authentication, the adverse event may belong to an adulterant or contaminant, not the labeled herb.
✓
5. What was the dose and rate of exposure?
Bolus ≠ sipping. Fasted ≠ fed. Animal gavage ≠ human oral consumption. Does the exposure scenario in the evidence map onto any realistic use pattern for your client? Rate of absorption determines peak plasma concentration and whether toxic thresholds are breached.
✓
6. Were alternative explanations adequately excluded?
Polypharmacy, concomitant medications, comorbidities, adulteration, misidentification, underlying disease. Any of these can be the actual cause of the event attributed to the herb. Were they assessed or just listed as "possible confounders"?
✓
7. Is the denominator known?
How many people use this herb without incident? Without exposure prevalence data, you cannot calculate incidence. A high absolute count of adverse event reports means nothing without a denominator. Commonly used herbs will always have more reports than rarely used ones.
✓
8. What does the signal mean for THIS client?
Is your client's use pattern the one implicated in the evidence? "The evidence is mixed" is not a clinical answer. Identify the preparation form, dose context, and co-use context specific to your client, then map the evidence to those variables. That is your clinical translation.

0 of 8 questions worked through

Causality Method Classes

Expert Judgment — Narrative probability; flexible; high inter-observer variability. Best for sparse or complex botanical cases.

Probabilistic — Bayesian likelihood; rigorous; requires data volume rarely achievable in botanical pharmacovigilance.

Algorithmic — Structured decision tree; most widely used; criterion weighting often arbitrary. Naranjo is one implementation — designed for pharmaceuticals. The method class matters more than any instrument.

IPPSC Authentication

I Identity — Correct plant, correct preparation form, voucher + chemical confirmation

P Purity — Free from adulteration, contaminants, solvent residuals

P Potency — Active constituent at clinically meaningful dose

S Strength — Concentration standardized and batch-consistent

C Composition — Full constituent profile characterized, not just one marker

A source claiming authentication that demonstrates only one element (e.g., DNA barcoding for Identity alone) does not satisfy the standard. All five must be addressed for the toxicity signal to be interpretable.

The portable three-part answer for any herb-safety claim:
(1) What does the evidence actually implicate — herb category, preparation form, or specific use context?
(2) Were causality and authentication standards met in that evidence?
(3) Does your client's actual use pattern match the implicated scenario?
Only after answering all three can you make a clinically useful risk statement.

Framework grounded in: Barnes (2003); Meyboom et al. (1997); Edwards (2017); Théophile et al. (2010, 2012); Agbabiaka et al. (2008); Stanciu et al. (2023); EFSA/EMA GTE assessments; WHO-UMC causality criteria. IPPSC as applied in botanical pharmacovigilance per USP and regulatory guidance.