Why Kratom Lab Results Differ: Repeatability, Reproducibility and Split-Sample Testing

The short answer

Two laboratory reports can disagree even when both refer to the same
production lot. The difference may begin before either laboratory
receives a sample.

A kratom lot is not necessarily uniform at every location. Two
containers may differ slightly. Two portions taken from one container
may differ. Sample grinding, mixing, extraction, calibration,
instruments, reference standards, calculations, moisture correction,
rounding and reporting conventions can add more variation. If the
laboratories do not use the same method or report on the same basis, the
two numbers may not be directly comparable at all.

This does not mean every disagreement is harmless. A large or
unexplained difference can expose a sampling failure, mislabeled sample,
unit error, unsuitable method, weak calibration, contamination,
data-transcription mistake or a result near an important limit. It
deserves a documented review.

The right first question is not “Which lab is wrong?” It is:

Are these reports measuring the same analyte in comparable portions, with comparable methods, units and reporting bases?

Only after that question is answered should a reviewer compare the
numerical results.

A COA is a
measurement report, not the product itself

A certificate of analysis, or COA, records results for the sample the
laboratory received and tested. It is not a direct measurement of every
gram, capsule or bottle in the commercial lot.

That distinction creates an evidence chain:

production lot → sampled locations → combined or individual samples → laboratory sample → prepared test portion → instrument response → calculation → reported result

A problem or source of variation can enter at every arrow.

For example, imagine that a 500-kilogram powder lot is held in 25
lined drums. A single grab from the top of one drum is sent to
Laboratory A. A separate grab from the bottom of another drum is sent to
Laboratory B. The reports share a lot code, but the laboratories did not
receive equivalent samples. Calling the work “split-sample testing”
would be misleading.

A sound comparison begins with traceability: who sampled, where,
when, how much, under which plan, with what tools, how the material was
mixed or divided, which seals were applied and whether both laboratories
received portions from the same homogenized parent sample.

FDA’s dietary-supplement current good manufacturing practice
framework requires representative samples in specified circumstances and
requires laboratory methods to be appropriate for their intended use and
scientifically valid. FDA also states that kratom is not lawfully
marketed as a dietary supplement. Kiody should therefore treat 21
CFR Part 111 as a useful quality-system reference, not as proof that
a kratom product is FDA approved or lawfully marketed.

Seven reasons two results
may differ

1. The submitted
samples were not equivalent

This is often the most important possibility.

Botanical powder can segregate during handling. Particle size,
density, static charge, vibration, container filling and settling can
change where different particles accumulate. A composite sample may
average several locations, while a grab sample reflects only one
location. Capsules introduce additional questions about blend uniformity
and fill weight. Liquid extracts can stratify or precipitate if the
sampling instructions do not address mixing.

Even excellent laboratories cannot repair an unrepresentative
sampling plan after the fact.

Useful questions include:

• Did both results refer to the same defined lot?
• Were the source containers identified?
• Was a written sampling plan followed?
• Were increments collected from multiple locations?
• Was the parent sample homogenized before it was divided?
• Did the sampler use a splitter designed to avoid selective particle
  loss?
• Were reserve portions sealed at the same time?
• Were temperature, moisture and transit conditions controlled where
  relevant?

“Same lot” is not the same as “same sample.” “Same container” is not
the same as “same test portion.”

2. Sample preparation
differed

The laboratory rarely places the submitted powder directly into an
instrument. It may first grind, sieve, mix, weigh, extract, dilute,
filter, centrifuge or derivatize a test portion.

Differences in these steps can affect recovery. Examples include:

• one laboratory mills the sample more finely;
• different test-portion masses are weighed;
• extraction solvents or pH conditions differ;
• extraction time, temperature or agitation differs;
• one method uses one extraction and another uses repeated
  extraction;
• one laboratory corrects reference-standard purity while another
  calculation is unclear;
• filtration retains or releases different material;
• a high-concentration extract requires a larger dilution than leaf
  powder;
• the analyte changes during poorly controlled preparation or
  storage.

A method name such as “LC-MS/MS” describes an instrument family, not
the complete procedure. Two LC-MS/MS methods may use different sample
preparation, columns, mobile phases, transitions, calibration ranges,
internal standards and acceptance criteria.

3. The methods measure
differently

A method should be fit for the product matrix, analyte and
concentration range. Leaf powder, pure-leaf capsules, resinous extracts,
sweetened liquids and enhanced products are not interchangeable
matrices.

Matrix components can suppress or enhance an instrument signal. A
method validated for botanical powder may not perform the same way with
a concentrated extract or a gummy. A broad screening method may identify
an analyte but not quantify it as reliably as a validated quantitative
method. A chromatographic peak may include an unresolved interferent if
selectivity is inadequate.

FDA’s Validation
and Verification of Analytical Testing Methods guidance for
tobacco-product testing is not a kratom rule, but its analytical
concepts are useful: validation considers accuracy, precision,
selectivity, sensitivity, linearity, range and robustness, while
verification demonstrates that a laboratory can meet method-performance
expectations. Those concepts help explain why an instrument label alone
is not enough.

When methods differ, request the method identifier and revision,
intended matrix, analyte definition, calibration range, recovery data,
precision data, limits of detection and quantitation, and relevant
quality-control acceptance criteria. Proprietary details may be
protected, but a defensible report should still disclose enough to
evaluate fitness for purpose.

4. Normal measurement
variation is present

Repeated measurements do not usually produce infinitely identical
numbers.

Small differences can arise from balances, volumetric equipment,
temperature, extraction recovery, instrument response, calibration
fitting and analysts. A result is best understood as an estimate
produced by a defined measurement process—not as a perfect statement of
a product’s “true value.”

This is where repeatability, intermediate precision, reproducibility
and measurement uncertainty matter. They are related, but they do not
mean the same thing.

5. Units or reporting bases
differ

Two reports may look contradictory when they are mathematically
consistent.

Common units include:

• percent by weight;
• milligrams per gram (mg/g);
• micrograms per gram (µg/g);
• parts per million (ppm);
• milligrams per milliliter (mg/mL);
• milligrams per serving;
• colony-forming units per gram (CFU/g).

For a solid product on the same reporting basis:

• 1% = 10 mg/g
• 0.1% = 1 mg/g = 1,000 µg/g = 1,000 ppm
• 0.01% = 0.1 mg/g = 100 µg/g = 100 ppm

But a dry-weight result and an as-received result are not
automatically interchangeable. A liquid result in mg/mL
cannot be converted to percent by weight without appropriate density and
basis information. “Per serving” depends on the declared or measured
serving size. A percentage of total alkaloids uses a different
denominator from a percentage of total product weight.

Never compare only the digits.

6. Detection
limits and reporting rules differ

ND, <LOD, <LOQ, zero and
“not tested” have different meanings.

• ND generally means not detected under the stated
  method and reporting conditions. It does not prove absolute
  absence.
• Below LOD means the signal did not meet the
  method’s stated detection criterion.
• Below LOQ usually means the analyte may be detected
  but cannot be quantified with the method’s stated performance.
• Zero is a numerical value and should not be
  substituted for a censored result unless a defined statistical procedure
  requires and explains that substitution.
• Not tested means no result was generated for that
  analyte.

Laboratory A might report 7-OH as <0.01%; Laboratory
B might quantify 0.006%. Those entries do not necessarily
conflict. Laboratory B may simply have a lower quantitation limit.

Likewise, one microbiology laboratory might report
<10 CFU/g and another <100 CFU/g. The
reporting limits differ by one order of magnitude. Treating both as zero
hides useful information.

7. One or both reports
contain an error

Errors do occur. Possibilities include:

• sample or lot misidentification;
• transcription or decimal-placement error;
• incorrect unit conversion;
• calculation formula error;
• expired or incorrectly characterized reference material;
• calibration or quality-control failure;
• carryover or contamination;
• instrument integration error;
• wrong method revision;
• use of a method outside its validated range;
• unauthorized data replacement or incomplete reporting.

The response should be evidence-based. Do not accuse a laboratory of
fraud merely because two numbers differ. Request an amended report,
raw-data review, quality-control review and documented investigation
when the facts warrant them.

Repeatability,
intermediate precision and reproducibility

These terms describe variation under different conditions.

Repeatability

Repeatability asks how closely results agree when conditions are kept
as similar as practical: same method, same laboratory, same equipment,
same analyst or tightly controlled analyst conditions, and a short time
interval.

For example, one analyst prepares and measures six test portions from
a well-mixed laboratory sample in one run. The spread among those values
estimates one aspect of repeatability.

Repeatability does not reveal all variation. It may exclude different
days, analysts, instruments, reagent lots, calibrations and
laboratories.

Intermediate precision

Intermediate precision broadens the conditions within one laboratory.
The lab may repeat work on different days, with different analysts,
instruments, columns or reagent lots.

This can be more informative for routine performance than one tightly
controlled repeatability run because ordinary laboratory work happens
over time.

Reproducibility

Reproducibility examines agreement under broader changed conditions,
especially between laboratories. It can include different analysts,
instruments, environments and method implementations.

Reproducibility is normally worse—that is, more variable—than
repeatability because it includes more sources of variation. That does
not make the data defective. It means the comparison reflects a larger
measurement system.

FDA’s chemical-method glossary describes multi-laboratory validation
as a collaborative study in which repeatability and reproducibility are
measured in at least two laboratories. The FDA analytical-method
glossary is a useful terminology source, though it does not create a
kratom-specific acceptance limit.

Precision is not accuracy

Results can be tightly clustered and still be biased.

Suppose a laboratory reports six replicate values of
1.51%, 1.50%, 1.51%, 1.50%, 1.51% and 1.50%. The
repeatability looks excellent. But if a calculation error causes every
value to be 10% high, the method is precise without being accurate.

The reverse can also occur. A mean may be close to an accepted
reference value while individual replicates are widely dispersed. The
average looks accurate, but the precision is weak.

Laboratory quality review therefore considers multiple
characteristics:

• accuracy or trueness: closeness to an accepted
  reference or recovery expectation;
• precision: closeness among repeated results under
  defined conditions;
• selectivity: ability to distinguish the analyte
  from interferences;
• linearity and range: whether response supports the
  concentrations tested;
• sensitivity: whether relevant levels can be
  detected and quantified;
• robustness: resilience to small, deliberate
  procedural variations;
• measurement uncertainty: a quantified description
  of doubt associated with the result.

No single r², recovery percentage or duplicate result
proves the whole method is suitable.

Measurement
uncertainty: when do results truly conflict?

Measurement uncertainty helps describe the range of values reasonably
attributable to a measured quantity under defined assumptions. It is not
a mistake allowance and does not make every result acceptable.

Consider two fictional dry-weight mitragynine results:

• Laboratory A: 1.42% ± 0.10%
• Laboratory B: 1.50% ± 0.12%

If both statements use comparable expanded-uncertainty conventions,
their intervals overlap. The difference between the reported centers may
be unsurprising.

Now consider:

• Laboratory A: 1.10% ± 0.05%
• Laboratory B: 1.70% ± 0.06%

That gap is much harder to explain through the stated analytical
uncertainties alone. But it still does not prove which laboratory is
right. Sampling heterogeneity, reporting-basis differences or an
unrecognized systematic bias may dominate.

NIST’s Technical Note
1297 explains the U.S. approach to evaluating and expressing
measurement uncertainty. A laboratory should disclose the uncertainty’s
coverage factor or confidence convention, the measurand and whether
sampling uncertainty is included. Two bare “±” values may not mean the
same thing.

Near a legal or internal limit, uncertainty and the decision rule
become especially important. A business should not improvise a pass/fail
rule after seeing a result. The rule should define in advance how
uncertainty is considered, what happens in a guard band, who reviews an
indeterminate or near-limit result and which authority controls.

Relative
percent difference: useful, but easy to misuse

For two positive results measuring the same quantity on the same
basis, a reviewer may calculate relative percent difference (RPD):

RPD = |A − B| ÷ ((A + B) ÷ 2) × 100

If Laboratory A reports 1.40% and Laboratory B reports
1.50%:

RPD = 0.10 ÷ 1.45 × 100 = 6.9%

RPD describes the size of the difference relative to the two-result
mean. It does not decide whether the difference is acceptable.

An acceptance criterion must be connected to method performance,
sampling variation, product risk, specification distance and the
comparison’s purpose. A universal “10% rule” for all kratom analytes and
matrices would be invented.

RPD becomes unstable near zero. It also should not be calculated by
replacing ND or <LOQ with zero unless a
documented procedure justifies the treatment. For microbiological
counts, logarithmic comparisons may be more informative than arithmetic
RPD. For qualitative results, a presence/absence agreement framework is
needed instead.

What split-sample
testing really means

A defensible split-sample study starts with one parent sample that is
made as homogeneous as practical and then divided into equivalent,
traceable portions.

It does not mean collecting unrelated grabs from the same commercial
lot and mailing them to different laboratories.

A practical split-sample
design

1. Define the question. Specify the product, lot,
   analyte, decision threshold and reason for the comparison.
2. Quarantine when appropriate. If the dispute affects
   lot release or a serious specification, prevent distribution under the
   applicable quality procedure while the issue is reviewed.
3. Review the original sampling history. Identify
   whether the disagreement may reflect nonequivalent samples.
4. Choose an independent, qualified sampler. Avoid
   allowing either laboratory to selectively collect its preferred
   portion.
5. Create a representative parent sample. Follow a
   written plan suited to the product format and lot configuration.
6. Homogenize appropriately. Mixing should reduce
   variation without changing the analyte or contaminating the sample.
7. Divide with a defensible technique. For powders, a
   suitable sample splitter may be more reliable than alternately spooning
   portions.
8. Prepare at least three sealed portions. Send one to
   each laboratory and retain an adjudication portion under controlled
   conditions.
9. Blind where practical. Use neutral sample
   identifiers so prior results do not influence handling or
   interpretation.
10. Align the measurement plan. Decide whether both
    labs should use the same validated method or whether the study
    intentionally compares routine methods.
11. Predefine the comparison rule. State how units,
    moisture basis, uncertainty, censored values and acceptance criteria
    will be handled before results return.
12. Reconcile complete reports. Review methods, sample
    receipt, condition, preparation, quality controls, calculations,
    uncertainty and deviations—not only the final number.

The retained portion matters. If two initial reports remain
irreconcilable, a qualified third laboratory may test it under a
predefined adjudication plan. The third result should not automatically
be treated as a majority vote. Three laboratories can share a bias, and
the retained portion may not be equivalent if division was poor.

Same method or different
methods?

Both designs can be useful, but they answer different questions.

Same-method comparison

If both laboratories follow the same fully specified method, the
comparison emphasizes laboratory implementation and reproducibility. It
can reveal differences in analyst execution, equipment, calibration,
environmental conditions or quality controls.

This works only when “same method” truly means the same revision,
sample preparation, calculations and acceptance rules. Naming the same
published method while making unrecorded modifications weakens the
comparison.

Routine-method comparison

If each laboratory uses its ordinary validated method, the comparison
reflects the results buyers are likely to receive in practice.
Differences can arise from both laboratory execution and method
design.

This approach may be useful for supplier qualification or risk
assessment, but it is harder to diagnose. A difference cannot be
attributed to a laboratory alone when the methods are not
harmonized.

Orthogonal confirmation

Sometimes a different measurement principle can help investigate
interference or identity. An orthogonal method is not simply a second
run on another instrument of the same type. It brings sufficiently
different selectivity or measurement logic to address the suspected
problem.

Orthogonal results still require a fit-for-purpose method, compatible
measurand and valid comparison. A qualitative identity technique does
not adjudicate a quantitative potency disagreement.

Proficiency
testing and interlaboratory comparisons

Proficiency testing, or PT, evaluates a laboratory’s performance
using samples and an assigned or consensus value under a defined
program. It can reveal bias, calculation problems, method weaknesses,
contamination or training needs.

PT is valuable evidence, but it has boundaries:

• participation is not the same as successful performance;
• one successful round does not prove every analyte, matrix and
  concentration is handled well;
• a generic botanical PT may not cover kratom;
• a kratom alkaloid PT may not cover heavy metals, pathogens or
  manufactured derivatives;
• an assigned value may depend on a reference method, reference
  laboratory or participant consensus;
• a PT score evaluates the submitted PT result, not the quality of
  every commercial lot;
• PT does not replace routine blanks, spikes, duplicates, controls,
  calibration checks or method validation.

NIST describes proficiency testing as a form of interlaboratory
comparison and notes that PT design depends on the test item, method and
number of participants. NIST materials also explain that unacceptable PT
performance requires investigation within an accreditation program. See
NIST
Handbook 150 for the general framework.

When reviewing PT evidence, ask:

• Was the laboratory enrolled for the specific analyte and
  matrix?
• What dates and rounds are covered?
• What score or acceptance criterion was used?
• Did the laboratory pass?
• Were any unacceptable or questionable results investigated?
• Does the tested concentration resemble the commercial decision
  range?
• Does the PT apply to the location and method shown on the COA?

An ISO/IEC 17025 accreditation certificate may be useful, but the
technical scope matters. Accreditation does not mean every test offered
by a laboratory is accredited. Verify the laboratory address, analyte,
matrix, method and current scope with the accreditation body.

Five fictional comparison
examples

These examples are educational and do not describe Kiody products or
real laboratories.

Example 1: different
units, same result

Laboratory A reports mitragynine at 1.35% w/w.
Laboratory B reports 13.5 mg/g.

For the same solid-product basis, these are equivalent because
1% = 10 mg/g. There is no numerical disagreement after
conversion.

Review point: confirm both are as-received or both are dry-weight
results before declaring agreement.

Example 2:
detection limit explains the apparent conflict

Laboratory A reports 7-OH as <0.010%. Laboratory B
reports 0.006%.

Laboratory A did not report zero. Its method did not quantify below
0.010% under the stated conditions. Laboratory B’s result
fits beneath Laboratory A’s reporting limit.

Review point: obtain LOD, LOQ and reporting-limit definitions. Do not
relabel Laboratory A’s result “none.”

Example
3: two grabs reveal lot heterogeneity, not lab performance

A supplier fills 40 drums. A top grab from Drum 1 goes to Laboratory
A; a bottom grab from Drum 40 goes to Laboratory B. Alkaloid results
differ by 22% RPD.

The study cannot isolate analytical reproducibility because the
samples may differ. A representative parent sample and controlled
division are needed.

Review point: investigate blending, filling and sampling before
assigning blame to either laboratory.

Example
4: dry-weight correction changes a near-threshold result

Laboratory A reports an analyte at 380 ppm as received.
Laboratory B reports 422 ppm dry weight. The product
contains 10% moisture.

A simplified conversion of Laboratory A’s result gives:

380 ppm ÷ (1 − 0.10) = 422 ppm dry weight

The reports may reflect the same concentration on different
bases.

Review point: use the moisture value linked to the tested sample and
confirm the laboratory’s calculation. Do not assume a generic moisture
percentage.

Example 5:
matching numbers hide a weak study

Two laboratories both report 1.4% mitragynine. One
report rounds from 1.35%; the other rounds from
1.44%. Neither provides uncertainty, and the samples were
unrelated grabs.

The displayed values match, but the evidence does not prove strong
reproducibility or lot uniformity.

Review point: apparent agreement is only as meaningful as sampling,
method and reporting detail.

A 12-step
workflow for reviewing conflicting COAs

Step 1: Preserve the evidence

Save the original signed reports, attachments, sample-submission
forms, chain-of-custody records, emails and any machine-readable
exports. Record when each file was received. Do not overwrite a report
with an amended version.

Step 2: Confirm document
authenticity

Contact the laboratory through independently verified contact
information. Confirm the report number, sample ID, issue date, revision
and authorized signatory. A QR code is convenient, but it is not the
only authenticity control.

Step 3: Match the product
identity

Compare product name, form, lot code, batch number, package size,
manufacturing date and sample description. Watch for parent-lot and
finished-product lot codes that look related but are not identical.

Step 4: Reconstruct sampling

Document who selected each sample, the number and location of
increments, parent-sample preparation, division method, seals, custody
transfers and storage. Decide whether the samples were truly
comparable.

Step 5: Align analyte
definitions

Confirm the chemical identity and form. “Total alkaloids,” “total
mitragynines,” “7-OH-related compounds” and a named analyte are not
interchangeable. Check whether salts, hydrates, stereoisomers,
degradants or derivatives are included.

Step 6: Normalize units and
bases

Convert only when the necessary mass, volume, density, serving and
moisture information exists. Preserve the original values and
calculations. Never convert a percentage of total alkaloids into a
percentage of product weight without an actual total-alkaloid
measurement and defined denominator.

Step 7: Compare method scope

Record method identifiers and revisions, matrix, preparation,
instrument, calibration range, LOD, LOQ, reporting limit and
accreditation status. Note whether a laboratory modified a published
method.

Step 8: Review quality
controls

Ask whether blanks, duplicates, spikes, certified reference
materials, calibration verification, system-suitability checks and
continuing controls met their criteria. A final COA may not display all
of this information.

Step 9: Evaluate
uncertainty and precision

Request available uncertainty and precision information. Determine
whether the reported uncertainty includes sampling. Calculate RPD only
when appropriate and compare it with predefined, scientifically
justified expectations.

Step 10: Look for simple
errors

Recalculate unit conversions, dilution factors, dry-weight
corrections and rounding. Confirm decimal placement and transcribed
sample IDs. Simple mistakes deserve correction before an expensive
retest begins.

Step 11:
Use a predefined retest or split-sample plan

Do not repeatedly test until a preferred answer appears. Define the
number of portions, laboratories, methods, acceptance logic and
disposition authority in advance. Preserve all valid results.

Step 12: Make
and document the quality decision

State what the evidence establishes, what remains uncertain and what
action follows. Possibilities include release, continued hold,
rejection, supplier investigation, resampling, method investigation,
label correction or recall review. A third result should not erase two
earlier results without a scientifically justified explanation.

Twenty warning
signs in a laboratory comparison

Pause and investigate when any of these appear:

1. The reports have the same lot code but no sampling records.
2. Samples came from different containers or locations without a
   representative plan.
3. “Split sample” is claimed, but no parent-sample homogenization or
   division is documented.
4. One result is dry weight and the other is as received.
5. A liquid mg/mL result is compared with a solid
   mg/g result without density or formulation data.
6. A percentage of total alkaloids is treated as a percentage of
   product weight.
7. ND is entered as zero.
8. Different LODs or LOQs are ignored.
9. Method names are limited to instrument acronyms.
10. The method was validated for leaf but applied to an extract, gummy
    or liquid without matrix verification.
11. The result sits outside the calibration range.
12. Dilution factors or purity corrections cannot be reconstructed.
13. Quality-control failures are omitted from the discussion.
14. An amended report replaces the original without a revision
    history.
15. Only favorable retests are retained.
16. The laboratory address does not match the accredited location.
17. The accreditation scope does not include the analyte, matrix or
    method.
18. A PT participation claim does not disclose the analyte, matrix,
    round or outcome.
19. A third laboratory is treated as an automatic deciding vote.
20. A numerical match is presented as proof that every unit in the lot
    is identical.

One warning sign does not automatically invalidate a result. It
identifies a question that should be resolved before the report supports
a high-consequence decision.

Comparison record template

A review record can include the following fields:

Product and lot

• review record number;
• product name and format;
• declared ingredients;
• production-lot identifier;
• finished-product lot identifier;
• batch size and container count;
• manufacturing and packaging dates;
• storage conditions;
• applicable specification or legal threshold;
• reason for comparison.

Sampling and custody

• sampler name and organization;
• sampling date and location;
• written sampling-plan identifier;
• source containers and locations;
• increment count and mass;
• parent-sample mass;
• homogenization procedure;
• splitting equipment and procedure;
• split-portion identifiers and masses;
• seal numbers;
• reserve-portion location;
• custody transfers;
• shipping and receipt condition.

Laboratory and method

• laboratory legal name;
• testing-site address;
• report and sample numbers;
• accreditation body and certificate number;
• applicable scope entry;
• method identifier and revision;
• matrix within method scope;
• analyte and measurand definition;
• sample-preparation summary;
• instrument platform;
• calibration range;
• reference-standard identity and purity correction;
• LOD, LOQ and reporting limit;
• quality-control results and acceptance;
• deviations or amended-report history.

Results and reconciliation

• original result and unit;
• as-received or dry-weight basis;
• moisture result and method;
• normalized result and calculation;
• unrounded result if available;
• uncertainty statement and coverage factor;
• duplicate or replicate data;
• RPD or other predefined comparison statistic;
• predefined acceptance criterion;
• laboratory explanation;
• investigation findings;
• third-laboratory plan, if used;
• disposition decision;
• quality reviewer and date;
• legal-review status when a controlled-substance threshold is
  involved.

This record is not a universal regulatory form. It is a practical
starting point that should be adapted to the product, risk and governing
requirements.

Product type changes the
comparison

Whole leaf and powder

Particle size, stem content, vein fragments, moisture and segregation
may affect representativeness. A sampling plan should address multiple
locations and suitable homogenization without changing the material.

Pure-leaf capsules

Testing pooled capsule contents may characterize a composite but
conceal capsule-to-capsule variation. Testing individual units addresses
a different question. Total capsule weight is not the same as botanical
fill weight.

Extracts

Extraction ratio claims such as 10:1 do not define
measured alkaloid concentration. Resin, powder and liquid matrices may
require different preparation and dilution. Compare actual analyte
results on compatible bases.

Enhanced products

Adding extract or isolated alkaloids to leaf can increase segregation
risk. A composite result may not reveal localized high-concentration
units. Uniformity and intentional enhancement should be evaluated
separately.

Concentrated 7-OH products

Concentrated or intentionally enriched 7-OH products should not be
treated as ordinary botanical leaf. Near a legal threshold, sample
identity, dry-weight or other statutory basis, uncertainty and decision
rules can materially affect the conclusion. Kiody does not sell
concentrated 7-OH.

MGPI, MGM-15 and MGM-16

These compounds are separate from ordinary mitragynine and 7-OH. An
assay for mitragynine does not establish their absence. The method must
specifically address the named analyte at a suitable reporting limit and
in the relevant matrix.

Federal status note
as of September 3, 2026

Federal status must be read by compound and action.

HHS’s threshold proceeding for 7-OH above a specified level remains
pending. The agency extended public comments through September
10, 2026. The notice is an information request concerning a
proposed threshold; it is not a final scheduling order for 7-OH. See the
August
26 HHS extension notice.

By contrast, DEA’s temporary Schedule I order for mitragynine
pseudoindoxyl, MGM-15 and MGM-16 took effect August 26,
2026. The order is separate from the pending 7-OH threshold
proceeding. See the DEA
temporary scheduling order.

DOJ updated its announcement on September 1, 2026 to
state that it will exercise enforcement discretion when only incidental
trace MGPI is confirmed in a product otherwise consistent with botanical
kratom. DOJ expressly says this is not a legal exemption and does not
change MGPI’s Schedule I status. The policy does not apply to MGM-15,
MGM-16, or manufactured, concentrated, fortified or intentionally added
MGPI. See the DOJ
announcement and September 1 update.

This makes laboratory comparability especially important. “Trace,”
“detected,” “below LOQ” and a quantified concentration are not
interchangeable, and DOJ announced no numerical trace threshold. A
business should not convert enforcement discretion into a product claim
or legal safe harbor.

State and local rules may be more restrictive and may use different
analytes, thresholds and denominators. Check the current law for the
destination and do not infer legality from a passing COA alone.

Frequently asked questions

Why
did two laboratories report different mitragynine percentages?

Possible causes include nonequivalent samples, natural lot variation,
different moisture bases, sample preparation, calibration, extraction
recovery, method bias, rounding or error. Compare the full evidence
chain before deciding.

Which result should I trust?

Trust should follow evidence, not the higher or more favorable
number. Evaluate sample representativeness, chain of custody, method
fitness, accreditation scope, quality controls, uncertainty, report
authenticity and investigation history.

Does the
lower result automatically win for safety?

No. Choosing the lower number is not a scientific decision rule. A
conservative action may be appropriate in some circumstances, but the
quality decision should be defined, documented and connected to the
applicable risk and requirement.

Does
the higher result automatically control a legal threshold?

Not universally. The governing law, official method requirements,
sampling rules, uncertainty policy and enforcement authority matter.
Place affected material on hold and obtain qualified legal and technical
review.

Is retesting
allowed after a failing result?

Retesting can be part of a written investigation, but repeated
testing should not be used to test a failure into compliance. Preserve
the original result, investigate scientifically assignable causes and
follow a predefined procedure.

What is testing into
compliance?

It is the practice of generating repeated results and selectively
relying on favorable ones without a justified investigation and
predefined rule. It distorts the evidence.

Should I average two
laboratory results?

Not automatically. An average can hide nonequivalent samples, method
bias or a genuine failure. Averaging is appropriate only when the study
design and decision procedure support it.

Is a third laboratory the
tiebreaker?

Not by simple majority vote. A third laboratory can add evidence if
it tests a properly retained equivalent portion under a predefined plan.
Its result must be reviewed with the first two.

What is a blind sample?

A blind sample uses an identifier or submission design that withholds
information that could influence handling or interpretation. Blindness
does not correct poor sampling or an unsuitable method.

What is a duplicate?

The word can refer to field duplicates, laboratory sample duplicates,
preparation duplicates or replicate instrument measurements. The report
should state which kind, because each captures different sources of
variation.

Can
two results be statistically different but both pass?

Yes. A product specification and a statistical comparison answer
different questions. Both values may meet the limit even when their
difference signals a sampling or process issue.

Can two results agree
but both be wrong?

Yes. Shared reference-standard errors, common method bias, similar
matrix interference or identical unit-conversion mistakes can produce
agreement without accuracy.

Does
ISO/IEC 17025 accreditation guarantee identical results?

No. Accreditation is evidence of assessed competence within a defined
scope. It does not eliminate measurement variation or guarantee that
every test, analyte and matrix is included.

Does proficiency
testing certify my product?

No. PT evaluates a laboratory’s performance for a defined challenge.
It does not certify a commercial lot, prove every method performs
correctly or establish product safety.

What does “same lot” prove?

It connects reports to a production identifier if the records are
authentic. It does not prove the submitted samples were representative,
equivalent or handled the same way.

Why does moisture basis
matter?

Removing water from the denominator increases a dry-weight
concentration relative to the corresponding as-received value. The
difference may be material near a threshold.

Is ND the same as
zero?

No. ND means the method did not detect the analyte under stated
conditions. The analyte may be present below the method’s detection
capability.

Is 1.4% more precise
than 1.40%?

The extra digit implies finer reporting, but it does not guarantee
better measurement. Meaningful precision depends on method performance
and rounding rules.

Can a COA prove a product is
legal?

No. A COA may provide evidence about a submitted sample. Legal status
also depends on product category, ingredients, concentration basis,
destination, current federal/state/local law and sometimes licensing or
labeling.

Does Kiody sell concentrated
7-OH?

No. Kiody does not sell concentrated 7-OH, MGPI, MGM-15 or MGM-16 and
limits its educational and botanical-leaf content to adults 21 and
older.

Sources and further reading

1. FDA:
   FDA and Kratom — current federal agency position, product-marketing
   status and safety warnings.
2. 21
   CFR Part 111 — representative samples, scientifically valid methods,
   laboratory controls and records within the dietary-supplement CGMP
   framework.
3. FDA: Validation
   and Verification of Analytical Testing Methods — analytical concepts
   including accuracy, precision, selectivity, sensitivity, linearity,
   range and robustness.
4. FDA
   analytical-method glossary — terminology for multi-laboratory
   validation, repeatability and reproducibility.
5. NIST
   Technical Note 1297 — U.S. guidance for evaluating and expressing
   measurement uncertainty.
6. NIST
   Handbook 150 — general accreditation, proficiency-testing and
   interlaboratory-comparison concepts.
7. HHS:
   7-OH threshold proceeding comment-period extension — comments
   extended through September 10, 2026.
8. DEA:
   Temporary Schedule I placement of MGPI, MGM-15 and MGM-16 —
   effective August 26, 2026.
9. DOJ:
   Emergency scheduling announcement, updated September 1, 2026 —
   enforcement-discretion statement for incidental trace MGPI and its
   limitations.