In the twelve months to December 2025, United States provisional data recorded 1,221 drug overdose deaths involving mitragynine, kratom’s main alkaloid. In the same window it recorded 36,627 involving fentanyl.
The word doing the work in that sentence is involving. In the data it comes from, a drug is recorded as involved in a death when the drug is named somewhere in the text of the death certificate. The National Center for Health Statistics says so in terms: substances mentioned in those text fields “are assumed to be involved in the death unless contextual information indicates otherwise.”
So 1,221 is a count of death certificates on which somebody wrote the word. It is not a count of deaths that kratom caused, and the agency publishing it does not claim that it is.
Both sides of the public argument depend on blurring this. “Kratom has killed more than a thousand people a year” is not supported by the figure. Neither is “kratom has never killed anyone”. What follows is what the records actually say, and where they stop.
How a drug gets counted as involved
Overdose deaths in American statistics are found by a code, and the drugs inside them are found by reading.
The code comes first. A death is a drug overdose death when its underlying cause falls in a defined set of categories in the tenth revision of the International Classification of Diseases: X40 to X44 for unintentional, X60 to X64 for suicide, X85 for homicide, and Y10 to Y14 where intent is undetermined. That decides whether a death is an overdose at all. It says nothing about which substance.
Which substance is a separate question, and for most of the drugs people want counted the classification has no code of its own. So the NCHS searches the free text of the certificate instead: the causes of death in Part I, the contributing conditions in Part II, and the description of how the injury happened. The method is called Drugs Mentioned with Involvement, and it maps spellings found in that text onto a list of referent groups.
Three consequences follow, and all three are printed in the agency’s own technical notes.
A mention is treated as involvement by default. The certifier has to actively indicate otherwise for a named drug to be excluded.
The categories are not mutually exclusive. A certificate naming fentanyl and mitragynine is counted once under each, so the totals for different drugs cannot be added together.
The quality of the input varies by jurisdiction, because death investigation in the United States is not one system. The notes say regional differences in the completeness of investigations “must be considered when reviewing these findings.”
What the mitragynine counts look like
The published series runs monthly from January 2019, as twelve-month rolling totals, and the version read here was current as of 5 July 2026. The twelve months ending January 2019 carried 645 deaths involving mitragynine. The highest twelve-month total in the series is 1,684, reached at a plateau across the years ending December 2021 and January 2022. The most recent is 1,221, for the year ending December 2025.
That shape is worth stating plainly and then leaving alone. The count roughly tripled to 2022 and has fallen by about a quarter since. Some of that is real change in what people consume. Some of it is change in what laboratories test for and what certifiers write down, and the published data cannot separate the two. The agency’s own caveat is that counts by specific drug “are subject to change as additional search terms for drugs are identified.”
The search terms collapse leaf and concentrate into one number
The list of terms the method looks for is published as a table, and reading it changes how the mitragynine figure should be understood.
There is one referent group, MITRAGYNINE, and twenty-four rows map into it, twenty-three of them distinct (one misspelling is listed twice). Most are misspellings of mitragynine and kratom collected from real certificates: MITRAGYNNE, MITRAGYRINE, KRATAM, KRATON. Two of the terms are HYDROXYMITRAGYNINE and 7-HYDROXYMITRAGYNINE.
That means the national count does not distinguish a death certificate mentioning kratom leaf from one mentioning 7-hydroxymitragynine, the far more potent compound sold in concentrated products and now the subject of federal scheduling activity. Both land in the same number. The distinction that the entire American regulatory fight turns on, which we set out separately, is not visible in the mortality statistics at all.
One other entry in that table is worth naming, because it affects how tightly the number should be read. MIRTZAPINE also maps to MITRAGYNINE, evidently as a misspelling of mitragynine. Mirtazapine is a widely prescribed antidepressant, and it has no referent group of its own in this table, which covers twelve drugs rather than all of them. A certifier who wrote “mirtzapine” meaning the antidepressant would have that death counted as a mitragynine mention. We do not know how often that happens and neither, from the published data, can anyone else.
What happens when somebody reads the reports
The national count is an index. These studies are where the question of causation is actually examined, and four of them are worth setting out. All are human observational work near the bottom of the evidence ladder: mostly case series and record reviews, which can establish that something occurred and cannot establish how often. The exception is the Florida study set out below, a case-control design, which is a stronger observational structure than the rest.
CDC, 27 states, July 2016 to December 2017. This is the study that separated the two words most cleanly, because it recorded both. Of 27,338 overdose deaths entered into the State Unintentional Drug Overdose Reporting System, 152 decedents, or 0.56%, tested positive for kratom on postmortem toxicology. A medical examiner or coroner determined kratom to be a cause of death for 91 of those 152. Seven of the 91 had kratom as the only substance detected, and the authors immediately qualify even that: the presence of additional substances cannot be ruled out, because testing protocols were not documented and varied between and within states. Among the kratom-positive decedents, fentanyl or a fentanyl analogue was listed as a cause of death in 65.1%, heroin in 32.9%, benzodiazepines in 22.4%, prescription opioids in 19.7% and cocaine in 18.4%. About 80% had a recorded history of substance misuse.
Oklahoma, 2022 to 2023. Twenty-eight death investigations with quantified mitragynine. Mitragynine was listed in the cause of death in 22 of them, and 21 of those 22 also had at least one other drug reported on the death certificate. Heart blood concentrations ran from 130 to 2,600 ng/mL, femoral blood from 130 to 7,500 ng/mL.
Florida, five years. A case-control design, which is the most useful structure anyone has applied to this question. It compared 38 deaths in which mitragynine was listed as a cause with 111 in which it was detected but not deemed causal. Every one of the 111 had co-exposures, including fentanyl in 84, cocaine in 49 and ethanol in 29. The 38 attributed deaths had no toxic levels of any other substance, and a much higher mean blood mitragynine concentration. The authors read this as a concentration-dependent association with fatal toxicity.
Kern County, California, 2020. Of 214 accidental opioid-associated overdose deaths in one county’s coroner reports, four had mitragynine. All four had at least one other substance; fentanyl was present in three.
The number in the Florida abstract does not add up
We are citing that Florida study and we are also going to say what is wrong with how its result is printed, because a reader checking our work would find it.
The published abstract gives the mean blood mitragynine concentration in the attributed deaths as 2,486 µg/L with a 95% confidence interval of 1,607 to 2,863, and in the merely-associated deaths as 179 µg/L with a 95% confidence interval of 217 to 375. A confidence interval that does not contain its own mean is not possible. One of those three numbers is mistyped, and from the abstract alone there is no way to tell which.
The direction of the finding, that the attributed deaths had far more mitragynine in them, is not in doubt. The figure should not be quoted to the unit until the full text resolves it. We have read the abstract and not the full text, which we flag again in the method note below.
Why a postmortem concentration is weaker evidence than it looks
Four separate problems sit between a number on a toxicology report and a conclusion about cause of death. Each is documented in the forensic literature rather than inferred by us.
Testing is not routine. Mitragynine is not on standard panels. A laboratory finds it when it looks for it, so the figures measure testing practice as much as exposure. The CDC authors say their count may underestimate kratom-positive deaths by an amount that cannot be determined.
The analysis is hard to do correctly. A 2019 analysis of 1,001 mitragynine-positive blood specimens reported concentrations from 5.6 to 29,000 ng/mL, with a median of 130 ng/mL and a mean of 410 ng/mL. Its central methodological point is that mitragynine must be resolved from its diastereomers and related natural products, and that failing to do so can produce falsely elevated concentrations that are then wrongly compared against reference ranges.
7-hydroxymitragynine degrades in the sample. On the pharmacology, 7-OH would be the better marker of kratom toxicity because it is the more potent compound. A study of 51 postmortem cases found it stable for only about seven days refrigerated, with concentrations dropping more than half after 60 days. A low 7-OH result may describe the storage history of the specimen rather than the deceased person.
There is no established lethal concentration or dose. This is the load-bearing point and it comes from a peer-reviewed commentary by five kratom researchers. No causative lethal blood concentration, and no lethal dose, has been established for kratom or mitragynine in humans or in animals. That commentary was written to correct a published paper, the Kern County report above, which had described a blood concentration as a lethal dose. It notes the error was not unique to that paper.
What the two advocacy readings each get wrong
A 2025 review examined 95 patients from published case reports of acute adverse effects involving kratom. Mitragynine was toxicologically confirmed in 55, of whom 35 had died. In 32 of the deceased cases and 7 of the surviving ones, the toxicology showed other substances capable of causing or contributing to the effect, despite many of those effects having been attributed to kratom alone. The authors describe a pattern of weak or inadequate toxicology testing and medical history in the source reports, and conclude that the literature is insufficient to show that kratom consumption alone increases the risk of severe acute adverse effects.
That conclusion should be read alongside its declared interests, which the paper states and which we repeat here rather than summarise away. All five authors were employed by the same consulting firm, and one has served as a paid expert in kratom litigation and may do so again.
The disclosure cuts both ways, so here is the other side of it. The lead author of the commentary we relied on above, for the point that no lethal concentration has been established, discloses in separate work honoraria from a forensic research centre, payments from law firms as an expert witness in kratom cases, and a professional society presidency.
Neither disclosure makes either paper wrong. It does mean that much of the published work on kratom mortality is written by people who are paid to give evidence about it, in both directions, which is worth knowing before weighing any single claim on this subject.
The prohibitionist reading fails on the same ground from the other side. The 1,221 figure is a mention count, most mentions sit beside fentanyl, and no dose or concentration has been established at which kratom kills. Treating the count as a body count is the error the agency’s own technical notes are written to prevent.
What we do not know
Whether kratom alone, at quantities people actually consume, causes death at any measurable rate. The attributed single-substance cases are real and they are few, and “few” in a count of certificates is not a rate, because the denominator, the number of people using kratom and how much, is not known with any precision.
How much of the rise to 2022 and the fall since is consumption, and how much is testing and certification practice.
How many of the deaths counted under MITRAGYNINE involved concentrated 7-hydroxymitragynine products rather than leaf. The data as published cannot answer this, and it is the question most worth answering.
Whether the concentration-dependence the Florida study reports holds up in other jurisdictions. One case-control study in one state is a reason to look, not a threshold.
How we found these papers
Europe PMC was searched on 12 September 2026 with three queries: (kratom OR mitragynine) AND (death OR mortality OR fatal OR postmortem OR "medical examiner" OR coroner), which returned 747
records; kratom AND "cause of death", 116 records; and mitragynine AND (postmortem OR "post-mortem") AND concentration, 82 records. Included were agency mortality reports, medical
examiner and coroner case series, and analytical toxicology papers bearing on how the numbers are
produced. Excluded were single case reports without toxicology, animal work, and conference
abstracts with no full text.
Four papers were read in full from the open-access text: the CDC report, the Kern County study, the commentary on it, and the case-report evaluation. Four were available to us as abstracts only, and their figures are reported here as abstract-limited: the Oklahoma series, the Florida case-control study, the 51-case distribution study, and the Papsun analysis of mitragynine’s interpretative and analytical issues. We have not read their full texts and do not characterise anything in them beyond what their abstracts state.
The CDC material is not from the literature search. It was read from the agency’s technical notes, its published search-term table and its own data endpoint, rather than from any report about it.
If you can point us to a full text that settles the Florida confidence interval, or to jurisdiction-level data that separates leaf from concentrate, write to editorial@wikikratom.com.