Questioning the story

Caveat on this rating: Read from the abstract and plain language summary only. Most included trials excluded young people at risk of suicide; suicide-related estimates are low certainty with wide intervals. The 2026 re-appraisal authors report that the full dataset behind this review was not retained.

Caveat on this rating: An exploratory re-analysis with no preregistered protocol, by authors who have argued elsewhere that fluoxetine's benefit in young people has shrunk over time. Fluoxetine still beat placebo beyond chance in every analysis; what changed is the size. The 2006 trial had not been retracted or corrected when the paper was written.

Caveat on this rating: One trial in veterans (84% men) who had already attempted, with modest lithium levels and a broad composite outcome; deaths were too few to compare. It tests adding lithium to usual care, not lithium started as the main treatment. The older meta-analysis of 48 smaller trials points the other way.

Caveat on this rating: Ever-use only, no dose; cannabis variants below genome-wide significance; pleiotropy tests possibly underpowered. Gage 2017 found a smaller cannabis effect and a stronger reverse direction.

Caveat on this rating: A methods critique: it argues effects are overstated but does not measure by how much, and it does not show the drugs have no effect. Read from the abstract only.

Caveat on this rating: Observational, no placebo, self-selected through psychedelic-friendly venues, 180 comparators against 953 microdosers; the app maker's staff and a supplement seller with patents on the tested combination are co-authors. Randomised trials find no clear effect.

Caveat on this rating: Weak genetic instruments for cannabis and an ever-used measure that cannot see dose; the authors say their cannabis estimate is probably conservative. A 2018 MR study found a larger causal effect.

Caveat on this rating: Small correlations, considerable heterogeneity, and a fail-safe number below criterion in every analysis; no link with live birth. The authors call the influence 'somewhat limited'. Boivin 2011 found no link. Read from the abstract only.

Caveat on this rating: Cross-sectional, self-selected app users recruited through psychedelic-friendly venues, with small effects; the app maker's founders and a patent holder are co-authors. Randomised trials (Meshkat 2026) find no clear effect.

Caveat on this rating: An editorial by psychiatry researchers, several with declared pharmaceutical ties and one the journal's editor. It critiques survey methods but concedes long-term withdrawal exists and that trials are too short to measure it. It does not dispute that patients are often not told.

Caveat on this rating: The theory's author defending it; it argues the critique misrepresents the theory rather than presenting new experimental data on the disputed physiology. The 39-author evaluation it answers is in this library.

Caveat on this rating: Many co-authors are longstanding critics of the theory, so this states an established position forcefully rather than coming from a neutral panel. It evaluates the theory's physiology and evolution, not any therapy. Porges's reply (same issue) says it attacks a proxy of the theory.

Caveat on this rating: This is the regulator that approved the drug, describing its own doubts and why it approved anyway. It cannot say whether the placebo relapses were withdrawal, unblinding or true relapse; it says the reviewers could not tell.

Caveat on this rating: Its author wrote it, and it answers its critics in its own pages. A 39-author evaluation of this paper (Grossman et al. 2026, same journal) calls the theory untenable; Porges replied in the same issue. The clinical evidence it cites for the Safe and Sound Protocol is one single-arm study, one single-child report and two preliminary controlled trials.

Caveat on this rating: Sixty-six healthy people, one dose, three weeks, and a laboratory task rather than a reported feeling; the authors say the result 'may reflect' blunting. Semi-randomised allocation. No other cognitive measure changed. It says nothing about duration, reversibility or patients with depression.

Caveat on this rating: Post hoc, by the authors' own description. An odds ratio per prior course, with no absolute response rates in the abstract. The continuation phase showed no association. Confounding by harder-to-treat illness cannot be excluded from the abstract. It does not test restarting the same drug.

Caveat on this rating: A secondary analysis of a trial designed for a different comparison; one drug, 16 weeks. The authors themselves make the practice implication conditional on replication in rigorous studies, and the abstract does not report that replication. Confounding by illness severity cannot be fully excluded.

Caveat on this rating: A scoping review maps the literature without pooling or grading it, so it gives no rate and no effect size. The two management strategies are what clinicians do, not what has been tested; the abstract reports no comparative trial. Its own central finding is that symptom and side effect have not been separated.

Caveat on this rating: A self-selected internet survey with self-rated symptoms and no untreated depressed comparison group; the authors name self-evaluation and sample size as limitations. Blunting tracked depression severity, so the 46% mixes drug effect and residual illness by the authors' own account. Two of four authors are affiliated with a pharmaceutical research institute.

Caveat on this rating: Recruited from a forum for people who believe they have PSSD, retrospective and self-reported; the authors state the design does not allow estimation of incidence. Only 23 of 532 met the strict criteria. Dose did not predict severity and depression did, so the illness remains part of the picture.

Caveat on this rating: Qualitative: it describes and does not count. Part of the material came from patient websites, a self-selected source. Participants' confidence that they could separate side effect from symptom is a self-report, and the same group's later survey found blunting tracks depression severity. The six themes are not in the abstract.

Caveat on this rating: Self-submitted reports to a website built to collect them, assembled by the website's founder. A case series describes; it cannot count incidence or establish causation in any case. The cross-class overlap is unexplained. No rate, mechanism or treatment.

Caveat on this rating: A single-author literature review proposing a hypothesis its author says is untested and hard to test. The rechallenge and paradoxical effects are reported 'in a few patients' and 'in susceptible individuals', not measured as rates. The same abstract calls antidepressants crucial for major depressive episodes.

Caveat on this rating: A narrative review that does not grade or count its sources. Seven proposed mechanisms and none established. The phototherapy 'promising results' rest on evidence the abstract does not describe. The review's purpose is partly to raise recognition, which is stated.

Caveat on this rating: Eleven hospitals in one state in one year, funded by the malpractice insurer of the hospitals studied. Preventability and severity are reviewer judgements made from the record after the fact. The other direction: this is a count, not an extrapolation, and its death finding (one preventable death in 2,809 admissions) is far closer to the low end of the published range than to the 251,454 headline, while its harm finding (nearly one admission in four) is far higher than what hospital reporting systems capture.

Caveat on this rating: The minority stress model is a framework the author proposes, not a tested causal chain; the paper itself lists future research as an aim. Prevalence estimates depend on how orientation was measured in the underlying studies.

Caveat on this rating: Both figures are extrapolations from chart reviews of adverse events in two states and one year each, scaled to the whole country. The reviews were designed to count injuries from care and negligence, not to judge whether a given death would have happened without the error, so the death counts inherit that gap. The other direction: the report is the reason systematic counting of harm began, and the later studies that produced smaller, more careful numbers exist because of it.

Caveat on this rating: This row is the contest. The 251,454 is an arithmetic average of four earlier estimates, one from a for-profit ratings company, resting on studies with 14 or fewer deaths each; the 25,000 is a different extrapolation from a 3.6 percent preventability rate. Neither is a count. What is not contested: the death certificate does miss human and system factors, care does harm people far more often than reporting systems admit, and the two sides disagree about the size of the death toll, not about whether error kills.

Caveat on this rating: The 1.5 million is a sum of projections, each from a single site or a single practice scaled to the country, and the committee itself called them likely underestimates, which cuts both ways: the true figure may be higher, and no part of it is a direct national count. These are events, not deaths. The other direction: medication harm is the best-documented kind of harm from care, it was the most common category in the 2023 count, and it is the kind a patient can do something about by keeping one list and asking about interactions.

Caveat on this rating: Observational: nobody controlled the treatment, and the drugs in use were those of 1978 to the 1990s. The unit is the maintenance interval, not the person, and the 103 people were selected as medication responders. The word tachyphylaxis names the event, not a proven mechanism.

Caveat on this rating: Retrospective interviews in a purposive sample from one specialist clinic; adults recalling adolescence. No comparison group. Designed to inform the ICD-11 decision and co-authored by a working group member, so it is evidence gathered by the people proposing the change.

Caveat on this rating: A regulator page and a patient fact sheet, not evidence that asking these questions lowers harm; no trial on this row tests the list. The dose-for-age-and-kidneys question is not AHRQ wording and the row says so. The other direction: the questions cost nothing, the harms they target (interactions, unintentional overdose in older adults, name confusion, medicines dropped or duplicated at discharge) are the best-documented harms from care, and none of the questions asks anyone to change a medicine.

Caveat on this rating: Controls are the general population, not transgender people without surgery, so the study cannot attribute the excess to surgery and the authors do not. Small counts give wide intervals (suicide aHR 5.8 to 62.9). The abstract itself credits surgery with "alleviating gender dysphoria".

Caveat on this rating: The surgery finding was formally corrected by the journal in the same issue after published letters; the correction text was not read for this row. The design compares different people at different times since treatment in a single year, not the same people before and after.

Caveat on this rating: A participant's account of a process he chaired. The paper reports no data and takes a position; the opposing position was argued in the same journal in the May 1973 symposium (PMID 4784866, record only, not read). The DSM-III category the paper discusses stayed in the manual until 1987.

Caveat on this rating: The paper states that its recommendations rest on social attitudes, policy and human rights as well as research. It reports no outcome data of its own. An erratum was published in 2017 and was not read for this row.

Caveat on this rating: A narrative review of heterogeneous studies with strict inclusion criteria; the authors say the findings may not generalise. The 9%-to-57% range is a definition problem as much as a measurement. The abstract does not separate drug tolerance from natural recurrence, and none of the four strategies is established.

Caveat on this rating: A narrative review from 1998 pooling trials with different relapse definitions; the range is as wide as the definitions. It lists seven explanations and rules none in or out, and it tests no strategy. Its call for double-blind studies has largely gone unanswered.

Caveat on this rating: A narrative history by one author with a stated position. The 58% referendum and the disrupted meetings are the case that the decision was political as well as scientific; the paper reports both. Kinsey's and Hooker's figures appear here as Drescher reports them, not from the originals.

Caveat on this rating: Written by participants in the classification work from a single specialist clinic. The paper's own point, that social and political factors shaped the diagnosis, applies to every edition including the current one.

Caveat on this rating: A position paper by members of the working group whose recommendation it argues for. It names human rights as a criterion alongside evidence, so the reasoning is explicitly values-led. It reports no outcome data.

Caveat on this rating: An observational comparison of meeting and nonmeeting dates in Medicare data, not a trial, and the authors say plainly that they cannot establish the mechanism. It cannot say who was actually in the building. The result is confined to high-risk heart failure and cardiac arrest patients at major teaching hospitals; there was no difference for low-risk patients anywhere or for any patient at nonteaching hospitals, and the high-risk heart attack group had the same mortality with fewer stents. It is one study. It is evidence that a particular intensity of care may not help the sickest, not evidence that hospitals or cardiologists are dangerous.

Caveat on this rating: Short trials that were not designed to measure suicidality; events identified retrospectively from adverse event reports. Trials exclude the most acutely suicidal patients. Absolute event rates were low in every age group.

Caveat on this rating: One insured population in one health plan, 1992 to 2003. No untreated comparison group, so it describes timing, not drug effect. Attempts that did not lead to admission are not counted.

Caveat on this rating: Observational cohort: confounding by indication and by who stays in care. Pre-treatment attempts partly cause the diagnosis and prescription, which inflates the before-treatment rate. Attempts, not deaths. The 18 to 25 group, where the effect was not significant, is the group covered by the FDA boxed warning.

Caveat on this rating: Pools randomized and open trials of very different designs and durations; the 12.5 and 7.5 percent are averages, not a controlled comparison. Hypomania definitions vary. No hospitalization outcome.

Caveat on this rating: Counts published studies only; unpublished or ongoing postapproval work is invisible to it. All therapeutic areas; the abstract gives no psychiatry breakdown.

Caveat on this rating: Small (103 followed up), open label, and after the 18-month trial treatment was at clinicians' discretion, so the two arms converged in practice. Applies only to remitted first-episode patients. Not yet replicated at this length.

Caveat on this rating: One city and a cohort that aged over the study. Structured lay interviews can miss disorders a clinician would diagnose, and disorders outside the five assessed (for example PTSD, dysthymia, chronic pain indications) are not counted.

Caveat on this rating: A composite outcome that treats a safety communication and a withdrawal alike. A higher event rate can reflect closer surveillance and wider, longer use as much as greater hazard. Covers approvals through 2010; the newest psychiatric brands are not in it.

Caveat on this rating: Observational. Within-individual design removes stable differences between people but not the fact that stopping a drug and relapsing can share a cause. Industry-employed co-authors and the ranking of the sponsor's product first warrant the usual caution.

Caveat on this rating: Observational; the cohort is people already hospitalized once, so it says little about milder bipolar illness. The gabapentin result is unexpected and did not survive the sensitivity analyses.

Caveat on this rating: Cross-sectional prevalence, not incidence, and heavily confounded by prior exposure to older drugs, age and illness duration, which the authors model. Severity was not analysable. Prevalence on a rating scale overstates clinically disabling TD.

Caveat on this rating: Repeated cross-sections, not the same people followed over time, so duration is self-reported at one interview. Longer use may partly reflect guideline advice to continue treatment after recurrent episodes.

Caveat on this rating: Naturalistic cohort of 139; treatment was not assigned, so confounding by illness severity is the central objection and the authors acknowledge other factors. Outcome is work, not relapse or admission.

Caveat on this rating: Observational between-person comparison with adjustment; confounding by reason for stopping cannot be removed. The five-year-plus group is small (wide interval, 2.78 to 19.05).

Caveat on this rating: Ecological, short (US suicide data only through 2004), and widely debated: later analyses disagree about whether the warnings changed suicide deaths. The authors infer untreated depression as the mechanism; the data cannot show it.

Caveat on this rating: Mania identified from register diagnoses, so milder hypomania is missed. The monotherapy estimate has a wide interval (1.12 to 7.19). People given monotherapy may differ in ways a within-person design does not capture, such as a recent change in illness course.

Caveat on this rating: Included studies span decades, 12 countries and very different methods, which is why the authors report medians and not a pooled rate. Not specific to psychiatric drugs or to the US system.

Caveat on this rating: Visit-level survey data: a diagnosis may exist in the chart from an earlier visit. Off-label but evidence-supported uses (pain, insomnia) are counted as "no psychiatric diagnosis". Data end in 2007.

Caveat on this rating: A descriptive brief of death certificate data; it makes no claim about causes or policy. Method of death is reliably recorded; intent (suicide versus undetermined or accidental, especially for overdoses) is less so.

Caveat on this rating: Counts stays, not people, and only readmissions within the same state data. 2012 is the most recent brief of this design. Readmission is a measure of system performance as much as of illness.

Caveat on this rating: Before 1933 the figures cover only the death registration states, not the whole country, and the table warns that revisions of the cause-of-death classification "do create breaks in the comparability". Suicide has been undercounted to differing degrees in different eras. That more people were in treatment while the rate rose is a juxtaposition, not evidence of cause in either direction (treatment use: the adult medication row in this library). A flat rate since 2018 hides opposite movements by age and sex.

Caveat on this rating: Very small matched sample from two hospitals; confounding by indication is the central objection. Mixed diagnoses. Not replicated in a large cohort.

Caveat on this rating: Ecological before-and-after design with no comparison country. Supporters of headspace point to evaluations of people who used the service; this paper measures the whole youth population, most of whom did not.

Caveat on this rating: An association across national time series; it cannot show that any individual death followed a job loss. The rise from 2000 to 2018 ran through the longest expansion on record, so the business cycle is one driver among several.

Caveat on this rating: A review of repeated national surveys, which differ in instruments and response rates across waves and countries. The authors could not test whether greater willingness to disclose symptoms hides a real improvement.

Caveat on this rating: A narrative evaluation of explanations, not a systematic review with a pooled estimate. Prevalence surveys have their own measurement problems, and the authors weigh rather than rule out the possibility that rising willingness to report symptoms masks an improvement.

Caveat on this rating: One health system, retrospective, and dependent on a serotonin syndrome diagnosis having been recorded; mild cases managed without a diagnosis code would be missed, which would push the true rate up. The study argues against the advisory; the FDA has not, to our knowledge, withdrawn it, and the labels still list triptans.

Caveat on this rating: Retrospective claims analysis: association, not causation, and the authors say so ("If this association is causal"). Privately insured adults under 65 only; the elderly and Medicaid populations, where both drugs are common, are not represented.

Caveat on this rating: Healthy volunteers, 14 days of pretreatment, one psilocybin dose in a laboratory; partly funded by a psychedelic drug developer. The Gukasyan 2023 survey in this wave reports blunting in ordinary use that this design could not detect.

Caveat on this rating: Observational studies pooled; confounding by indication (people with depression may use more NSAIDs or alcohol) cannot be excluded. The authors note earlier reviews had overestimated the SSRI-alone risk.

Caveat on this rating: The 2019 FDA notice, as reported at the time, said the evidence for breathing problems in otherwise healthy people taking a gabapentinoid alone was limited; the label warning is specific to co-prescription with CNS depressants and to existing lung disease. That notice was not read for this row because it is no longer hosted.

Caveat on this rating: Sixteen healthy volunteers, single doses, intravenous citalopram rather than chronic oral SSRI treatment; a person on months of an oral SSRI may respond differently. Whether the cardiovascular effects of MDMA were also reduced is not stated in the abstract and is not claimed here.

Caveat on this rating: The label lists examples, not a closed list; the size of the concentration increase varies by drug and is not given in the summary section read.

Caveat on this rating: Twenty five states, six months, and "presence" means detected at death, not cause. The decrease reported was small and the authors do not claim a trend.

Caveat on this rating: Nested case-control in administrative data: association, not causation; residual confounding by pain severity or illness is possible. Ontario only, 1997 to 2013. The NSAID negative control strengthens but does not settle the causal reading.

Caveat on this rating: A contraindication states that the regulator considers the combination unsafe; no incidence figure is given and none was read.

Caveat on this rating: Observational; the paroxetine label (this wave) states that other studies have failed to find the risk. Women 66 and older only. The other-SSRI null result is reassuring but the study was not powered to rank them.

Caveat on this rating: The 2011 notice was criticised in the literature for resting on QT interval data rather than on measured arrhythmias; the label itself reports postmarketing cases, which are unquantified. The Beach 2013 review in this wave notes there are no well-controlled studies of torsades risk with psychiatric drugs.

Caveat on this rating: A label warning states a class risk; it gives no rate. The Tori 2020 and Gladden 2019 rows in this wave give the surveillance figures.

Caveat on this rating: A label warning states a risk the regulator considers established; it does not give the size of the risk. The size is in the Sun 2017 BMJ analysis in this wave (about twice the odds of an overdose visit with concurrent use), which is observational and cannot prove the combination caused the overdoses.

Caveat on this rating: Death-certificate data: co-involvement depends on what a medical examiner tested for and recorded, which varies by state and year, so both the levels and the trend are partly measurement.

Caveat on this rating: The label lists classes, not doses; it does not say how much alcohol changes the effect, and no figure for that was read.

Caveat on this rating: Observational; the confidence interval on the main odds ratio runs from 1.22 to 9.18. The companion fixed-effects analysis addresses self-medication and confounding but cannot rule them out entirely. UK cannabis market, 2002 to 2015.

Caveat on this rating: Observational; reverse causation (prodromal dementia leading to prescriptions) is the standard objection, partly addressed by the longer exposure windows. The 10.3% attributable fraction assumes causation. Not all antidepressants or antipsychotics are anticholinergic; SSRIs other than paroxetine mostly are not.

Caveat on this rating: The label balances two risks it does not quantify against each other. No figure for overdose deaths with buprenorphine plus benzodiazepines versus untreated opioid use disorder was read for this row.

Caveat on this rating: Online retrospective self-report; selection and recall bias both plausible, and dose was self-reported. The Becker 2022 controlled trial found positive effects preserved after 14 days of escitalopram, so the size of the blunting in ordinary use is uncertain.

Caveat on this rating: Derived from overdose admissions at one Australian toxicology service; performance in therapeutic-dose interactions, which are usually milder, was not the setting studied.

Caveat on this rating: The label says other estrogen products such as HRT "have not been studied"; the contraceptive effect is established for combined oral contraceptives and inferred for the rest.

Caveat on this rating: Observational; the "previous user" group is small (4.6 percent) and self-selected. The underlying EMBLEM cohort was a pharmaceutical-sponsored study, which does not bear on the cannabis comparison but is worth knowing.

Caveat on this rating: The imipramine and chlorpromazine findings are single-dose volunteer studies, not outcomes in patients. The label gives no figure for how much alcohol changes next-day impairment.

Caveat on this rating: Twelve healthy young volunteers, fixed treatment order, one St John's wort product; extracts vary in the hyperforin content thought to drive induction, so other products may induce less or more.

Caveat on this rating: The label itself records both directions on tamoxifen. The Kelly 2010 row in this wave is the study that found the risk.

Caveat on this rating: Small pooled sample from two earlier trials, retrospective analysis, cut-off derived from the same data it is tested on. Association, not proof that the benzodiazepine caused the non-response.

Caveat on this rating: The cannabidiol finding comes from trials of prescription-strength cannabidiol; whether typical over-the-counter CBD doses reach the same effect is not stated in the label and was not read.

Caveat on this rating: The label lists NSAIDs and thiazides among the level-raising drugs; the Juurlink 2004 population study in this wave found neither independently associated with hospital admission for toxicity in older adults, while loop diuretics and ACE inhibitors were. Both are on this page.

Caveat on this rating: Fifty two women, cross-sectional, levels not outcomes. The size of the effect is consistent with the label, which was written after studies of this kind.

Caveat on this rating: A narrative review from 2013; the specific drug rankings could have shifted with later data, and the review itself says the comparative evidence is limited.

Caveat on this rating: Self-selected online reports with no denominator; the percentages are of reports, not of users, and severe events are over-represented. The lithium versus lamotrigine contrast within the same sources is the finding, not the 47 percent itself.

Caveat on this rating: The label's list mixes drugs "shown" and drugs "expected" to raise the level, without distinguishing which is which for each entry.

Caveat on this rating: A narrative review; "known or predicted to interact" includes predictions from pharmacology, not only documented cases. Most of the serious harms tabulated are for cardiovascular and transplant drugs, not psychiatric ones.

Caveat on this rating: A label lists warnings the regulator requires; it does not give incidence. No figure for how often tramadol plus an SSRI produces serotonin syndrome was read for this row, and the Orlova 2018 row in this wave is a reminder that a labelled serotonergic interaction can turn out to be rare when measured.

Caveat on this rating: The 10 msec figure is from a single-dose volunteer study cited in the label, not from patients on chronic pimozide.

Caveat on this rating: Spontaneous reports cannot give a rate. The label lists triptans among serotonergic agents; the Orlova 2018 row in this wave measured the triptan plus SSRI combination and found it rare.

Caveat on this rating: Wide confidence intervals (the ACE inhibitor interval runs from 2.6 to 22). Adults 66 and older only. Case-control design in administrative data: association, not proof. The NSAID null result conflicts with the label and with pharmacology; the authors report it, they do not explain it.

Caveat on this rating: The label describes prescription cannabidiol at epilepsy doses (commonly 10 to 20 mg/kg/day); over-the-counter products deliver far less and often not what the packet says, so extrapolation to them is uncertain in both directions.

Caveat on this rating: The one-third rule is a label instruction, not a measured average; individual changes vary widely, which is why the label pairs it with monitoring.

Caveat on this rating: The alprazolam and propranolol figures are from small volunteer studies cited in the label at fluvoxamine 100 mg; the label notes the interaction "may be more pronounced" at 300 mg and was not studied there.

Caveat on this rating: Small naturalistic sample from nursing homes; the clozapine concentration difference (50%) did not reach conventional significance (p=0.058) although the ratio difference did.

Caveat on this rating: Two case reports; the abstract attributes the effect to smoke constituents without separating cannabis from tobacco, so the cannabis contribution on its own is inferred, not measured.

Caveat on this rating: A case series of six intensive-care patients; seizures in critically ill people have many causes and the series cannot apportion them. The direction and size of the level drop are consistent across the literature the authors reviewed.

Caveat on this rating: The food list is old and broad; Gillman 2011 (this wave) argues tyramine content in modern foods is lower and the hypertension risk lower than the label era assumed. That is a dispute about the food rules, not about the drug contraindications.

Caveat on this rating: A contraindication list is the regulator's judgement, not a set of measured rates; the Gillman 2011 row in the interactions wave argues the food and some drug risks of MAOIs are smaller than the label era assumed, while not disputing the stimulant and serotonergic contraindications.

Caveat on this rating: The trial percentages are for esketamine nasal spray at approved doses under observation; they do not transfer directly to street ketamine of unknown dose and purity, in either direction.

Caveat on this rating: Four forensic cases; "probable cause" is the authors' judgement from toxicology and circumstances. A case series cannot give a rate, and moclobemide is not the MAOI US patients take.

Caveat on this rating: A single-author narrative review from an author who has long argued MAOIs are underused. The tyramine claim is plausible and widely cited but rests on food-content surveys not read for this row. The label contraindications stand regardless.

Caveat on this rating: NSDUH surveys the civilian, noninstitutionalized population aged 12 or older, so people in hospitals, prisons, nursing homes, on active duty or unsheltered are not in it; SAMHSA states that estimates from 2021 onward are not comparable with 2020 or earlier, so no line here runs back further than 2021. A survey answer is not a diagnosis. 'Any mental illness' and 'serious mental illness' are model estimates calibrated on clinical interviews, not counts of diagnoses.

Caveat on this rating: NSDUH surveys the civilian, noninstitutionalized population aged 12 or older, so people in hospitals, prisons, nursing homes, on active duty or unsheltered are not in it; SAMHSA states that estimates from 2021 onward are not comparable with 2020 or earlier, so no line here runs back further than 2021. A survey answer is not a diagnosis. The death count is from the National Vital Statistics System, not from the survey, and covers all ages.

Caveat on this rating: NSDUH surveys the civilian, noninstitutionalized population aged 12 or older, so people in hospitals, prisons, nursing homes, on active duty or unsheltered are not in it; SAMHSA states that estimates from 2021 onward are not comparable with 2020 or earlier, so no line here runs back further than 2021. A survey answer is not a diagnosis. Marijuana is past-month use, not a disorder; the two series are different questions and no causal link is stated by the report or claimed here.

Caveat on this rating: A principal-diagnosis count depends on coding practice and on the definition AHRQ uses, which changed in 2016; 'first' means first among non-birth diagnoses in community hospitals, and the count says nothing about outcomes.

Caveat on this rating: NSDUH surveys the civilian, noninstitutionalized population aged 12 or older, so people in hospitals, prisons, nursing homes, on active duty or unsheltered are not in it; SAMHSA states that estimates from 2021 onward are not comparable with 2020 or earlier, so no line here runs back further than 2021. A survey answer is not a diagnosis. The report gives no cause for the decline and none is claimed here.

Caveat on this rating: This row asserts an absence, which is the claim most easily overstated. What is established is that the delayed onset is not explained by the immediate neurochemical change, and that the field's own reviews present competing accounts rather than a settled one. What is NOT established is that either account here is correct: the emotional-processing model and the neuroplasticity model are both active research programmes with supporting and non-supporting findings, and this row deliberately reports that they are unresolved instead of ranking them. The claim that no clinical test measures a chemical imbalance is a statement about available assays, not a claim that no biological contribution exists. A narrative review is also a lower grade of evidence than a systematic one, and it is cited here for what the field says about its own uncertainty rather than as a pooled result.

Caveat on this rating: This literature does not point one way, and the row would be dishonest if it implied otherwise. A randomised study from the same research group as the bogus-test experiment found self-blame and perceived helplessness GREATER after a cognitive-behavioural explanation than a biological one (Lee et al. 2016, J Soc Clin Psychol 35(7):571-588), and a pre-registered sham-genetic-test experiment found no effect on beliefs about depression at all (Lamontagne et al. 2023, PMID 36869258). Both are named here and neither is quoted, because their full texts were not obtained. The Kvaale effect sizes are small to moderate, the dangerousness result may reflect publication bias by the authors' own account, and the samples across this field are dominated by undergraduates and online panels rather than people in treatment. The absence of any study of debunking, or of any behavioural outcome, is stated as an absence found by search - it is not proof that no such work exists anywhere.

Caveat on this rating: Every efficacy number here is contested at the edges, and none of the disputes touches the conclusion this stop rests on. Cipriani 2018's own authors graded certainty moderate to very low and 9% of included trials at high risk of bias; its between-drug ranking is not established and is not used here. Stone 2022's three-class result comes from a statistical model, not an observed grouping, and its reading has been argued over since publication - though its co-author list includes the field's most prominent placebo-effect sceptic, which cuts against overstating benefit rather than for it. Both meta-analyses are dominated by short, largely industry-sponsored acute-phase trials in adults meeting formal criteria for major depression, and neither measures long-term outcomes. Cowen and Browning's emotional-processing account is a hypothesis with supporting experiments, not a settled mechanism. The ANTLER figures describe trial conditions, not any individual reader's risk.

Caveat on this rating: This stop is the dispute, so the caveat is about what it does NOT establish. It does not establish that the serotonin hypothesis is disproved: the review's own framing is absence of support, and Jauhar and colleagues argue that its method manufactures that absence. It does not establish that the rebuttal is correct either. The two sides also disagree about what tryptophan depletion is for - whether the test is inducing depression in healthy people, or worsening it in people with a history - and that disagreement is unresolved rather than adjudicated here. Four further published critiques and a second author reply appeared in 2024 and are named in the body only as "a second exchange", because their full texts are paywalled with no PMC deposit and this library does not quote a paper it has only read about. Declared interests are printed for both sides precisely because neither set of interests decides the science.

Caveat on this rating: The central dispute in this row is unresolved and is left that way on purpose. Pies argues that no professional body, textbook or peer-reviewed consensus ever advanced a chemical-imbalance theory and that critics cherry-pick quotations; Ang and colleagues report that highly cited reviews and every textbook they sampled did support the serotonin theory between 1990 and 2010. Two of the three authors on that side are also authors of the 2022 umbrella review, which is a stake in the outcome and is stated rather than hidden. Their paper was a sample of highly cited work, not a census, and its full text could not be obtained, so nothing beyond its published abstract is quoted here. The France 2007 survey is 251 US undergraduates in one place at one time and does not describe the general public. Nothing in this row establishes that anyone intended to mislead; intent cannot be documented from outside, and the essay itself does not claim it.

Caveat on this rating: Definitions do the heavy lifting here and are stated rather than assumed. 'Pregnancy-related' means causally related to the pregnancy and within one year of its end - a wider window than the 42-day definition used in international comparisons, which is why UK figures for suicide as a direct cause within 42 days (0.80 per 100,000) look nothing like the 6-week-to-1-year figures quoted. The MMRC category bundles suicide with overdose and poisoning related to substance use disorder, so it is not a suicide count. The 2020 US figure covers 38 states, not the nation, and a 2020 coding change moved substance-use-related overdose deaths into this category, which is part of why 22.5% is not comparable with the 11% from 2008-17. A separate CDC report gives 26% for mental health conditions as a CONTRIBUTING circumstance rather than an underlying cause; that figure answers a different question and is not used. None of these numbers say anything about the risk facing any individual reader.

Caveat on this rating: Every study in this row is observational and self-selected, and that cuts against the row's own direction as much as for it. Cohen 2006 recruited at three specialist perinatal psychiatry centres, so its population is women with severe recurrent illness rather than a general one, and its hazard ratio of 5.0 cannot be read as the causal effect of stopping. Grote 2010 measured antenatal depression rather than treatment status, and its effect sizes swing with the measurement instrument and the country. Jarde 2016's own conflict-of-interest subgroup nearly doubles the preterm-birth odds ratio, which is a warning about the literature and is printed here for that reason. No randomised trial of continuation versus discontinuation exists, so 'stopping is riskier' and 'continuing is riskier' rest on the same grade of evidence - which is the point of the stop, and why it closes on a conversation rather than a conclusion.

Caveat on this rating: What this row does not claim: that SSRIs in pregnancy are risk-free. Adjustment is not randomisation, and the propensity-score models can only correct for what was measured in claims data - smoking, alcohol, illness severity and body-mass index are poorly captured in Medicaid records, and residual confounding runs in both directions. The primary-PPHN result still clears one after adjustment. Brown 2017's sibling analysis has wide intervals rather than a null, and its authors wrote that a causal relationship cannot be ruled out. Sujan 2017's surviving preterm-birth association may itself reflect illness severity within a family rather than the drug. Levinson-Castiel is a single centre with 60 exposed infants. The defensible reading is that the large early relative risks shrank substantially under better designs, that the absolute excess risks are small where they persist, and that neither 'proven harmful' nor 'proven safe' is what the evidence supports.

Caveat on this rating: The uncertainty here is real and is printed rather than resolved. Patorno's adjusted risk ratio for cardiac malformations is 1.65 with a lower bound of 1.02, and the right ventricular outflow tract estimate is 2.66 with a lower bound of exactly 1.00 - both only just clear chance, and the lithium numerator for that outcome is suppressed under the Medicaid data-use agreement, so this row makes no claim about how many Ebstein cases the study did or did not observe. Patorno excluded Ebstein anomaly as a named outcome on purpose, because clinicians may be likelier to code a defect as Ebstein in an infant known to have been lithium-exposed. Munk-Olsen 2018's first-trimester malformation result is significant while its any-time and cardiac-specific results are not. Fornaro 2020's meta-analysis and Diav-Citrin 2014 point the same direction with wide intervals. The honest summary is a small absolute increase, dose-related, on a very small base - not the register's 400-fold, and not nothing.

Caveat on this rating: Named limits. EURAP is a prospective registry of women in epilepsy care, not a randomised comparison, and women prescribed valproate differ from women prescribed lamotrigine in ways registries cannot fully adjust away - the strongest single caution on every figure here. NEAD is observational and its age-6 arm retained 224 of 311 children. Christensen's whole-cohort autism estimate (4.42% vs 1.53%) is confounded by indication and the epilepsy-restricted hazard ratio of 1.7 (0.9-3.2) is NOT statistically significant; both are printed so the reader can see the difference the comparison group makes. The North American registry's 9.3% valproate figure is consistent with EURAP but its published abstract carries no confidence interval, so it is not quoted here. What is not in dispute across registries, cohorts and three regulators is the direction, the dose-dependence and the rough size.

Caveat on this rating: What is NOT established: that the narrative PLLR labels are better understood by clinicians in practice. No study shows that, and the Namazy survey points the other way - 68% of respondents did not find the narrative concise, and 95% still used the letters. This row asserts the documented change and the documented reasons the agency gave for it, not that the replacement works. The Namazy survey is also a 12% response rate (184/1500) among allergy/immunology clinicians, not a general prescriber sample. The Addis comparison is from 2000 and concerns three systems, two of which still exist.

Caveat on this rating: Contested, and symmetrically - both reviews have published critiques. Jauhar and Hayes argue Davies and Read inverted the evidence hierarchy, giving most weight to self-selected online surveys of people already experiencing withdrawal, without registering inclusion criteria in advance and excluding low-incidence studies after the fact; on that reading 56% is not a population rate and cannot be attributed to the drug's pharmacology. Henssler and colleagues' placebo-controlled estimate drew five published comment letters in Lancet Psychiatry and carries a published erratum; the authors themselves report substantial heterogeneity and note that residual or re-emerging illness has to be considered when reading their figure. What neither paper disputes is that withdrawal is real, that it was underestimated for years, and that abrupt discontinuation is the avoidable part. The headline percentage is unsettled in both directions.

Caveat on this rating: Population and construct, both load-bearing. These are 476 consecutive adults already referred to a national specialist autism service - people who had passed a referring clinician's judgement - so the predictive values are conditioned on that population and do not describe someone completing the questionnaire at home, where the proportion who are autistic is different and unknown. Sensitivity and specificity are properties of the instrument against a clinical reference standard, not verdicts on any individual score. The finding replicates in the wider literature rather than standing alone: Sizoo et al. 2015 concluded that none of the common self-report instruments has sufficient validity to predict an autism diagnosis in outpatient settings, and Wigham et al.'s 2019 systematic review found questionnaires perform markedly worse in referral streams than in samples of already-diagnosed people.

Caveat on this rating: Genre limits, from the studies' own limitation sections, and they bind the numbers next to them. This is top-video sampling of one hashtag at one time point: there is no denominator, and it does not reproduce what any individual's feed shows. "Accurate" is rater-defined and the definitions differ between studies, so these percentages are not comparable with accuracy figures published for other conditions or platforms. Cross-sectional content ratings measure content, not harm - no study here follows viewers to any outcome. Rater agreement is itself imperfect: in a study of the same genre, expert raters reached only moderate agreement. Nothing in this literature measures how many people self-identify because of social media, or whether any of them are right; both are separately verified absences. The later adult-diagnosis study is Brennan et al. 2025 (doi:10.1007/s10803-025-07123-0).

Caveat on this rating: Citation provenance, stated plainly. The anchor is Volkmar and McPartland's peer-reviewed history of the autism diagnostic concept, which is the house-citable route for the DSM-III-to-DSM-5 timeline because DSM edition text is paywalled and the publisher's history page does not resolve. The quoted sentence is not from that review: it is verbatim from the American Psychiatric Association's own "Highlights of Changes from DSM-IV-TR to DSM-5" document, held on file and verified 2026-08-26, and it is attributed to the manual in the body rather than to the review. Diagnostic criteria are a definition, not a finding, so nothing here is evidence about causes, prevalence or what any individual has. The manual also states that people with a well-established earlier diagnosis keep it under the new category, which is why the 2013 change did not remove existing diagnoses.

Caveat on this rating: Design limits, stated because they are the whole caveat. This is a cross-sectional online survey of a self-selected convenience sample recruited through a national campaign, not a random population sample - the same methodological objection Jauhar and Hayes raise against this research group's withdrawal-incidence work applies here, and it is why no national disclosure rate can be read off this row. Recall of a past consultation is also self-reported and unverified against the record: some prescribers may have mentioned withdrawal without the patient retaining it, which the study cannot separate from non-disclosure. What the figure supports is that the conversation is frequently absent or not memorable, not that a precise percentage of prescribers omit it. The 752-participant UK figure is Read, Gee, Diggle and Butler 2019 (doi:10.1016/j.addbeh.2018.08.021), carried in prose and subject to the same sampling limit.

Caveat on this rating: The denominator is the whole caveat and it is stated in the body rather than hidden here. The paper's own words: "We find one out of five STM articles suffering from reference rot" and "When only considering STM articles that contain references to web resources, this fraction increases to seven out of ten." One in five is of ALL science, technology and medicine articles in their corpus - not of all scholarship, and not of web references. Dropping the denominator is the standard misquotation of this paper and would make our own figure the kind of claim this trail exists to warn about. The figure describes scholarly citation, not government websites; it is offered as context for why link rot is normal, never as a measurement of federal pages.

Caveat on this rating: The 2003-2007 chronology behind this warning is DELIBERATELY ABSENT from this row. The October 2003 advisory, the October 2004 announcement, the 2005 label change and the 2007 extension to adults under 25 could not be confirmed against a primary source during verification, so not one of those dates is stated. What is stated is what was verified: the boxed warning's current wording, read off the live label, and the fact that the FDA pages that explained it now 404. An unverified date on a row about a documentary record would be the exact failure this trail is about.

Caveat on this rating: The quoted sentence is NOT from the cited paper. It is from Thomas Insel's NIMH Director's Blog post "Transforming Diagnosis" of 29 April 2013, whose nimh.nih.gov URL returns 404, and it is transcribed from the Internet Archive capture at web.archive.org/web/20210709195956/https://www.nimh.nih.gov/about/directors/thomas-insel/blog/2013/transforming-diagnosis, retrieved 2026-08-27. web.archive.org is not an allowed citation host here, so the row's citation is Insel's own peer-reviewed statement of the same position, co-authored with Cuthbert in the same year - a source that is still hosted and still resolves. The blog's full sentence opens "Unlike our definitions of ischemic heart disease, lymphoma, or AIDS,"; the row quotes from "the DSM diagnoses" onward, verbatim.

Caveat on this rating: The register here is disclosure, not accusation, and the row is written to keep it that way. Retiring old pages is ordinary site maintenance, the FDA states the pre-2020 position itself rather than concealing it, and no claim of deletion or suppression is made anywhere on this row. What is stated is the effect: part of the public record of American drug regulation is reachable only through a nonprofit archive. The reader draws the inference. Note also that the older fda.gov path for this index (.../drug-safety-and-availability-drug-safety-communications) now 404s in its own right; the live address is the shorter one this row cites.

Caveat on this rating: The serotonin pairing is mandatory wherever this claim appears and is carried in the body as well as here. Moncrieff et al. 2022 (Mol Psychiatry, doi 10.1038/s41380-022-01661-0, PMID 35854107) found no consistent evidence of an association between serotonin and depression; Jauhar and colleagues replied in the same journal (doi 10.1038/s41380-023-02095-y) that the conclusion is overstated on grounds of method and interpretation. This is not settled in either direction. Schildkraut 1965 (Am J Psychiatry 122:509-522, doi 10.1176/ajp.122.5.509, PMID 5319766) is named but NOT quoted: its full text could not be read during verification, so the two-lines-of-evidence characterisation is taken from Hillhouse and Porter 2015, which is the row's citation.

Caveat on this rating: This stop deliberately breaks its own trail's pattern, because the record does. Zeller and colleagues published monoamine-oxidase inhibition by these compounds in 1952 (Experientia 8:349-350, doi 10.1007/bf02174413); Loomer, Saunders and Kline published the antidepressant evaluation in 1957 (Psychiatric Research Reports 8:129-141, PMID 13542681). So the mechanism preceded the psychiatric indication by five years, and iproniazid is NOT an example of a mechanism found after the fact. What was built backward is the inference from the drug's action to the cause of the illness, which is what the body says. No synthesis year is stated: two peer-reviewed sources disagree (1953 versus clinical use in 1952) and the conflict is unresolved.

Caveat on this rating: The serendipity is real and it is commonly overstated, so this row states both. Cade had a hypothesis - a toxin in manic urine - and used lithium urate because uric acid dissolves poorly; he then tested lithium carbonate alone, which is how the lithium rather than the urate was identified as active. Lopez-Munoz et al. 2018 (Int J Mol Sci 19:2143, doi 10.3390/ijms19072143) gives the balanced reading: chance played an important role, within a deliberately designed series. The modern claim that Cade's guinea pigs were simply lithium-toxic is NOT asserted here: no source for it was located, and an unsourced debunk is the same error in the other direction. A free reprint of the 1949 paper is at PMC2560740 (Bull World Health Organ 2000;78:518-520).

Caveat on this rating: The eleven-year figure is the gap to the first published MECHANISM account, not to the disease theory, and the row says so. Carlsson and Lindqvist 1963 (Acta Pharmacologica et Toxicologica 20:140-144, doi 10.1111/j.1600-0773.1963.tb01730.x) showed accelerated catecholamine metabolite turnover and INFERRED receptor blockade; direct receptor-binding demonstration came in 1975, twenty-three years after the first dose. A common miscitation is corrected here: the dopamine history is Baumeister and Francis 2002, NOT "Baumeister and Hawkins", which is their 2004 paper on the serotonin hypothesis.

Caveat on this rating: Two corrections to the popular version are carried here deliberately. The first psychiatric dose was given at Val-de-Grace by Hamon, Paraire and Velluz on 19 January 1952; Delay and Deniker began at Sainte-Anne on 24 March 1952 and are correctly credited with establishing chlorpromazine as an antipsychotic, not with the first dose. The synthesis date is NOT stated on this row: Ban 2007 gives 11 December 1951 and popular accounts give December 1950, and only the Ban date is confirmed from a peer-reviewed source we could read, so the row says nothing rather than picking one. Delay and Deniker's original 1952 papers are not in Crossref and have no DOI; their dates rest on Ban 2007.

Caveat on this rating: The absence is the finding, and it is labelled as one: no accuracy percentage for self-identification exists anywhere in this literature, so none appears here in either direction. The 73-89% referral-yield range (Ashwood et al. 2016; Robinson et al. 2025) is not concordance data and is presented as what it is, a selection-biased upper bound from clinic-referral samples. A sampling caution binds much of the surrounding literature: samples recruited through social media show a reversed sex ratio, higher employment and education, and fewer people with intellectual disability than population expectations (Rodgaard et al. 2022, doi:10.1002/aur.2777). Separately verified and equally absent: no peer-reviewed prevalence estimate of social-media-driven self-identification, and no longitudinal study following self-identifiers through to formal assessment. Industry figures circulating on those points are unverified and are not cited.

Caveat on this rating: Scope and construct, both mandatory here. The 59-72% range is an exploratory projection, not a measurement: recorded diagnoses in English primary-care data were compared against community prevalence estimates derived from separate instrument-based studies, so the range inherits the uncertainty of both inputs. It is England; no equivalent US series exists, and the citable US statement on adult assessment access is qualitative rather than statistical. What the age gradient documents is a change in who gets identified, not a change in who is autistic, and nothing here is a claim about any individual's diagnosis or support needs. The waiting-time figures behind the queue sentence are NHS England's "Autism Waiting Time Statistics" for June 2026 - 294,792 open referrals, 86.8% open at least 13 weeks - published as official statistics in development, whose own caveats note that community-paediatric activity is not yet separately identifiable.

Caveat on this rating: The bridge-study numbers cut against a tidy narrative in both directions, so both halves are printed here and neither may travel alone. The 2018 case-definition change is a genuine construct break, and the series must not be spliced across it. And the change's measured size - approximately no difference for 2014, 7% lower for 2016, with about 86% of children meeting both definitions - is far too small, and points the wrong way, to explain the 18.5 to 23.0 per 1,000 step it sits under. Quoting either half without the other misleads. One scope limit: the bridge analysis compares the two definitions on ADDM data for 2014 and 2016 only, and is not a general estimate of the change's effect in later surveillance years.

Caveat on this rating: Both flat verdicts are excluded by the record, in both directions. No peer-reviewed decomposition attributes 100% of the rise to changed counting, and none supports calling the whole rise real and unexplained either; every quantified study leaves a residual it cannot account for, and this row prints the intervals that carry it. The scope caveats are load-bearing: Hansen is Danish registry data and two named reporting changes, with a CI from 33 to 87 that leaves a large possible residual; King & Bearman is one US state's service registry and one pathway, published alongside four commentaries in the same issue of the journal and actively debated in print. The two figures are not commensurable and are neither summed nor averaged here. This row also makes no overdiagnosis claim: none of these studies measures overdiagnosis, none supports "most diagnoses are wrong," and none licenses any statement about an individual's diagnosis.

Caveat on this rating: Scope, stated as the paper states it: Sweden, birth cohorts 1993-2002, with the phenotype measured by a validated parental telephone interview in a twin cohort and diagnoses taken from the national patient register. The authors say administrative changes "are important for" the increase - not that they account for all of it - and no decomposition in this literature attributes 100% of any rise to changed counting. A flat symptom trend in one country over one decade is strong evidence about that country and that decade; it is not a measurement of US surveillance counts, which use a different instrument again, and it must not be read as one.

Caveat on this rating: The register here is fixed by the verified record and is not optional. This row states that the predicted drop did not appear in surveillance counts, and never that DSM-5 caused the rise: the rise long predates 2013, and DSM-5's measured fingerprint in the impact literature is a DECREASE. Kulage 2014 predicted -31% (CI 20-44); this paper measured -20.8% (16.0-26.7); Maenner 2014 found 81.2% of CDC surveillance cases retained under DSM-5; Huerta 2012 found 91% retention in clinic samples. A second distinction is load-bearing: these meta-analyses pool STUDY SAMPLES, and the surveillance series counts records at sites - they are different quantities measured by different instruments, and this row keeps them apart rather than subtracting one from the other.

Caveat on this rating: The doubling is not contested; its reach is limited by design, and the limits are the point. NHIS measures parent-reported "ever told," ages 3-17, with no records checked - it cannot be placed on one axis with CDC's records-based site network, and it cannot be read across its own 2014 and 2019 breaks as a single trend. One precision point carried from the verified record: NCHS's own language about the 2019 redesign is a hedge, not a prohibition. A flat NCHS sentence forbidding comparison across 2019 does not exist - it was searched for and not found - and the blunt "should not be directly compared" phrasing that circulates belongs to a different CDC programme page, not to NCHS survey documentation. Attribute it correctly or not at all.

Caveat on this rating: Not a contested finding - a contested reading. The 32.2 per 1,000 figure is real, primary and correctly published; what the record does not support is calling it a national rate, or a count of diagnoses. ADDM measures records-based IDENTIFICATION among 8-year-olds at a competitively selected, non-representative and changing set of sites, and CDC states both limits in the report's own limitations section. The site range is printed here because it is the agency's own device for showing it. This figure must never be placed on one axis, or in one sentence as a trend, with parent-reported survey series (NHIS, NSCH) or with instrument-assessed adult epidemiology: those are different constructs, not different estimates of the same quantity.

Caveat on this rating: The mandatory contested label: burnout-versus-depression distinctness is unresolved in either direction - Bianchi, Schonfeld & Laurent 2015 ("the distinction between burnout and depression is conceptually fragile") against Maslach & Leiter 2016 (the construct stands), with Rotenstein et al. 2018 in JAMA separately documenting how inconsistent the literature's definitions are. Burnout as an occupational phenomenon is strong; burnout as a medical diagnosis is a non-entity by the WHO's own design, and "nervous breakdown" was never a diagnosis in any manual.

Caveat on this rating: The construct grades strong; what fails is the pop typology, and against the same literature. Taxometric and modern replication work (Fraley & Waller 1998; Fraley, Hudson, Heffernan & Segal 2015) supports dimensions over types, and Pinquart, Feussner & Ahnert 2013 finds infancy-to-adult continuity weak to moderate - both planks of the quiz version, not the construct. One negative is labeled rather than asserted: no peer-reviewed critique of the book Attached was located as of 2026-08-26.

Caveat on this rating: Tier labels are mandatory here and are stated in the body: the peer-reviewed debunk corpus does not exist (the only indexed paper on dopamine fasting is descriptive, in a low-barrier venue - Desai et al., Cureus 2024), so the mechanism debunk rests on institutional expert commentary (Harvard Health, Grinspoon 2020) backed by the peer-reviewed dopamine literature (Schultz, Dayan & Montague 1997; Berridge & Robinson 1998 and 2016). The citation on this row is the mechanism literature, not a debunk study - no such study exists.

Caveat on this rating: The house pairing, stated in the body and repeated here: Grossman 2023 ("likely refutations of the five basic premises") and the vagal-anatomy critique (Neuhuber & Berthoud 2022) on one side, Porges's 2023 reply on the other - unresolved in the literature, so polyvagal theory is presented as neither established nor refuted. HRV and autonomic research stands independently of that dispute and grades moderate; "nervous system dysregulation" as a clinical entity exists in no diagnostic system.

Caveat on this rating: The labels here are mandatory, per the verified record: body involvement in trauma grades strong (Pole 2007; the interoception literature); literal somatic memory storage has no mechanism in the literature; the book's own claims are contested in a peer-reviewed appraisal (Scheeringa 2025, BJPsych Bulletin) and the repressed-memory framing is contradicted (Otgaar et al. 2019). This row cites the founding 1994 paper as lineage, not endorsement - the critiques travel with it, here and in the body.

Caveat on this rating: The descriptive finding (the word's extension measurably widened) is settled; the evaluative claim (that the widening is harmful) is not. Cascardi and Brown ("Concept Creep or Meaningful Expansion?") and Brandt and Proulx, in the same issue of Psychological Inquiry as Haslam's target article, argue the expansion can be principled and protective and that concept change is normal science. This row states the widening as fact and the harm claim as the live dispute it is.

Caveat on this rating: CDC publishes this data as four disjoint segments that must not be joined into one line, so nothing here is a continuous 1997-2024 trend. Two federal surveys measured the same years and disagreed. The underlying measure is parent-reported "ever diagnosed", which drifts upward mechanically because it is cumulative rather than current.

Caveat on this rating: The NIMH passages quoted here are from the Director's Blog of April 2013. That URL now returns 404 and the text is quoted from an Internet Archive capture dated 27 June 2021 - the second primary US government document in this series that has to be read from a private archive. The often-quoted figure for how many diagnoses each DSM edition contains is deliberately omitted: no citable primary source states it, and the number varies with whether subtypes, residual categories and provisional conditions are counted. It is the most quotable statistic in this subject and the easiest to refute.

Caveat on this rating: The umbrella review is strong and it is contested, and both belong beside it. It was pre-registered with PROSPERO, unfunded, and published in a top-quartile journal - and it loses points on allegiance in our own rubric because its lead author is the most prominent proponent of the position it supports, which is a recorded fact about the paper rather than an accusation. Jauhar and colleagues published a direct rebuttal in the same journal arguing the conclusion is overstated on grounds of method and interpretation. This is not settled in either direction.

Caveat on this rating: The trial-duration count quoted here is our own audit of this library, not an independent finding: the denominator is the 72 randomised trials in this corpus, not the psychiatric literature as a whole. The FDA's phase participant numbers and durations are descriptive ranges the agency publishes about its own process, not measured averages across approvals.

Caveat on this rating: The declarations of interest read literally: "KH has previously acted as a temporary consultant for Pfizer (manufacturers of sildenafil). MT has been paid to lecture and received travel expenses from Bristol-Myers Squibb (manufacturers of buspirone) and Otsuka; his spouse is an employee of GlaxoSmithKline (manufacturers of bupropion)." Two of the six authors therefore have financial ties to the makers of the two drugs the review endorses, which is why independence is scored false despite the funding being public. The reviewers also warn that partial reporting of subscale results in the source trials could bias the effect estimates upward.

Caveat on this rating: A regulatory label change is a precautionary act on a safety signal, not proof of incidence, causation, or permanence. The evidence base the PRAC weighed was EudraVigilance reports, literature, social media, and manufacturer reviews. Clomipramine and vortioxetine were part of the same signal assessment but were explicitly not covered by the labelling recommendation. US FDA labelling does not carry equivalent wording.

Caveat on this rating: The authors state that including open-label studies and pooling across different sexual-function scales "could reduce the significance of our findings," so the 25.8-80.3% band is wide partly because the source studies were not uniform. No funding statement or conflict-of-interest disclosure was accessible for this paper.

Caveat on this rating: Dated: the search closed 31 December 2011 and the review has not been updated, so it predates the trauma-focused work of the last decade. Its conclusion applies to brief generic counselling as trialled up to 2011 and should not be read as evidence against structured psychological treatment for post-loss PTSD, depression, or complicated grief.

Caveat on this rating: The high trust score reflects the review's method, funding, and journal, not the strength of the underlying trials: 398 men across 11 studies, several from the 1970s and 1980s, with waitlist controls and confidence intervals touching 1.0. The single trial claiming group therapy outperformed sildenafil alone (WMD -12.40 on the IIEF) had 20 participants and was run by one of the review's own authors, which is a direct conflict on the review's most eye-catching result.

Caveat on this rating: "Inadequate evidence" for exercise, education, and the drug options means not enough good trials existed in 2019, not that those approaches were shown to fail. The recommendation is also US-specific and now several years old.

Caveat on this rating: The between-group difference disappeared by 24 weeks, but that is because women found to be depressed at 12 weeks were referred for treatment for ethical reasons, which contaminates the later comparison rather than showing the effect wore off. The trial also found no benefit for loneliness or health-service use.

Caveat on this rating: Only 12 trials, several rated at high risk of bias by the authors' own Cochrane assessment, and the anxiety and stress estimates rest on four and five studies respectively. Too few trials reported partner or infant outcomes to say anything about them.

Caveat on this rating: The high trust score reflects the quality of this meta-analysis as a piece of evidence synthesis, not the strength of the effect it found. The pooled benefit rests on a subjective, dichotomised outcome in six trials with demonstrated publication bias, and the authors' own conclusion is the hedged "valerian might improve sleep quality," followed by a call for better studies. Twenty years on, those studies have largely not materialised.

Caveat on this rating: These are screening-instrument prevalences drawn largely from cross-sectional studies with substantial heterogeneity, so they measure symptom burden rather than diagnosed disorder, and the pooled percentages are sensitive to which scales and cut-offs the contributing studies used.

Caveat on this rating: This review reaches the opposite conclusion to Bent 2006 from overlapping trials, and the difference is instructive: Taibi weighted trial quality and recency, and found that the better and newer the study, the smaller the effect. That pattern is the signature of a treatment whose apparent benefit comes from weak methods.

Caveat on this rating: Most contributing studies were cross-sectional and used self-report screening scales such as the EPDS rather than diagnostic interviews, so this is a symptom-threshold prevalence and will run higher than the rate of diagnosed major depression. The article also carries a published Correction (PMID 34930896).

Caveat on this rating: This is a rigorous review of an essentially empty evidence base, and it is now twenty years old. Absence of evidence is not evidence of absence — but people buying valerian for anxiety should know that a full Cochrane search found one 36-person pilot study, and it was negative.

Caveat on this rating: The abstract's conclusion — that "valerian could be a safe and effective herb to promote sleep" — is not supported by the review's own primary pooled result, whose confidence interval crosses zero. The whole-root subgroup finding that the authors lean on rests on four trials and a confidence interval that nearly touches zero, and it is a post-hoc subgroup analysis, the weakest kind of evidence. One co-author is affiliated with Heartwood Education, a UK herbal-medicine education organisation, so the review is not independent of the herbal-medicine field it evaluates.

Caveat on this rating: The control groups were not uniform — they mixed live births, difficult live births, and non-pregnant community samples — and "perinatal loss" spans early miscarriage through stillbirth, so the pooled risk ratios blur real differences between very different experiences.

Caveat on this rating: Both arms improved similarly on the primary desire and arousal outcomes, so this trial does not show that mindfulness specifically works — it shows that eight weeks of structured group attention works. There was no no-treatment or waitlist arm, so regression to the mean and expectancy effects cannot be separated out. All 148 participants were cisgender women, and the authors note the need to diversify samples.

Caveat on this rating: Only psychiatric conditions severe enough to require hospitalisation were captured, so ordinary depression and anxiety treated in primary care are invisible here — the absolute risks are small and the affective-disorder result should not be read as evidence that unsuccessful treatment protects against depression.

Caveat on this rating: The pregnancy-rate result is the weak half of this paper: the pooled studies were heterogeneous, not all randomised, and the finding directly contradicts the larger and better-controlled Boivin 2011 BMJ meta-analysis showing distress does not affect cycle outcome. Read this study as evidence that psychological support reduces distress, not as evidence that it makes you pregnant.

Caveat on this rating: The effect size depends heavily on how depression was measured, and the authors applied trim-and-fill because of publication bias. These are observational associations: depression clusters with poverty, poor nutrition, smoking, and limited prenatal care, and the pooled estimates cannot fully separate those.

Caveat on this rating: A WEAK GRADE recommendation against a treatment is not a finding that it is dangerous; it means the trials are too few, too small, and too inconsistent to justify recommending it. The guideline also notes that it downgraded evidence quality across the board because most insomnia drug trials are industry-funded and carry publication-bias risk. It is now nine years old.

Caveat on this rating: The author notes that methodological quality varied widely across the included trials, and the pool mixes properly randomised with quasi-randomised studies. The abstract does not report a pooled participant total or the numeric effect sizes, so sample size is recorded as unknown rather than guessed.

Caveat on this rating: Heterogeneity between studies was substantial and the 3-to-6-month peak rests on a small subset with a very wide interval (17.3-36.1%). Most included studies used screening instruments validated in women, which may not capture how depression presents in men.

Caveat on this rating: The confidence intervals around the diagnosed-disorder estimates are wide (15.2%, 95% CI 9.0-21.4 antenatally), reflecting real heterogeneity between studies and instruments. Rates were higher in low- and middle-income countries, so a single global figure understates the spread.

Caveat on this rating: Attrition was substantial and grew over follow-up: 67% of the loss cohort completed the one-month questionnaire, 58% at three months, and 46% at nine months, so the later percentages rest on a shrinking and possibly non-representative subgroup. These are validated screening thresholds, not clinical diagnoses.

Caveat on this rating: The incidence estimate has a very wide confidence interval (4-20%), and prevalence differed sharply by region (26% in the Middle East vs 8% in Europe), so the single pooled number hides large real variation. No funding statement was retrievable.

Caveat on this rating: Partner response rates were low and fell over time (60%, 48%, and 39% at 1, 3, and 9 months), and all partners in the study were male, so nothing here speaks to same-sex-partner or non-birthing-mother grief. Men may also under-report distress on these instruments, which would bias the gap wider than it truly is.

Caveat on this rating: Every pooled effect is against a waitlist, not an active or placebo control, so attention and expectancy are baked into the d = 0.58 figure. The search ends at 2009, predating the internet-delivered treatments that now dominate access. No total participant count is reported in the abstract, and the MEDLINE record lists the publication type "Research Support, Non-U.S. Gov't", indicating unnamed external support.

Caveat on this rating: Only eight citations survived screening, yielding six studies per direction, and the depression-to-dysfunction association lost statistical significance in the partially adjusted model (RR/OR 1.41, 95% CI 0.90 to 2.23) until an outlying study was removed. Heterogeneity between studies was significant. Prospective cohorts show sequence, not causation.

Caveat on this rating: The funding statement reads: "This study was supported by an independent investigator-initiated grant from Pfizer Inc, New York, New York. Pfizer Inc provided sildenafil and matching placebo," with the role-of-sponsor statement adding that "Pfizer had no other role in the design and conduct of the study." Pfizer makes sildenafil, and several authors disclose financial relationships with it. The Cochrane review of this literature concluded that for women it remains uncertain whether sildenafil is more effective than placebo, and flagged that unpublished data could change the picture — a direct disagreement a reader should know about.

Caveat on this rating: A convenience-sample study design was associated with substantially higher odds ratios for premature delivery (OR 2.43) than better-sampled studies, which suggests the headline association is inflated by weaker studies. The authors explicitly call for higher-quality research. No funding statement could be read; the full text is paywalled.

Caveat on this rating: The authors call this a pilot trial, and it reads like one: the headline erectile-dysfunction result missed significance at P = 0.065, the premature-ejaculation arm was flatly negative, and the control condition was a waitlist rather than an active comparison. Sexual self-confidence — the outcome closest to performance anxiety — was a secondary measure, and improved only marginally (P = 0.05). No funding statement was available.

Caveat on this rating: The funding statement reads: "This study was supported by an independent investigator-initiated grant from Pfizer Inc, New York, NY." Pfizer makes sildenafil. Several authors also disclose research support, consulting fees, or speakers-bureau payments from Pfizer. The trial predates routine registration, so there is no pre-registered protocol to check the reported outcomes against.

Caveat on this rating: The authors say plainly that because the study is cross-sectional, "causality versus comorbidity cannot be determined" — this is overlap, not a demonstration that depression causes low desire. Depression was defined partly by antidepressant use, which itself lowers desire, so the two exposures are entangled. The MEDLINE record lists the publication type "Research Support, Non-U.S. Gov't", indicating unnamed external non-government support.

Caveat on this rating: This is a one-time self-report survey (the journal graded it Level of Evidence III), so it can measure prevalence and correlation but not cause or direction. The MEDLINE record carries the publication type "Research Support, Non-U.S. Gov't", meaning the study had external non-government support that the record does not name; the funding statement itself could not be reached, so funding is recorded UNKNOWN rather than guessed. Prevalence figures from surveys of this kind have been used commercially to size markets for desire drugs, which is a reason to read the distress threshold carefully rather than the headline 43.1%.

Caveat on this rating: This is a single-author narrative review with no systematic search protocol and no pooled analysis; the author's affiliation is his own consultancy, Pykonsult LLC, and no funding or conflict-of-interest statement was accessible. Its treatment section is explicitly hypothesis-generating — it applies "pharmacologic principles" to drugs and phytotherapies including buspirone, trazodone, yohimbine, l-citrulline, and bitter orange that have not been tested for sexual performance anxiety. Treat the prevalence range as the usable part and the supplement suggestions as untested.

Caveat on this rating: The review's own evidence tables record that a substantial share of the included trials were industry-funded, had a manufacturer author, or were supplied with the product by the manufacturer, and that funding was simply unreported for several others. The authors also note that adverse events were poorly reported in the source trials and that none of the studies was designed to detect rare harms — including the drug interactions that are St. John's wort's main danger.

Caveat on this rating: The paper carries a published erratum (Transl Psychiatry 2021;11(1):465) and a published critique (Transl Psychiatry 2022;12(1):298); neither is a retraction, but both belong alongside it. Dose and formulation subgroups drawn out of a meta-analysis of heterogeneous small trials are exploratory — the Cochrane review reached a similar overall effect size and judged it very low certainty and likely biased upward.

Caveat on this rating: Nine small trials totalling 317 people, published in 2010, and the benefit is a shift of roughly 40 minutes to an hour in when sleep starts — real, but modest, and it delivers no extra total sleep. Effect depends heavily on taking the dose at the right time relative to one's own melatonin onset; taken at the wrong hour it can push the clock the wrong way.

Caveat on this rating: The higher depression rate was a borderline result on one of two coprimary outcomes while the other (mood scores) showed nothing at all; the authors did not claim omega-3 causes depression, and this was a prevention trial in people without clinically relevant depressive symptoms at baseline, so it does not test omega-3 as a treatment for existing depression. Study capsules and matching placebo were donated by Pronova BioPharma, an industry in-kind contribution, though the trial itself was NIH-funded, the funders had no role, and the result was unfavourable to the product.

Caveat on this rating: A 39-person trial has limited power, and the authors discuss whether the frequent study visits and psychoeducation inflated the placebo response — so this is a failed replication rather than proof that NAC does nothing in children. It is included precisely because the positive adult trial is usually cited without it.

Caveat on this rating: All 14 recommendations in this guideline are WEAK, including the ones in favour of prescription hypnotics, and the task force explicitly notes that evidence quality is routinely downgraded because most hypnotic trials are industry-funded. So the guidance against melatonin sits in a field of uniformly thin evidence rather than reflecting a uniquely poor showing. Several task-force members disclosed industry consulting relationships and recused themselves from the corresponding drug recommendations.

Caveat on this rating: Entirely observational. Mostly cross-sectional studies, which cannot establish that low magnesium came before the depression rather than after it, and dietary magnesium intake is a proxy for overall diet quality, socioeconomic position, and physical activity. No supplement was tested in any of the pooled studies.

Caveat on this rating: The authors themselves report evidence of moderate publication bias (contour-enhanced funnel plot, significant Egger's test, and a significant subgroup analysis on timing of the pregnancy test). Separately, the first author disclosed speaker fees from EMD Serono and Merck and a research grant from Merck Serono in the three years before publication, though the paper itself received no organisational support.

Caveat on this rating: Blinding was almost certainly compromised: 95.6% of methylene blue participants reported blue-green urine discoloration versus 21.1% on placebo. The subgroup finding that drives the headline result was predicted in advance but comes from splitting 42 people by post-training fear level, so it needs replication before it means much.

Caveat on this rating: A null result across a wide credible interval means "not proven to work", not "proven not to work" — the L-theanine arm of this network rests on few small trials, so a modest real effect cannot be ruled out. The analysis came from a traditional Chinese medicine research institute, which if anything would bias it toward finding herbal benefits, making the null finding for L-theanine harder to dismiss.

Caveat on this rating: With 34 participants the study is underpowered to detect small true effects, and the authors' own Bayesian analysis in the successfully blinded subgroup was inconclusive rather than clearly negative.

Caveat on this rating: These are screening-questionnaire symptom rates, not diagnosed major depressive disorder, and heterogeneity between the pooled studies was high. The published confidence interval for the HADS estimate (21.01%, 95% CI 15.61-34.42) is asymmetric around the point estimate as printed in the paper.

Caveat on this rating: All three review authors disclosed financial relationships with Schwabe, a hypericum manufacturer: one had received a research grant and conference speaking fees, and all three had accepted travel reimbursement. The review itself flags that country of origin and trial size both predicted the result, which the authors say "complicates the interpretation" — plain language for probable bias in the underlying literature. The review is from 2008 and has not been updated.

Caveat on this rating: The competing-interests statement records that the work was "partly funded by APIRE/Eli Lilly Psychiatric Research Fellowship" and "the AACAP/Eli Lilly Pilot Research Award" alongside NIH grants, so the funding is mixed rather than fully independent — though Eli Lilly markets no melatonin product, so there is no proponent conflict on this specific question. The pooled 7-minute and 8-minute effects are statistically significant and clinically trivial; readers who see melatonin described as "effective" are usually seeing this paper's conclusion divorced from its numbers.

Caveat on this rating: The authors' own stated limitation is the small sample: 37 people cannot rule out a modest real benefit, so this is an absence of evidence rather than firm evidence of absence. The 120 mg/day dose is also low relative to the 248 mg used in the Tarleton trial, which the authors themselves flag as a possible explanation.

Caveat on this rating: Open-label with a no-treatment control rather than a placebo. That single design choice makes the headline 6-point PHQ-9 improvement uninterpretable as a drug effect: placebo responses of that magnitude are routine in depression trials, and here nothing separated the magnesium effect from the effect of knowingly starting a treatment. The authors' conclusion that "magnesium is effective for mild-to-moderate depression" outruns what an unblinded design can show.

Caveat on this rating: Only four eligible studies existed, none of them powered to measure pregnancy rates, so the evidence establishes a mechanism and a warning signal rather than a quantified failure rate. Different St. John's wort products produced different degrees of interaction, which means a negative result for one brand does not clear the others.

Caveat on this rating: The abstract reports p<0.001 alongside pooled risk ratios whose 95% confidence intervals comfortably include 1.0, which is internally inconsistent; read the confidence intervals, which show equivalence rather than superiority. This meta-analysis also inherits the industry funding and country-of-origin biases present in the underlying trial literature, and it did not assess drug interactions at all.

Caveat on this rating: This is a failed trial, not simply a negative one: escitalopram, a drug with established efficacy, also failed to beat placebo, which means the study lacked assay sensitivity and cannot cleanly rule SAMe out. Two later papers re-analysed subsets of the same dataset and reported positive findings — a single-site subsample (Sarris 2014, J Affect Disord, PMID 24856557, n=144, SAMe superior to placebo) and a male-only subgroup (Sarris 2015, Pharmacopsychiatry, PMID 26011569, significant in 51 men, not in 62 women). Post-hoc subgroups from a failed trial are hypotheses, not results.

Caveat on this rating: The meta-analytic estimates rest on very little randomised data — two parallel RCTs (three comparisons, n=117) for efficacy and two RCTs (n=109) for adverse events — so the confidence intervals are wide and the review is better read as demonstrating absence of evidence than proving absence of effect. Several authors are affiliated with psychedelic research centres and one is chief medical officer of a psychedelics company (disclosed), so the null result is not coming from sceptics.

Caveat on this rating: Participants sourced and weighed their own material, so actual doses were unverified, and the sample was self-selected enthusiasts rather than people with a diagnosed condition. The authors are based at Imperial College's Centre for Psychedelic Research with a co-author from the Beckley Foundation, both proponents of psychedelic research — which cuts against, not toward, a bias explanation for this null result.

Caveat on this rating: The mineral analysis pooled iron, zinc, and magnesium together rather than testing magnesium alone, and the population was perinatal, so this is a null result in a specific setting rather than a definitive negative for magnesium in depression generally. The wide confidence interval reflects few trials.

Caveat on this rating: One trial, 50 people, one site, and a very large effect — the combination that most often fails to replicate, and in this case did fail to replicate in children four years later. No funding or disclosure statement could be read (the journal page returned HTTP 403 on 2026-08-18), so funding is UNKNOWN, conflicts false, and independence left null.

Caveat on this rating: A clean null on the primary outcome, with post-discontinuation secondary signals that the authors themselves flag as needing explanation. Study medications and placebos were supplied by commercial nutraceutical companies (Nutrition Care, BioCeuticals and Catalent, Australia) alongside public NHMRC and CRC funding, hence MIXED; the senior author's disclosed consultancy to Bioadvantex, a NAC commercialiser, is why independence is marked false. A published erratum (BMC Med 2019;17(1):35) applies.

Caveat on this rating: Uncontrolled and unblinded: participants knew they were dosing, sourced their own substances, and were self-selected enthusiasts, so nothing here can separate drug effect from expectation. The authors say as much and explicitly call for dose- controlled research.

Caveat on this rating: Two of the authors have direct commercial ties to SAMe: the disclosure statement records that Dr Bottiglieri chaired the advisory board of and holds stock options in Methylation Sciences Inc. and is a scientific advisor to Gnosis S.p.A., and Dr Gerbarg receives royalties from books covering SAMe. The review received no funding, and it reports the mania risk plainly, but it is not written by parties independent of the product. It also pools trials of widely varying quality, dose, and route without a meta-analysis, so "promising but limited" is a judgement rather than an effect size.

Caveat on this rating: The comparator was not inert: the "placebo" arm was 15 mg/day of methylene blue, chosen to preserve blinding because the drug discolours urine. Since a 1987 trial by Naylor reported that 15 mg/day itself acted as an antidepressant, the true drug-versus- nothing effect could be either larger or smaller than the reported difference.

Caveat on this rating: The review did not pool the trials or grade certainty of evidence, so "may assist" is a narrative judgement rather than a measured effect; none of the nine trials studied people with a diagnosed anxiety disorder, and several were manufacturer-funded.

Caveat on this rating: 73 participants across two arms is small enough that a handful of responders moves the result, and the trial has not been independently replicated in the sixteen years since — the Cochrane review, published six years later, still cites it as a single study. No funding statement or disclosure statement was read, so both funding and conflicts are recorded conservatively.

Caveat on this rating: This is a systematic review of case reports, not of trials. Case reports capture only the people who reached a hospital and were written up, so they show what phenibut can do but cannot tell you how often it does it.

Caveat on this rating: ISNPR is a society organised around nutritional psychiatry, so it is the proponent of the intervention it is writing guidelines for, and the panel includes investigators whose own omega-3 trials form part of the evidence base being summarised. Two letters challenging the guideline appeared in the same journal (Psychother Psychosom 2020;89(1):48, PMID 31655818, and the authors' reply at 89(1):49, PMID 31743919). No funding statement was located, so funding is recorded UNKNOWN.

Caveat on this rating: The positive signals appeared at week 16, after treatment had been discontinued at week 12, and in secondary outcomes and a severity subgroup — the classic shape of findings that do not replicate. The trial is from the research group that has published most of the positive NAC literature, and its senior author's disclosures elsewhere record consultancy to Bioadvantex, a NAC commercialiser, hence independent false. No funding or disclosure statement was readable for this paper.

Caveat on this rating: Two of the five authors — Berk and Dean — are investigators on trials included in the pooled estimate, so this is not an independent synthesis of other people's work, and Berk's disclosures elsewhere in this literature record consultancy to Bioadvantex, a company commercialising NAC. The search ended in November 2014 and so predates the larger negative bipolar trial listed below. No funding statement was readable, so funding is UNKNOWN.

Caveat on this rating: Thirty-one samples of Canadian products bought in 2016; it does not measure how common the problem is across the whole market, and the later US gummy analysis (Cohen 2023, below) did not detect serotonin in any product it tested, so the serotonin contamination finding has not been replicated in the US supply.

Caveat on this rating: The published abstract gives no sample size, no effect size, and no confidence intervals, so the strength of the result cannot be assessed from the record; sample size is recorded as unknown rather than guessed. The dose it found effective, 15 mg/day, is the same dose the same group's 1986 trial and Alda's 2017 trial both used as the comparator — a direct internal contradiction in this small literature that has never been resolved.

Caveat on this rating: Nearly half the enrolled participants did not complete, the comparator was a low dose of the same drug rather than an inert placebo, and methylene blue's blue-green urine makes blinding difficult at any dose — all conceded in the paper. Everyone was also on lithium, so this tests methylene blue only as an add-on.

Caveat on this rating: Poison-centre data are passive surveillance: exposures are self-reported by callers, are not laboratory-confirmed (phenibut is not detected on routine urine drug screens), and most people who use phenibut never call, so these numbers are a floor, not an incidence rate. Some of the rise after 2015 reflects "phenibut" being added as a searchable term that year.

Caveat on this rating: The authors state plainly that the whole data set is case reports and is therefore subject to publication bias, that reported doses may not reflect what unregulated products actually contained, and that people describing single-substance use may have been using other drugs too.

Caveat on this rating: The authors state their own limits: only 25 products, one sample per brand, gummies only, and no test of whether tablets and capsules behave the same or whether a given brand varies batch to batch. It measures label accuracy, not clinical harm.

Caveat on this rating: The acknowledgments state that two of the three authors "received funding from Sanofi to conduct the initial systematic publication database search" and that "Sanofi also provided access to data from 3 unpublished Mg intervention studies"; the senior author separately discloses membership of the Sanofi Consumer Healthcare Advisory Board. Sanofi markets magnesium supplements, so the sponsor had a direct commercial interest in the review's conclusion, and unpublished sponsor-held data entered the evidence base without independent scrutiny. Many included studies tested magnesium combined with up to five other ingredients.

Caveat on this rating: "This study and the journal's Rapid Service Fee was funded by Ethical Naturals Inc., California, USA", the company that sells the AlphaWave L-theanine tested; the trial was run at a contract research organisation. With 16 participants and one participant lost after period 1, the trial is far too small to support the conclusion in its own discussion that the product "should be considered as a nutraceutical supplement to manage acute stress."

Caveat on this rating: "The study was funded by a grant from Neurobrands CA, USA", a beverage company with a commercial interest in the L-theanine drink tested; the authors state the funder had no input into design, conduct, or interpretation. The tested product was a multi-ingredient drink, not isolated L-theanine, so the effects cannot be attributed to L-theanine alone.

Caveat on this rating: Funded by Taiyo Kagaku Co., Ltd., which manufactures the Suntheanine-brand L-theanine; two of the seven authors are employees of that company and it supplied both the active and placebo tablets. Several outcomes were measured within-person before-and-after rather than against placebo, which inflates apparent effects.

Caveat on this rating: Only four brands met the inclusion criteria, so this is a small snapshot rather than a market-wide survey; it demonstrates that labelled dose cannot be trusted, not the exact proportion of products that are mislabelled.

Caveat on this rating: A single-state case series of 56 calls, with coingestants documented in 36% of patients, cannot separate phenibut's effects from other drugs and cannot estimate how common these outcomes are among people who use phenibut. It scores low on the evidence rubric by design — it is a description of hospital cases, not a study of whether phenibut helps or harms.

Caveat on this rating: Three of the four authors are employees of Max Zeller Söhne AG, a manufacturer of a St. John's wort herbal medicinal product, and this is a narrative review rather than a systematic one — so it sits at the bottom of the evidence rubric and its framing should be read with the authors' commercial interest in mind. Note the direction of that interest: authors paid by an SJW manufacturer nonetheless document the interactions, which makes the interaction findings harder, not softer.

Caveat on this rating: Open-label by necessity (neither ketamine nor ECT can be masked), so subjective outcomes may favor the treatment patients preferred; population excluded psychotic depression, where ECT performs best.

Caveat on this rating: One co-author (Ware) became an employee of Canopy Growth Corporation, a licensed cannabis producer, after the work was completed, and the senior author discloses an unrelated cannabidiol grant involving Aurora Cannabis; both were disclosed. The pooled estimates come from observational studies, so residual confounding by shared risk factors for cannabis use and depression cannot be excluded.

Caveat on this rating: A 5-point PHQ-9 difference is large for a supplement, which is itself a reason for caution in a 100-person pilot that was explicitly powered for feasibility rather than efficacy and was conducted at a single site in a setting where access to other care is limited. No trial registration appears in the PubMed record, so the primary outcome cannot be checked against a pre-specified protocol. It has not been replicated.

Caveat on this rating: The credible interval runs from -9.70 to -0.17, meaning the true effect could be near zero; the ashwagandha estimate rests on a handful of small trials, most of which were supplied or funded by extract manufacturers. The analysis was conducted at the Beijing Research Institute of Chinese Medicine, an institution with an institutional interest in herbal medicine generally, though its funding came from the university and Chinese national science programmes rather than from any supplement company.

Caveat on this rating: The rubric score overstates this trial badly. It was single-blinded with a no-supplement control rather than a placebo capsule, so expectancy is uncontrolled; it enrolled 30 people; and the mood change it reports is half a point on a scale where the group started at 1.2 out of 15, which is not clinically interpretable as treating depression. It is included as the most recent randomised 5-HTP evidence that exists, which is itself the point: this is what "current evidence" looks like for 5-HTP.

Caveat on this rating: This umbrella review's STRONG tier reflects its independent, systematic method — its actual conclusion is that the underlying MDMA evidence is weak. Read the tier as confidence in the review, not in MDMA.

Caveat on this rating: The disclosure statement records that the senior author, Dr Renshaw, "is also an inventor on related patent applications assigned to McLean Hospital and the University of Utah" — an intellectual-property interest in the intervention being tested, which is why independence is marked false here even though the funding is public. The trial was women-only, single-country, and n=52; a later trial in female adolescents found no difference from placebo.

Caveat on this rating: Diphenhydramine is a weak psychoactive control; the paper's own blinding assessment showed most participants correctly guessed their assignment, so expectancy effects cannot be excluded.

Caveat on this rating: Investigators are prominent psychedelic-research advocates; blinding was imperfect in both arms. The null primary endpoint is often glossed over in media coverage.

Caveat on this rating: The declaration of conflicting interests records that co-author Dr Candow "serves on the Scientific Advisory Board for Alzchem and Create (companies that manufacture creatine)" and has received industry creatine research funding and product donations, and that Dr Ostojic "is the co-founder of KRE-ALL, a company which develops food products enriched with creatine" — commercial ties to the intervention being reviewed, which is why independence is marked false despite the review itself being unfunded and its conclusions being cautious. No PROSPERO registration was found on the published page.

Caveat on this rating: Waitlist-controlled (no placebo), very small N, single site with strong psychedelic-research identity; effect sizes this large rarely survive larger placebo-controlled replication.

Caveat on this rating: The remission odds ratio of 3.6 looks dramatic but rests on three trials and is contradicted by the response outcome from two trials pointing the other way — a pattern that usually means too few events, not a real dissociation. The reviewers' own bias analyses point the same direction. No funding or competing-interests statement could be read (the Cambridge Core page returned only the abstract), so both are recorded conservatively.

Caveat on this rating: Single site, N=29, 7-day primary horizon, and ayahuasca's intense somatic effects (nausea, vomiting) make blinding doubtful despite the inert placebo.

Caveat on this rating: The pooled odds ratio looks impressive but rests on 64 patients with a confidence interval running from 1.28 to 13.15 — statistically positive and practically uninformative. The review is from 2002 and predates almost nothing, because almost nothing usable has been published since.

Caveat on this rating: Crossover design with a discriminable active dose means most participants likely knew their condition; Heffter-funded investigators are proponents of the therapy.

Caveat on this rating: This is a case-control design, so it establishes association and a dose-response pattern rather than proving causation; reverse causation (early psychotic symptoms driving heavier use) and residual confounding cannot be fully excluded, which is why the authors flag the population-attributable fractions as conditional on causality.

Caveat on this rating: An SMD of -1.55 is roughly three times the effect size seen for SSRIs in anxiety trials, which is implausible on its face; combined with I-squared of 93.8% and the authors' own "low certainty" GRADE rating, the pooled number should be read as a summary of a biased literature, not as an estimate of benefit. The great majority of the pooled trials were run on manufacturer-supplied branded extracts.

Caveat on this rating: A pooled remission rate from mostly uncontrolled trials measures how many people improved, not how many improved because of 5-HTP; depression remits substantially on placebo and with time. Classified here as a systematic review of largely non-randomised evidence rather than a meta-analysis of RCTs for exactly that reason.

Caveat on this rating: Sponsor-employee coauthors and shareholders; N=34; the 3-month durability claim comes from an uncontrolled open-label phase.

Caveat on this rating: Small N, discriminable drug (unblinding likely), academic two-center trial; independent of MindMed/Definium, which is a strength.

Caveat on this rating: Built entirely from observational studies with heterogeneous definitions of "heavy" use; the pooled odds ratio was generated by simulating individual data points from summary statistics, a method that depends on assumptions about the underlying distributions.

Caveat on this rating: Sponsor-funded with sponsor employees as authors; central raters were blinded, but participants were functionally unblinded by design of the drug itself. Phase 3 toplines (Voyage, Emerge, 2026) are positive but press-release-stage, not yet peer-reviewed.

Caveat on this rating: Usona is a 501(c)(3), not a commercial company, but it is the sponsor-developer seeking FDA approval of this exact product — scored as industry funding for that reason. Niacin does not mimic psychedelic effects, so unblinding was likely.

Caveat on this rating: Cross-sectional design: this cannot establish that creatine intake reduces depression risk, and reverse causation is at least as plausible as the causal reading. The paper states the authors have no conflict of interest, but co-author Dr Renshaw is disclosed elsewhere in this literature (in the 2012 AJP creatine trial above) as an inventor on related patent applications, which is why independence is marked false here.

Caveat on this rating: The disclosure states that STI-Pharm supplied the CBD at no cost and that four authors are co-inventors of a fluorinated-CBD patent; the University of São Paulo has licensed that patent and holds an agreement with Prati-Donaduzzi to develop a synthetic CBD product for indications including anxiety disorders. Public grant money plus donated study drug and patent interests is why funding is recorded MIXED.

Caveat on this rating: No placebo arm and no power calculation for equivalence: a trial in which both arms improve and neither wins cannot distinguish "both work" from "neither works and people recovered anyway", and 60 completers is far too few to establish non-inferiority. A single-centre trial that has not been replicated.

Caveat on this rating: N=7; included for historical significance only.

Caveat on this rating: Same sponsor-developer structure and unblinding problem as MAPP1; FDA's CRL asked for a further trial, which the Aug 2026 resubmission does not include (it relies on an audit, a phase 1 cardiac study and VA follow-up data instead).

Caveat on this rating: Sponsor-developer-run trial. FDA's 2024 advisory committee voted 9-2 that effectiveness was not established, citing near-total functional unblinding and allegations of trial-conduct problems in the wider program; FDA issued a CRL in Aug 2024. The paper itself is not retracted, but three related phase 2 papers are.

Caveat on this rating: Sponsor-run trial by the company seeking FDA approval; benefit at week 3 faded for many by week 12 (sustained response was a minority outcome in the full paper).

Caveat on this rating: Conducted by the University of São Paulo CBD group (Zuardi, Crippa, and colleagues). The 2011 record carries no conflict-of-interest statement, but the same group's later publications disclose that several of these authors are co-inventors on a CBD patent that the university has licensed to pharmaceutical companies, so independence from a commercial interest in CBD cannot be assumed.

Caveat on this rating: Funded by Arjuna Natural Extracts Ltd, the manufacturer of the Shoden extract tested. The authors state the trial was independently managed by the site investigators and the lead author discloses prior funding from the same company for unrelated work, but the sponsor's commercial interest in a positive result is direct, and the study's own secondary psychological outcome (DASS-21) did not reach statistical significance.

Caveat on this rating: Treatment was delivered by Ambio Life Sciences, whose shareholders are coauthors and patent applicants; Stanford analysis was philanthropically funded and independent of the clinic. Uncontrolled design caps what this can show regardless.

Caveat on this rating: N=12 with 8 vs 4 allocation; MAPS-affiliated authorship; included for historical significance.

Caveat on this rating: The paper's formal declaration reads "Source of Support: Nil" and "Conflict of Interest: None", but its own text states that the high-concentration full-spectrum extract tested "is the KSM-66 Ashwagandha extract, provided by Ixoreal Biomed, Hyderabad, India" — in-kind supply of the investigational product by its manufacturer, which is industry support whatever the declaration says. Recorded here as INDUSTRY on that basis. The trial was also not preregistered and the reported p-values are implausibly uniform across every scale.

Caveat on this rating: The racemic-vs-esketamine difference rests on indirect comparison across trials with different designs and populations; one coauthor (Zarate/NIMH) is an inventor on ketamine-related patents. An erratum was issued in 2021 (checked: not a retraction).

Caveat on this rating: Funded by the Institute for Research on Cannabinoids, a nonprofit on whose board three of the authors sit (unpaid), and two authors disclose personal fees from numerous cannabis and cannabinoid companies. The result nonetheless cuts against those commercial interests, which weakens the usual direction-of-bias concern.

Caveat on this rating: The CONTESTED tier here reflects rock-bottom design score (uncontrolled, unregistered, self-report), not an active dispute; there is no controlled US evidence for ibogaine in opioid withdrawal at all.

Caveat on this rating: Funded by the American Kratom Association, a pro-kratom advocacy organization, and respondents were recruited through its channels — a design that samples satisfied current users and is structurally biased toward favorable results.

Caveat on this rating: The low score reflects study design, not disputed findings: this is a narrative review with no systematic search, no pooled estimate, and no readable funding statement, so it carries little independent evidentiary weight even though its cautious conclusion has held up.

Caveat on this rating: This was the single positive short-term pivotal trial; two sibling trials (TRANSFORM-1, TRANSFORM-3) missed their primary endpoints, which the approval package acknowledged. Sponsor employees authored the paper.

Caveat on this rating: Randomized-withdrawal designs can overstate maintenance benefit (abrupt discontinuation effects count as "relapse"); sponsor-funded and sponsor-authored.

Caveat on this rating: Many of the included trials were sponsored by sertraline's manufacturer; the review authors ran sensitivity analyses excluding manufacturer-funded studies.

Caveat on this rating: Publicly funded (NIHR and Japan Society for the Promotion of Science); some co-authors disclosed lecture or consultancy fees from drug manufacturers.

Caveat on this rating: This analysis covers trazodone at antidepressant doses for depression; the much more common low-dose use for insomnia is off-label and was not what was measured here. Several authors disclosed pharmaceutical lecture/consulting fees, though the study itself was publicly funded (NIHR and JSPS).

Caveat on this rating: Trazodone is the most commonly prescribed antidepressant for insomnia in the USA, but this use is off-label; the review found no evidence on long-term use and rated adverse-event evidence low certainty.

Caveat on this rating: Publicly funded (NIHR/JSPS), but several coauthors disclosed lecture or consultancy fees from drug manufacturers, including Pfizer.

Caveat on this rating: The trazodone evidence in this Cochrane review comes from one trial of 30 people over two weeks; effects on night-time awakenings and daytime sleep appeared negligible, and longer-term safety in dementia is untested.

Caveat on this rating: Funders were independent, but several co-authors disclose consulting or speaker fees from ADHD drug manufacturers; head-to-head evidence at 12 weeks says little about long-term outcomes.

Caveat on this rating: Publicly funded (NIHR and Japan Society for the Promotion of Science); some co-authors disclosed lecture or consultancy fees from drug manufacturers.

Caveat on this rating: Publicly funded (NIHR, JSPS) and PROSPERO-registered, but most of the underlying trials were industry-funded and several authors disclosed pharmaceutical-company fees.

Caveat on this rating: Later independent reanalyses of the broader STAR*D program have argued its published cumulative remission rates were inflated by protocol deviations; the level-2 switch comparison itself found no significant between-drug differences.

Caveat on this rating: One review author disclosed speaking honoraria from several drug manufacturers; the review itself had no external funding.

Caveat on this rating: No external funding, but several authors (Moncrieff, Horowitz, Kirsch, Hengartner) are prominent critics of antidepressants and disclosed royalties/consulting tied to antidepressant-deprescribing work — a perspective readers should weigh.

Caveat on this rating: Sleep-related adverse effects were about four times more frequent than with placebo (RR 4.31); funding and conflict-of-interest statements were behind the journal paywall and could not be read.

Caveat on this rating: Funding statement could not be retrieved (paywalled full text), so funding is recorded as unknown; superseded in scope by the same group's 2018 analysis.

Caveat on this rating: Includes unpublished regulatory data, so its estimates are less flattering than published-only analyses; one author disclosed industry speaker fees.

Caveat on this rating: Two reanalysis authors have served as paid expert witnesses against GSK in paroxetine litigation (disclosed); the original GSK-funded 2001 report claimed efficacy and remains unretracted but disputed.

Caveat on this rating: 41% of included trials were industry-funded, and this review series' very-low-certainty GRADE ratings are disputed by many ADHD researchers who consider them overly conservative relative to clinical consensus.

Caveat on this rating: Observational claims data; some co-authors disclose consulting fees or equity in health-analytics and pharmaceutical-adjacent companies, though the study itself was NIMH-funded.

Caveat on this rating: The review notes 6 of 20 included trials had pharmaceutical-industry funding and most were underpowered with inadequate randomization reporting.

Caveat on this rating: Small evidence base (7 trials) limited to short-term use; using trazodone for insomnia is off-label.

Caveat on this rating: Association, not proven causation; one senior author reports grants and personal fees from Takeda (an ADHD drug maker) outside the work. Contrasts with Habel 2011, which found no short-term rise in serious cardiovascular events.

Caveat on this rating: Co-funded in part by Angelini Pharma alongside public funders (NIHR Oxford Health BRC, MRC, Swiss NSF); registration status could not be read from the paywalled full text.

Caveat on this rating: Medication use after the 14-month randomized phase was self-selected, so height and symptom comparisons are observational; several senior authors disclose past consulting for stimulant manufacturers.

Caveat on this rating: All five underlying trials were conducted and funded by the manufacturer (GlaxoSmithKline); the statistical analysis was independent, and one co-author sat on the manufacturer's advisory board.

Caveat on this rating: Limited to published trials, so publication bias cannot be excluded; funding source not verifiable from fetched records.

Caveat on this rating: Funding statement could not be read (paywalled); this BOLDER I registration trial supported the manufacturer's bipolar-depression indication, so sponsor involvement is plausible but unverified.

Caveat on this rating: Funding statement is behind a paywall and could not be read; the two component maintenance trials are widely reported as manufacturer-sponsored, so weigh sponsorship accordingly.

Caveat on this rating: Government-funded, but several authors disclosed consulting or research ties to pharmaceutical companies; observational within-individual design.

Caveat on this rating: Observational studies only — people prescribed hypnotics differ from those who aren't, so confounding by insomnia itself and by frailty cannot be excluded; fall results were heterogeneous and not statistically significant.

Caveat on this rating: Included studies were small and observational; the same group's companion meta-analysis (PMID 15033227) found impairment improved but did not fully resolve after withdrawal.

Caveat on this rating: Observational study; despite propensity-score adjustment, people prescribed lithium may differ from those prescribed the comparison drugs.

Caveat on this rating: Open-label design and part-funded by Sanofi-Aventis (a valproate manufacturer), though the results favored lithium over valproate.

Caveat on this rating: Included evidence is mixed (RCT discontinuation phases, open trials, and case reports), so withdrawal incidence for venlafaxine cannot be precisely quantified; design scored conservatively as a review including observational data.

Caveat on this rating: Funded by Janssen-Cilag, which markets competing antipsychotics (paliperidone, risperidone), though the funder reportedly had no role in design or analysis; observational design cannot fully exclude confounding by indication.

Caveat on this rating: Included non-placebo-controlled studies (compared against a "virtual placebo" sample), total participant count and funding not verifiable from accessible text — scored conservatively; direction of the mirtazapine weight finding is consistent with the placebo-controlled Kamp 2025 meta-analysis.

Caveat on this rating: Publicly funded (NIHR and Japan Society for the Promotion of Science); some co-authors disclosed lecture or consultancy fees from drug manufacturers.

Caveat on this rating: Publicly funded (NIHR and Japan Society for the Promotion of Science); some co-authors disclosed lecture or consultancy fees from drug manufacturers.

Caveat on this rating: One review author disclosed speaking honoraria from several drug manufacturers; the review itself had no external funding.

Caveat on this rating: Most included trials were manufacturer-sponsored comparisons of escitalopram against rival drugs, a known source of sponsorship bias the review discusses.

Caveat on this rating: Review itself was publicly funded, but several authors declared lecture or consultancy fees from pharmaceutical companies, and most included trials were industry sponsored.

Caveat on this rating: Funders were independent, but several co-authors disclose consulting or speaker fees from ADHD drug manufacturers; head-to-head evidence at 12 weeks says little about long-term outcomes.

Caveat on this rating: Authors are FDA scientists analysing the agency's own trial database; no external funding was received.

Caveat on this rating: The reviewers note that 8 of the 11 included trials were funded by the same manufacturer, so industry sponsorship bias in the underlying trials cannot be ruled out.

Caveat on this rating: Analysis of EMA clinical study reports by Nordic Cochrane Centre authors who are prominent critics of industry trial reporting; stress urinary incontinence is not an FDA-approved duloxetine indication, and the authors concluded harms outweighed benefits for that use.

Caveat on this rating: Of 146 included studies reporting funding, 72 were funded by pharmaceutical companies; safety evidence was rated very low certainty.

Caveat on this rating: Government-funded (VA Cooperative Studies Program), but Bristol-Myers Squibb supplied the aripiprazole used in the comparison arm.

Caveat on this rating: Several authors reported grants from or ties to pharmaceutical companies; confidence intervals could not rule out a modest risk increase, healthy-user bias may flatter the results, and the later Zhang 2024 study found rising cardiovascular risk with more than 3 years of cumulative use.

Caveat on this rating: Observational despite target-trial-emulation design; one senior author reports grants and personal fees from Shire/Takeda outside the work; results cover ADHD medications as a class, mostly stimulants.

Caveat on this rating: Observational claims data; some co-authors disclose consulting fees or equity in health-analytics and pharmaceutical-adjacent companies, though the study itself was NIMH-funded.

Caveat on this rating: 16 of the 19 included trials were industry-funded and the review graded the evidence low to very low quality, so effect sizes may be optimistic.

Caveat on this rating: This is an EU regulatory action; the US FDA label does not impose the same 100 mg daily cap.

Caveat on this rating: NIH-funded, but Forest Laboratories (escitalopram's patent holder) supplied the study drug and placebo, and several authors disclosed industry relationships.

Caveat on this rating: All included studies were industry funded, and only four were not sponsored by duloxetine's manufacturer; the review authors cautioned that sponsorship bias may overestimate treatment effects.

Caveat on this rating: Eli Lilly, duloxetine's manufacturer, sponsored all but one of the included trials.

Caveat on this rating: The journal funding footnote sits behind JAMA's paywall, so funding is recorded as unknown; ClinicalTrials.gov (NCT00006286) lists the National Institute of Mental Health as the study sponsor.

Caveat on this rating: Funding statement could not be retrieved (paywalled full text), so funding is recorded as unknown; superseded in scope by the same group's 2018 analysis.

Caveat on this rating: Directly contradicts the dose-response signal in Billioti de Gage 2014; the dementia question is unsettled, and dependence/fall risks in older adults are established regardless.

Caveat on this rating: This trial tested buspirone as an add-on for depression, not for anxiety — an off-label use of the drug.

Caveat on this rating: Slightly elevated risk appeared at low but not high exposure - a pattern inconsistent with causation - yet the finding conflicts with the Billioti de Gage 2014 case-control study; the question remains unsettled.

Caveat on this rating: Contested: the prospective Gray 2016 BMJ cohort found no dose-response link between benzodiazepine use and dementia, so reverse causation (early dementia symptoms leading to prescriptions) remains a live alternative explanation.

Caveat on this rating: Full funding statement is paywalled and could not be read; the pooled trials were of the manufacturer's drug, so sponsorship could not be ruled out.

Caveat on this rating: Association, not proven causation - anxiety and insomnia can be early dementia symptoms that trigger benzodiazepine prescriptions years before diagnosis, and the prospective Gray 2016 cohort (BMJ) found no dose-response supporting a causal link.

Caveat on this rating: Funded by Pfizer and GlaxoSmithKline, the manufacturers of the drugs studied; several authors were Pfizer or GSK employees.

Caveat on this rating: Contemporary critiques in the same journal (e.g. PMID 2660772, 1989) disputed how durable the advantage over placebo was beyond the first weeks of this manufacturer-era trial and emphasized rebound and dependence on discontinuation.

Caveat on this rating: Funding and author industry relationships for this meta-analysis could not be read from open sources; several included trials were manufacturer-run.

Caveat on this rating: Pre-registration-era trial; the funding source could not be read from available records.

Caveat on this rating: Pre-registration-era trial; the funding source could not be read from available records.

Caveat on this rating: Small single-site pilot RCT from 1993; effect size is large but the study predates modern trial registration and reporting standards.

Caveat on this rating: Open-label design — patients and clinicians knew which drug was taken, which can inflate between-drug differences.

Caveat on this rating: Several authors disclosed speaker fees and grants from pharmaceutical companies (Janssen, Lundbeck, Otsuka, Angelini, and others).

Caveat on this rating: One author holds a position at Lund University funded by AB Svenska Spel, the state-owned Swedish gambling operator, as part of its responsible-gambling policy, and reports research funding from the same source alongside public research councils. The conflict is with the outcome being studied rather than with the drug, and it is disclosed in full in the paper. It is repeated here because a reader deciding how much weight to give a gambling finding should know who funds the chair that produced it.

Caveat on this rating: Study funding is public, but the author group carries extensive pharmaceutical relationships, all declared as unrelated to this work: honoraria from Lundbeck and Otsuka, consulting fees from Bristol-Myers Squibb, Novartis, Roche, Merck and Pfizer, stock options in several device and pharmaceutical companies, and participation in regulator-mandated phase IV studies funded by Alcon, Almirall, Astellas, AstraZeneca, Boehringer-Ingelheim, Novo Nordisk, Servier and LEO Pharma with funds paid to the institution. None of those are the makers of quetiapine, and the finding runs against the drug rather than for it, which is the direction sponsorship bias does not usually run. Recorded here so a reader can weigh it rather than discover it.