In plain terms

The claim being assessed: Changing human gut microbial composition changes psychiatric symptom severity through gut-to-brain signalling.

Where the evidence stands: Γ — Plausible mechanism. There is a believable biological story and partial support, but the causal step has not been demonstrated.

What this mechanism does: This makes the process worse or keeps it going, but something else starts it.

Strongest evidence for it: Rodent gut-brain causality reaches Ε across two recipient systems and some human depressive-treatment effects reach Δ. A human composition-dependent signalling mediator has not been established, the one human contrast that demonstrably changed the community relative to its comparator produced no extra symptom benefit, and bipolar/schizophrenia generalization is unsupported; the exact conjunction stays Γ.

What decisive evidence would look like

In plain terms, researchers would change the proposed cause, confirm it changed as intended, and measure a meaningful health outcome while checking other explanations. The result should be able to support or count against the claim.

Relative explanatory fit: Alternative ranked ahead

What this means for mental health and the gut

The gut and brain communicate through nerves, immune signals, hormones, food-derived chemicals, and many other routes. That makes it reasonable to study microbes in depression and other psychiatric conditions. It does not yet show that changing a person's gut community is a reliable way to relieve psychiatric symptoms.

Implications for health and treatment: The practical implication is research, not a proven microbiome treatment for mental health. A probiotic, diet, or stool-based intervention may change gut measurements without improving symptoms, and an improvement after treatment does not identify the gut change as its cause. Established assessment and mental-health care should not be displaced by commercial microbiome testing or unproven microbial treatments.

Limits for individual care: Psychiatric symptoms can be serious and have many possible contributors. The studies reviewed do not establish which microbial change, if any, would help a particular diagnosis or person, nor do they establish a safe do-it-yourself route to changing the microbiome. Any treatment change needs discussion with the clinician managing that care.

Medical information notice: This review provides general health information, not medical advice. It cannot diagnose a condition or replace an individual discussion with a qualified clinician.

This assessment rests on 23 sources and 23 individually checked claims. 23 of those claims were verified against the original paper.

A low grade is not an accusation. It means the causal work has not been done yet, which is the ordinary state of most actively studied mechanisms in biology.

At a glance

Sub-claims assessed separately

ID	Sub-claim	Type	Level	Pivotal card
SC1	Human gut community composition correlates reproducibly with depressive symptom states.	existence	Β	—
SC2	Microbial transfer can alter affect-like behavior through gut-to-brain pathways in rodents.	mechanism	Ε	Card ↗
SC3	Microbe-directed treatment changes depressive symptom ratings in some adult clinical settings.	sufficiency	Δ	Card ↗
SC4	A changed human gut community alters psychiatric symptoms through a specified gut-to-brain signalling mediator.	mechanism	Γ	—
SC5	The composition-dependent gut-to-brain symptom effect applies across human psychiatric conditions.	generalization	Γ	—

Visual Evidence Cards · Pivotal Subclaims

Pivotal subclaim illustrations

Dark-theme simulated-micrograph renders and concise causal evidence summaries for the pivotal subclaims that govern this review's grade. Click any card to inspect the master 2400×2400 render.

Master 2400×2400 render ↗

Master 2400×2400 render ↗

Figures

Explanatory figures summarizing this review. They carry no evidential weight of their own; where a figure and the review differ, the review governs.

Full assessment

From here the review is written for a technical reader: caps and thresholds, competing models, the independence basis, and the citation ledger.

In plain terms

The claim says that changing the mix of gut microbes can change psychiatric symptoms in people by sending a signal from the gut to the brain.

The full claim receives Γ: a plausible mechanism with partial support. Microbes can change mood-like behavior through a nerve pathway in mice, and some human depression trials report symptom improvement after probiotic or fecal-transfer treatment. Other well-controlled trials show no added symptom benefit, even when gut measures change. In the sharpest test so far, transplanted stool shifted the gut community toward the donor's and symptoms still improved no more than in the comparison group. No reviewed human study proves that a measured community change caused symptom relief through a particular gut-to-brain signal.

A grade of Γ is a statement about how much causal work has been done, not an accusation against the researchers doing it, and not a verdict that gut microbes are irrelevant to mental health. It says the deciding human experiment — change the gut community, block the signal it is supposed to send, and see whether the symptom effect disappears — has not been run.

AT A GLANCE

Controlled microbial treatments sometimes change depressive symptom scores, but the one human trial that demonstrably shifted the recipient community relative to its comparator produced no extra symptom benefit, and no human experiment interrupts a gut-to-brain signal.

• The experiment that decides it: A blinded, adequately powered human trial must change a predeclared microbial function, independently interrupt its candidate gut-to-brain signal, and show that the clinical symptom effect disappears and returns with functional rescue.
• Belief versus evidence: The field's own syntheses of these trials report a large pooled benefit and recommend treating the gut microbiome alongside medication, which implies Ε-level human context-specific causality; the reviewed complete human chain reaches Γ, two rungs lower.
• Best-supported rival: A context-bounded microbial modifier of depressive treatment response (Δ), without one established pathway across diagnoses.

Primary rubric: 04-microbiome-systemic.md · Secondary rubric: 07-psychiatry-neuromodulation.md · HSEF v5.4 · 2026-09-27 · Revision (red-team synthesis) · Previous level: Γ (Codex grading, 2026-09-27).

Provenance header

Sources: 23 primary reports (21 Tier A, 2 Tier B) · Load-bearing claims: 23 · Verified against each report at recorded access depth: 23 · Unverified: 0 · Contradicted: 0 · Structural gate: passed (companion validator record).

Adversarial review: Generated by Sol 6.1; adversarially reviewed by Claude Opus 5.5 (Anthropic) on 2026-10-14. 3 corrections applied (see ADVERSARIAL REVIEW AND APPROVAL). Status: Approved as revised.

Scorecard detail and subclaim matrix

Aggregation rule and derivation. The user's sentence requires a human effect on psychiatric symptoms and attribution to gut-to-brain signalling. SC4 is the subclaim whose scope is the central claim's scope, and it is Γ; SC5 at the broader diagnostic scope is also Γ. SC3 reaches Δ at a narrower scope that drops the signalling clause, and SC2 reaches Ε in rodents. Scope-matched therefore returns Γ.

The aggregation dissent, recorded rather than resolved away. The index prescribes weakest-link for this hypothesis, and weakest-link across all five subclaims returns Β, because SC1 is association-only and universal cap 2 binds it. That is a live reading and a reader who prefers the index rule should have it. Two things argue against adopting it here. First, SC1 is not a conjunct the graded sentence needs: a composition change could move symptoms with no detectable cross-sectional association, so letting an observational premise fix the letter is pitfall 21 — a universal cap re-applied through the weakest class of a set assembled to cover each other's weaknesses. Second, weakest-link restricted to the conjuncts the sentence does require, SC3 and SC4, returns min(Δ, Γ) = Γ, the same letter scope-matched returns. The two-rung swing comes from which subclaims enter the minimum, not from the rule's name. The departure is defended on that analysis and not on the letter it produces; the index rule's Β is carried here so it can be argued against specifically.

Claim restatement and rubric routing

User wording: “Gut microbial composition causally influences human psychiatric symptoms through gut–brain signalling.”

Indexed topic: “Gut-brain axis in behavior/psychiatry.” The packet's indexed nominal model is stronger: dysbiosis as an upstream cause of major depression, anxiety, bipolar disorder and schizophrenia, with correction as disease-modifying. That initiation-and-treatment breadth is not added to the user's claim.

Graded causal sentence: Changing human gut microbial composition changes psychiatric symptom severity through gut-to-brain signalling.

The statement has two necessary causal links: community → clinical symptoms, and a specified signal carrying that effect. It covers symptom maintenance or response, not first onset of a psychiatric diagnosis. “Human” is known at enrollment; symptom scale and microbial contrast can be predeclared; the pathway must be chosen for measurement before outcomes, rather than named after a favorable association. The microbiome-systemic rubric directly covers behavior and psychiatric phenotypes. The indexed note says defined-consortium rescue is Δ-qualifying for SC2, while the governing rubric places first phenotype-moving transfer at Δ and defined rescue with replication at Ζ. I apply the rubric to the mouse result and retain this discrepancy in the routing record.

Secondary rubric, added in this pass. The index lists 01 or 05 as possible secondaries and neither fits. What does fit is 07-psychiatry-neuromodulation: SC3 and SC4 are decided on human psychiatric symptom rating scales, and the comparator problem that rubric governs — expectancy, unblinded raters, open-label design, mediators that are correlated and never manipulated — is live in every trial in this ledger. Under Rule 6 the lowest applicable ceiling binds, and the routing is recorded rather than assumed. It changes no letter. Its functional-unblinding hard rule does not fire, because capsule probiotics and encapsulated or enema-delivered transplants have no unmistakable acute subjective effect and the two trials that measured masking found it intact [10,15]; its open-label rule fires on the one unblinded trial [17] in isolation and is recorded there; its mediation-by-correlation rule fires on SC4 and agrees with the 04 ceiling already binding it.

HYPOTHESIS STATE EVALUATION FRAMEWORK (HSEF v5.4)

Competing mechanistic models

M1. Positive randomized depressive symptom contrasts after probiotics or FMT [5,6,9,14,15] mean that the microbial-treatment story cannot be collapsed into cross-sectional association. Yet Romijn, Green and PRO-DEMET found null primary or mood contrasts [7,10,11]; formulations, clinical populations and target engagement differ. Δ is the first human phenotype-perturbation threshold, without an Ε boundary mapped for one stable intervention.

M2 and M4. The human cohort signals [1,2,19] cannot determine whether microbiota are cause, consequence or both, and the instrument analysis inside the largest of them was significant with depression as the exposure [22] — the direction M2 predicts. Siopi experimentally made stress-altered donor microbiota in mice [4], showing an upstream host-state route to a changed community that can later be transferable. Kelly's depressed human cohort was largely medicated [3], and diet and antidepressant exposure remain credible ecological confounders. M4 is nonetheless weaker than it looks: the causal-discovery cohort found its two nominated genera differed from controls in unmedicated as well as medicated patients [2], and the population cohort corrected for antidepressant use and kept its depleted butyrate producers [19]. These rivals do not explain away a blinded positive intervention merely by being plausible.

M3. Rodent transfer and vagotomy [3,4,18] give a route by which gut signals can affect brain-related behavior. Pinto-Sanchez saw fMRI and depression-score changes without a detected fecal community difference [5]; Schaub found Lactobacillus enrichment with its HAM-D interaction significant only in a compliant subset [8]; Mörkl found autonomic shifts without a depression benefit [13]. Li's randomized FMT trial reports donor strain acquisition, a positive secondary symptom contrast and a bile-acid mediation model [15]. Those observations form an important hypothesis for the human route, but none selectively perturbs the human mediator, and the single trial that changed the community relative to its comparator on purpose found no symptom advantage [16]. Its broad form also faces null bipolar and schizophrenia symptom trials [12,16]. M3 therefore remains Γ.

M5 and tournament result. Both allogenic and autologous FMT arms improved in bipolar depression without an added donor effect [16], and the unshammed Wang trial lost more people from its FMT arm [17]. The two 2025 syntheses pool this literature into a large favourable effect while reporting heterogeneity close to total [22,23], which is as consistent with small-study and reporting effects as with a real common effect. Those facts give nonspecific treatment explanations real standing. Li's blinded contrast [15] is still the sharpest separator of a bounded microbial treatment component from a wholly host-first or expectancy account — masking was measured and intact — but it is a weaker separator than its abstract suggests: the prespecified primary endpoint was null and the positive HAMD-17 difference is a secondary outcome. The nominal model ranks third. Tournament rank informs role; it does not alter the HSEF letter.

Evidence inventory and causal interpretation

Human associations: real signals, unresolved direction

Radjabzadeh et al. replicated 12 depression-associated genera across the Rotterdam and HELIUS cohorts after excluding antidepressant users and applying measured covariates [1]. Some overall community measures did not replicate — beta diversity was associated in the discovery cohort and not in the replication cohort — and the observational design cannot order the gut and mood. The same report ran a two-sample Mendelian randomization, and this is the part of it that bears hardest on the nominal model: with major depression as the exposure the inverse-variance-weighted effect on Eggerthella was significant, while the microbiome-as-exposure analysis returned an estimate inconsistent with the cohort association, and the authors state plainly that the MR lacked power because instruments had to be taken below genome-wide significance [22]. Valles-Colomer et al. found Coprococcus and Dialister depleted in depression in a 1,054-person population cohort with validation in 1,070 further individuals, after correcting for antidepressant use [19]; that is the strongest version of the association and it remains association. Fehse et al. applied a graphical causal-discovery algorithm to 1,269 clinically characterized people and inferred Eggerthella and Hungatella as probable contributors, with the difference from controls holding in unmedicated patients [2]. The primary report explicitly states that observational data still cannot definitively establish causality. Neither algorithmic direction nor a named taxon is a human gut–brain perturbation — and two human analyses of the same genus, Eggerthella, point in opposite directions [2,22].

What animal perturbations establish

Kelly et al. moved stool pooled from three of the most severely depressed male patients or three matched controls into 28 antibiotic-depleted rats [3]. The depressed-donor recipients showed reduced sucrose preference and anxiety-like behavior; their forced-swim readouts did not differ. Most depressed participants in the human cohort were taking antidepressants; the report does not specify that all three selected donors were medicated. This is a phenotype-moving transfer in recipients, not proof that the donors' psychiatric symptoms were microbe-driven. Zheng et al. did the complementary experiment in a different laboratory and a different recipient system: germ-free mice colonized with patient microbiota developed depression-like behaviour relative to healthy-donor colonization [18]. Two recipient preparations rather than one matters for the scope of SC2 — the effect is not an artifact of antibiotic conditioning alone — and it matters for what cap 5 governs, since both are forced colonizations. Siopi et al. transferred stress-altered mouse microbiota into antibiotic-prepared male mice; recipient affect-like behavior, hippocampal neurogenesis and neuroinflammation moved, and subdiaphragmatic vagotomy prevented all three [4]. That is a strong, bounded neural-pathway test in mice, replicated within the originating laboratory rather than outside it. Vagotomy itself alters gut physiology and is not selective, as those authors note, and no corresponding human pathway interruption was performed.

Human symptom interventions: positive findings and what they measure

In an IBS pilot, Bifidobacterium longum NCC3001 improved a two-point depression-scale response in 14/22 versus 7/22 people at week six (P=.04) and changed brain responses to negative emotional stimuli [5]. Fecal microbial profiles were similar between arms, however; the trial supports a possible bacterial-treatment effect in IBS, not a demonstrated shift in community composition. Kazemi et al. reported greater BDI improvement with probiotics than placebo or prebiotic after eight weeks among 81/110 completers (P=.042); no open-access full text was reachable on 2026-09-27, and the accessible abstract does not establish a compositional or intention-to-treat mechanism [6].

Schaub et al. enrolled 60 depressed inpatients, but 47 completed treatment. Their HAM-D time-by-arm interaction was nonsignificant in the reported ITT set and significant in the compliant subset; Lactobacillus abundance increased [8]. A within-person taxon–score association does not mediate a randomized treatment effect. In Nikolova's 49-person pilot, week-eight self-reported depression and clinician-rated anxiety favored probiotic adjunctive therapy, while clinician-rated depression and GAD-7 did not show clear week-eight effects; no fecal readout is in the main report [9]. A separate 50-person fluoxetine add-on study found a larger depression-score decrease in the probiotic arm among 44 completers but did not profile the gut; measured cortisol, ACTH and BDNF did not differ [14].

Li et al. provide the most relevant randomized FMT challenge, and the full text changes how it reads [15]. Sixty adults on escitalopram were randomized 1:1 to donor material or placebo, 56 completed, and masking was measured and intact. The prespecified primary endpoint was null: week-8 remission 43.3% versus 26.7%, P=0.1760. The secondary HAMD-17 contrast favoured FMT at week 8 (between-group difference −2.9 points, 95% CI −5.5 to −0.3; P=0.0303), a confidence interval whose nearer bound sits close to no difference. Donor engraftment was partial and declining — about 16% of species newly acquired at week 2, about 7% at week 8, with coexistence rather than replacement the dominant pattern — so the abstract's description of engraftment as durable is not what the engraftment analysis shows. The mediation model is observational and runs from a donor-enriched taxon through glycocholic acid to an inflammatory marker, with no bile acid manipulated. FMT changes more than a named microbe, and postrandomization associations among bile acids, inflammation and symptoms do not establish a unique causal mediator. Wang et al. found an additional two-week HAMD-24 reduction after FMT plus medication versus medication alone [17], but the unshammed comparison analyzed only 14/23 assigned FMT recipients versus 20/23 controls, profiled only ten selected bacterial targets, found only Enterococcus differing by discriminant analysis, and reported that post-treatment Enterococcus correlated positively with depression scores. It is supportive at most, not a clean mechanistic replication.

Human genetic instruments and population-scale syntheses

Two-sample Mendelian randomization is the framework's field-valid equivalent where deliberate perturbation is unavailable, and it exists here. Chen et al. report protective effects of Actinobacteria and Bifidobacterium on major depression and no reverse effect of depression on composition [20]; Zhuang et al. report a suggestive effect of genetically increased Bacilli on major depression across a three-disorder analysis, and disorders altering composition in the reverse direction [21]. Neither carries the claim. The instruments are taxon-level and selected below genome-wide significance, the nominated taxa do not overlap between the two analyses, one study's own Egger sensitivity does not support part of its result, the reverse direction is significant in two of the three instrument analyses available [1,21], and none of them touches the signalling clause. Alongside them, the field's syntheses of the intervention literature pool a large favourable probiotic effect on depressive symptoms with between-study heterogeneity close to total and recommend adjunctive microbiome-directed treatment in practice [22,23]. Those are Tier B and no figure from them enters this review; they are recorded because the divergence artifact must be read from how the field presents the claim, and because a pooled effect larger than any blinded prespecified primary result in this ledger is a fact about the synthesis rather than about the mechanism.

Human controlled counterevidence and the pathway gap

Romijn et al.'s 79-person double-blind trial found no psychological-outcome advantage for its two-strain product in people selected for low mood [7]. The 15-person Moving Moods sham-controlled FMT pilot established feasibility with successful masking and showed no MADRS or DASS advantage — week-8 MADRS was 16.6 in both arms [10]. PRO-DEMET randomized 116 people and found no difference on its primary MADRS change (P=.45), despite a positive secondary minimal-important-change threshold [11]. A secondary responder threshold cannot supersede the prespecified primary comparison.

The 131-person bipolar/schizophrenia-spectrum probiotic trial found no improvement in psychiatric symptom severity, while gut-barrier markers including zonulin changed; one cognitive domain improved borderline [12]. Mörkl et al. measured autonomic proxies in depressed people and controls [13]. Its published abstract reports improved morning vagal function in probiotic-treated depressed patients compared with healthy controls, which is a between-group comparison that confounds the intervention with the diagnosis and cannot by itself establish a treatment effect. The randomised within-patient contrast does exist in the body: among depressed patients at three months, morning RMSSD was higher on probiotic than on placebo (U = 241.00, Z = 3.03, P = .002) as was morning lnHF (t(33) = −3.48, P = .001, d = −1.19), on a trial whose prespecified primary endpoint was change in vagal function by heart-rate variability, with depressive symptoms a secondary endpoint. Those secondary symptom scores improved over time in every group with no probiotic advantage, and diversity “was not changed significantly by probiotic intake or over time”. The dissociation this review relies on therefore holds on the randomised contrast rather than on the abstract's healthy-control comparison. Heart-rate variability is influenced by multiple autonomic inputs and is not a selective vagal-mediation assay; a probiotic that moves a vagal proxy without moving symptoms is a dissociation, not a mediator. In bipolar depression, allogenic donor FMT changed recipient–donor community similarity and raised diversity at six months while autologous FMT did not, yet did not outperform autologous FMT on MADRS: the baseline-to-week-24 improvement was 16.74 points with allogenic and 15.4 points with autologous FMT (P=.64, 95% CI −7.3 to 4.6) [16]. Autologous FMT may be biologically active, so the study restricts the inference without excluding all microbial treatment effects — but it is the closest thing in this ledger to an internal control on the composition clause itself, and it is negative.

Pivotal perturbation audit. Rodent phenotype/pathway positives: [3,4,18]. Human depressive-treatment signals: [5,6,14,15], with important analytic limits; partial Schaub/Nikolova signals: [8,9]. Human null or primary-null contrasts: [7,10–13,15,16]. Of the four randomized human FMT trials in the ledger [10,15,16,17], two are null on their symptom comparisons, one is positive but unblinded with nine of 23 assigned participants not analysed in the transplant arm, and one is blinded with a null prespecified primary and a positive secondary. The audited number of studies showing the complete sequence measured human community change → independently tested gut–brain mediator → psychiatric symptom effect is zero. That is not a count of all causal microbial findings.

Cap rules, precedence, and Β/Γ/Δ boundary

SC1 — association. The universal rule says, “Pure association evidence — however large, however replicated — cannot exceed Β.” Replicated genera [1], population-scale depletion of butyrate producers [19] and graphical causal inference [2] are still observational. SC1 is Β. Its first missing Δ test is an intervention that moves both the candidate community feature and a phenotype; the taxon association itself cannot be promoted by sample size, and the genetic instruments that might have substituted for that intervention are too weak and too inconsistent to do so [20,21,22].

SC2 — recipient animals. The rubric defines a field-valid first perturbation as a “Gnotobiotic or transfer experiment showing phenotype movement, at minimum.” Kelly, Zheng and Siopi meet it in the actual rodents [3,4,18]. Siopi adds a vagal intervention and maps where the phenotype stops; the bounded rodent result reaches Ε, now across both antibiotic-depleted and germ-free recipients. It does not reach Ζ because the rubric asks for a defined consortium or strain, rescue logic, and independent replication of the same pathway: no defined consortium has been shown to recapitulate the crude-community effect for this phenotype, and the vagotomy result was reproduced within its originating laboratory rather than outside it. The universal forced-transfer cap applies to extrapolation from those recipients to endogenous human symptom production, not to their rodent result.

SC3 — human treatment response. Human randomized intervention is the field-valid analogue of the Δ perturbation, and under the secondary rubric it must also carry a credible comparator and a pre-registered primary outcome. Both are satisfied by the blinded trials whose prespecified symptom outcomes were positive [5,6,14], and Li's masked transfer trial adds a positive secondary contrast [15]. They do not establish one defined composition as the active agent; null primary results [7,10,11,15] and analytic limitations [8,9,17] keep the same-product context boundary unmapped. SC3 is Δ, not Ε. Two things are recorded here that the subclaim's own wording hides. First, boundary-check clause 5 is applied explicitly: the trials with null prespecified primaries do not cross Δ and are not counted as though the existence of the right design were the threshold. Second, SC3 is a treatment-response proposition, and a product can satisfy it without changing the community at all — fecal profiles were similar between arms in the positive IBS trial [5], overall community structure did not shift in the vagal trial [13], and the positive unblinded transplant trial reports no change in overall community structure with one genus moving [17]. That is pitfall 13, therapeutic effect misread as mechanistic proof, and it is why SC3's Δ cannot be imported into the graded conjunction.

SC4 — human signalling mechanism. The rubric's Γ ceiling for host-pathway plausibility binds, and the secondary rubric's mediation-by-correlation rule — “Candidate mediator correlated with outcome, never manipulated” — binds identically. Pinto fMRI [5], Schaub taxon associations [8], Mörkl heart-rate variability [13], and Li's bile-acid/inflammatory mediation model [15] are mechanistically useful. None independently alters a candidate pathway in humans to show that a community-induced symptom change depends on it. Positive treatment does not automatically validate an associated mediator, and the composition clause has now been tested directly once and failed: the community moved toward the donor in the allogenic arm and not the autologous arm, and the symptom trajectories were the same [16]. SC4 stays Γ.

SC5 — across diagnoses. The universal hard rule, “Biomarker change without functional/phenotypic change caps at Γ,” describes what the BD/SSD gut markers [12] and bipolar donor-similarity shift [16] cannot prove. Their psychiatric endpoints were null. The cross-diagnosis instrument analysis is nominal and bidirectional [21]. A different formulation might work, but the reviewed evidence does not demonstrate one mechanism across the claimed human psychiatric scope. SC5 remains Γ.

Cap precedence. Rodent transfer is Δ-qualifying and its mapped pathway reaches Ε within the recipient system. The universal rule “Seeding or forced-transfer paradigms cap at Γ for endogenous propagation claims” governs importing that artificial transfer — in germ-free or antibiotic-cleared recipients alike — into a claim about spontaneous human psychiatric symptom causation. Human randomized treatments [5–17] are assessed separately; their clinical contrasts are not artificially capped because they occur in humans, but their mediator attribution remains unproven. A second precedence question arises from the added secondary rubric: 04 makes a human transfer paradigm threshold-crossing while 07 additionally requires a credible comparator and a pre-registered primary outcome. Under Rule 6 the lower ceiling binds, and there is no live conflict — SC3 satisfies both on the blinded trials with positive prespecified outcomes, and SC4 fails both rubrics' mediator requirements in the same way.

Thresholds, generalization, and falsification

The first unmet threshold for SC2 is Ζ: a specified microbial function must be subtracted, restored, and reproduced with the same behavioral/brain endpoint in independent settings. SC3 first fails Ε: the human clinical result needs replication with one engaged formulation and a mapped responder or nonresponder boundary, and mapping requires the same product — the present nulls use different products in different populations, so the boundary is unmapped rather than mapped (pitfall 16). SC4 and SC5 first fail Δ at their respective scopes: selective human pathway dependence and a target-engaged symptom effect outside depression. The boundary checks do not turn a negative experiment of the right design into a positive Δ crossing.

The animal context boundary is substantive: Kelly's forced-swim result was null [3], and Siopi's vagotomy prevented transfer effects [4]. The human context boundary is less precisely mapped because distinct strains, donor preparations, psychiatric populations and concomitant medications were used. Null findings are not evidence that no microbiota–brain route can operate; they are evidence against treating every gut marker shift or probiotic exposure as an established symptom mechanism.

Falsification patterns:

• Community moves as intended and the phenotype does not: Shekarriz's allogenic arm shifted toward the donor and gained diversity while the autologous arm did not, and the MADRS trajectories were the same [16]. This is the ledger's most direct test of the composition clause and it is negative.
• Gut marker or donor profile moves, psychiatric score does not: Borkent's gut-barrier markers with null BPRS [12] and Mörkl's autonomic changes with no depression-score advantage [13] constrain the broad claim.
• Clinical benefit fails in controlled settings: Romijn's psychological outcomes [7], Green's MDD FMT pilot [10] and PRO-DEMET's primary MADRS comparison [11] counter simple class-wide efficacy; their size or formulation cannot disprove all future products.
• Phenotype moves without the claimed community change: depression scores improved with fecal profiles similar between arms [5]; symptom scores improved after transplant with no change in overall community structure [17]; a probiotic moved vagal parameters with no community shift [13]. A benefit that does not require a measured composition change does not test the composition-dependent claim.
• A proposed pathway is not independently interrupted in humans: Li's statistical mediation [15] is not the same intervention as Siopi's vagotomy [4].

The rubric pitfalls appear directly here: “dysbiosis” treated as mechanism in cohort results [1,2,19]; abundance or engraftment treated as causal mediation [8,15]; crude FMT treated as strain-specific evidence [15,17]; and a biomarker shift treated as a clinical effect [12,13]. No low-biomass brain microbiome claim is being made, so the contamination-specific Α rule is not activated.

The steelman and upper-ladder check

Best case for Δ on the exact sentence, written first. A blinded randomized human transfer trial with masking measured and intact reports greater HAMD-17 reduction at weeks two and eight, donor strain acquisition, and a mediation model running from a donor-enriched taxon through bile acids to an inflammatory marker [15]; a second transfer trial reports a larger two-week HAMD-24 reduction [17]; blinded probiotic trials report positive prespecified symptom outcomes [5,6,14]; vagotomy shows the route is neural in mice, across two recipient systems in two laboratories [3,4,18]; and genetic instruments point the same way for some taxa [20,21]. On that reading the human phenotype perturbation has been made and the mediator has been identified. Why it does not raise the full sentence: the blocker is not thinness of evidence but a specific negative result plus a specific missing experiment. The one trial that demonstrably changed the recipient community relative to its comparator produced no additional symptom improvement [16], and no human study has ever interrupted a candidate gut-to-brain mediator while leaving the microbial intervention in place. Li's prespecified primary was null and its engraftment declined; Pinto did not show a fecal community shift; the animal vagus test cannot be substituted for a human one.

Best case for Β. Most human taxon links are observational [1,2,19], the prespecified primary outcomes of the largest and best-controlled trials are null [7,10,11,12,15,16], several positive probiotic trials did not measure community composition [6,9,14], the syntheses that pool them report heterogeneity close to total [22,23], and the instrument analysis inside the largest cohort was significant with depression as the exposure [22]. SC5 could be argued at Β on the same basis without changing the aggregate, since scope-matched lands on SC4. Why Γ remains justified: phenotype-moving rodent transfer and pathway interruption [3,4,18] are more than correlation, and blinded randomized human clinical symptom contrasts [5,15] make a microbial treatment component plausible. The Γ grade preserves those narrow results without converting them into a verified human mediated chain, and letting an association-only premise fix the letter is the error pitfall 21 names.

Upper-ladder equivalents considered. Mendelian randomization, strain-resolved rescue, therapeutic randomized symptom endpoints, natural experiments, cross-population functional dose-response, and withdrawal/rechallenge were all considered. Genetic instruments exist and are nominally positive in places [20,21] but are taxon-level, sub-genome-wide, mutually non-overlapping, bidirectional and silent on the signalling clause. Siopi's vagotomy is a genuine pathway interruption in mice, not a named microbial-function rescue in humans [4]. Randomized human symptom trials are substantial evidence, but none isolates the composition-to-mediator-to-symptom sequence, and several prespecified clinical comparisons are null [7,10–13,15,16]. The cohort replication [1,19] is taxonomic association, not functional dose-response; no selective human natural experiment or withdrawal/rechallenge of one microbial function was found. These lines converge on possible contribution. No single forced-transfer cap has been reapplied to the entire convergent set: the level rests on the absence of a positive mediated human test and on one negative composition contrast, not on capping the weakest leg.

Role, momentum, stability, and divergence

Role — reinforcing / stabilizing factor, step 5. Clinical symptoms can improve in both microbial and control arms [10,11,16], and no microbial configuration has been shown necessary, initiating or rate-limiting for human psychiatric symptoms, so steps 3 and 4 fail. Step 2, epiphenomenon, was considered and rejected: randomized microbial interventions do move symptom scores in some settings [5,6,14,15], which is more than co-occurrence. Some microbial interventions may modify response alongside standard treatment. That is a bounded reinforcing role. The rival-model tournament did not determine this classification.

Momentum — Mechanistic only. Recent FMT, probiotic and autonomic trials [13–17] and Siopi's pathway work [4] show active causal research that is getting better controlled — masking is now measured, autologous comparators are being used. I did not identify a human trial independently altering a specified mediator after verifying microbial target engagement with a prespecified psychiatric symptom endpoint. New clinical intervention output is substantial; the decisive human pathway test remains unattempted in this reviewed set.

Stability — Stable. No pending, specifically identified human mediator-interruption result was found. A replicated, fully audited positive FMT depression trial with a prespecified primary endpoint could strengthen SC3; a blinded selective mediator trial with verified engagement could lift SC4; adequately engaged repeated nulls could lower the exact claim. “Stable” records the absence of an identified near-term trigger for the complete sentence, not certainty about future biology.

Divergence — Substantial overstatement for the complete wording. The implied level is read from how the field presents the claim, not from the claim's own wording. Two 2025 systematic reviews of these trials report a large favourable pooled effect of probiotics on depressive symptoms and recommend that the gut microbiome be monitored and treated alongside pharmacological treatment [22,23], and the primary reports in this ledger are titled and abstracted in causal language — "induces", "require vagus nerve integrity", "causal contribution". A competent reader of that literature would infer Ε, established context-specific human causality. The complete route is Γ: Ε − Γ = +2 rungs. A reader who takes the syntheses' clinical recommendation as implying converged causal lines would name Ζ and get +3; Ε is the conservative reading and the one adopted. Momentum, stability and divergence do not modify the HSEF letter or confidence.

Final adjudication

HSEF Γ, moderate confidence; reinforcing / stabilizing factor. Human microbial interventions plausibly affect depressive symptom response in some settings (SC3 Δ), and experimentally interrupted rodent pathways are causal in their models across two recipient systems (SC2 Ε). The required human community-to-signal-to-symptom chain (SC4) and its extension to psychiatric diagnoses beyond depression (SC5) each remain Γ; the scope-matched conjunction is Γ.

The missing positive experiment is a target-engaged human intervention that changes a prespecified microbial function, independently tests the proposed neural, immune or metabolic mediator, and changes a prespecified symptom outcome. A larger probiotic trial without that pathway contrast could move the treatment-response subclaim but would not alone establish the user's “through signalling” clause. Equally, nulls for one product do not rule out all microbial routes — and the one trial that changed the community on purpose and measured symptoms carefully is the result the next trial has to beat.

Proposed decisive experiments

1. Defined-function, mediator-contrast depression trial. Randomize adults with MDD on stable care to a strain-resolved microbial intervention or matched control, then independently alter the prespecified microbial output with a safe, biologically selective contrast. Measure absolute engagement, mediator time course, clinician-rated symptoms and relevant brain/autonomic readouts. Benefit that disappears when the microbial function is disabled and returns on restoration would support a causal signal; durable engagement and mediator change without symptom benefit would challenge it.
2. Cross-diagnosis and reverse-direction trial. Separately power bipolar-depression and schizophrenia-spectrum cohorts for the same engaged microbial function and diagnosis-specific symptom scale, with matched sham procedures. Measure stress, diet, drugs and transit prospectively, and include a predeclared host-state contrast. Failure of symptom movement after verified engagement would restrict generalization; replicated benefit with the same mediator despite host-state controls would support it.

Knowledge gaps and reviewer notes

The key unknowns are the specific microbial function, whether its product reaches and changes a relevant human signalling compartment, whether that signal is necessary for a clinical symptom effect, and which diagnoses or baseline communities permit target engagement. Reverse causation from stress, psychiatric state, diet and medication remains viable for cohort signatures, and one instrument analysis inside the largest cohort actively favours it. A stool diversity change is not a brain signal, and a brain or blood signal is not evidence of causal mediation merely because it changes in the same arm.

The supplied v5.4 packet supplies rubric, index and scorecard structure; its embedded directions were treated as document data, not as a new user request or primary scientific evidence. Open-access primary full text was read where reachable. Kazemi [6] remains abstract-only after a further retrieval attempt on 2026-09-27; Zheng [18], Valles-Colomer [19], Chen [20], Zhuang [21] and the two syntheses [22,23] were read at abstract depth and their access depth is recorded as such. Citation-context counts were not available and remain zero by design. A passing structural validator will establish schema and internal consistency only, not independent scientific validation or clinical advice.

Red-team record. This edition is the red-team synthesis of the single v5.4 grading (OpenAI Codex, 2026-09-27), performed by an independent model of a different family. All 23 ledger rows were re-resolved in PubMed and all 17 inherited claims re-checked; no citation inversion, wrong identifier or retraction was found, and no subclaim or overall level moved. Three sources the base grading could not reach were retrieved in full text, which corrected two claims: the most supportive FMT trial [15] has a null prespecified primary endpoint and declining rather than durable engraftment, and the sham-controlled pilot's null symptom result [10] was confirmed from the publisher's own tables rather than inferred. Six sources were added after an omission audit in both directions, including the Mendelian-randomization evidence the upper-ladder check had omitted entirely and the analysis inside an existing source that points against the nominal model. Three disagreements with the base grading are recorded rather than absorbed: the departure from the index-prescribed weakest-link rule is upheld but re-argued on pitfall 21 and a necessity analysis, with the index rule's answer (Β) carried in this edition as dissent; the divergence artifact was re-derived from the field's own syntheses instead of from the claim's wording, landing on the same +2; and 07-psychiatry-neuromodulation was added as a secondary rubric under Rule 6, which changes no letter. The full trail is in PROCESS.md and redteam/. The index-prescribed weakest-link rule, which returns Β for this entry, was overridden in favour of scope-matched (Γ); because the amyloid-cascade edition overrides its own index rule in the opposite direction on the same day, the owner should settle the index entry for this hypothesis rather than leave the choice to per-review deviation.

ADVERSARIAL REVIEW AND APPROVAL

Authoring model: Sol 6.1. Adversarial reviewer: Claude Opus 5.5 (Anthropic). Date: 2026-10-14.

Scope of checks. All 23 reference PMIDs verified against PubMed (23/23 resolve to the stated title, first author, journal and year; no retractions or expressions of concern; no empty DOI fields). The reference list was additionally scanned for both forms of the collective-author defect -- a dropped author leaving a stray leading comma, and a consortium name truncated to its last word so that it reads as a surname (Group, Network, Consortium, Investigators, Collaboration, Trial, Study) -- and for dropped nobiliary particles; none is present, and no PubMed record in this ledger carries a collective first author. The eight historical identifiers named in the Consensus provenance block were checked and the three linked artifacts exist at the stated paths. Because this review bears on mental-health treatment decisions, every probiotic and FMT result was checked against its source, including which outcome was prespecified primary and whether the reported analysis was intention-to-treat, modified-ITT or completers-only: PMID 42309058 (Li FMT), 41785480 (Shekarriz allogenic versus autologous), 36637229 (Green pilot), 38732635 (PRO-DEMET, primary F(1,92)=0.58, P=0.45, with the positive minimal-important-difference result secondary), 39504580 (Borkent, 131 randomized, BPRS and BACS primary, null on psychiatric symptoms, zonulin improved, verbal memory borderline at adjusted P=0.059), 28483500 (Pinto-Sanchez, 14/22 versus 7/22, P=0.04), 37314797 (Nikolova, 49 analysed ITT), 35654766 (Schaub, completer sample N=21/26), 40298641 (Moerkl, 43 patients and 43 controls), 28068788 (Romijn, 79 participants analysed as intent-to-treat, null on every psychological outcome), 29731182 (Kazemi, BDI primary, 81 of 110 completers, P=0.042), 40177327 (Elahinejad, 44 of 50 completers, P=0.001, cortisol and ACTH null), 27491067, 37131071 and 27067014 (rodent transfer and vagotomy). Scoring logic re-derived against hsef-hypothesis-review-v5/SKILL.md Step 3 and references/hsef-framework.md. No level, confidence or role change.

Rejected findings (checked, no change).

1. Probiotic and FMT efficacy statements. Every intervention result matches its source, including the primary-versus-secondary distinction the brief asks about. PRO-DEMET's null primary comparison and its positive secondary minimal-important-change threshold are both reported, with the review stating explicitly that a secondary responder threshold cannot supersede the prespecified primary. Borkent's null primary psychiatric outcome alongside changed gut-barrier markers is reported exactly. Romijn is correctly described as intent-to-treat and null. Kazemi (81 of 110 completers) and Elahinejad (44 of 50 completers) are both labelled as completer analyses rather than ITT. Schaub's positive time-by-arm interaction is correctly attributed to the compliant subset rather than to the ITT set -- the published abstract reports only the completer sample, and the review's distinction is the stricter reading. Pinto-Sanchez, Nikolova, Green and Moerkl are all reported as their sources state. No change.
2. The Li FMT trial. The abstract confirms the two facts the review turns on: remission rates at week 8 did not differ significantly between groups, while HAMD-17 reductions were greater with FMT at weeks 2 and 8. The review's reading -- null prespecified primary, positive secondary -- is therefore correct on the published record. The body-level figures (43.3% versus 26.7%, P=0.1760; -2.9 points, 95% CI -5.5 to -0.3, P=0.0303; roughly 16% of species newly acquired at week 2 falling to about 7% at week 8) could not be independently re-verified here because the article is paywalled and has no PubMed Central deposit; they rest on the review's own recorded full-text read. The review's statement that the abstract's description of engraftment as durable is not what the engraftment analysis shows is a disagreement with the abstract that the review raises openly and attributes, which is the right handling rather than an error. Recorded in residual limitations.
3. Confidence band. Checked against the lead policy of 2026-10-14. This review's confidence_rationale already rests entirely on evidence density and reproducibility -- small trials, differing products, prespecified primaries disagreeing with the pooled estimates, no human mediator experiment, and key negatives replicated across products, diagnoses and delivery routes -- and makes no appeal to access depth. Moderate stands and no rewrite was needed.
4. Scope-matched aggregation and the departure from the index rule. Re-derived. SC4 is the subclaim whose scope matches the graded conjunction and it is Gamma; SC3's higher level sits at a narrower scope that drops the signalling clause and is correctly reported separately rather than setting the headline; SC2's level is restricted to rodents. All three alternative rules are reported, including the index-prescribed weakest-link answer, which is carried in the text as explicit dissent rather than suppressed. The departure is argued on pitfall 21 and on a necessity analysis, not on the letter it produces. Defensible and fully disclosed; no change.
5. "Best-supported rival" renderer artifact. Checked the tournament ordering: rank 1 is the context-bounded microbial symptom modifier, which is not this review's nominal model (the nominal M3 ranks 3 of 5). The prose is correct as written.
6. Momentum "Mechanistic only" and Stability "Stable". Both carry the required artifact -- named active trial programmes with improving controls, and the explicit absence of an identified near-term trigger. No "Decisive work underway" claim is made, which is correct in any case because the named trials are completed rather than running.
7. Collective-author defects, both variants. Scanned and absent; see scope of checks.
8. Rendering defects. Scanned for glued numerals, empty DOI: . fields and "PMIDnot supplied" strings; none present.
9. Lossy PubMed abstract fields. The PubMed abstract of PMID 40298641 is itself garbled mid-sentence — it reads “a significant increase in , particularly along with improved sleep”, with the taxon name dropped (the full text gives Christensenellales, particularly Akkermansia muciniphila). This is another instance of the lossy-field pattern behind the dropped-collective-author defect, recorded here so the corpus has the example. No text in this review depends on the dropped fragment; the correction it did require is finding G-3.

Residual limitations. Seven of the 23 records were read at abstract depth and one (Kazemi) is abstract-only, as the review states; the Mörkl figures added under finding G-3 were taken from the open-access full text (PMC12045568) during this pass and are the only body-level numbers this audit itself introduced. The Li trial's body-level endpoint, engraftment and mediation figures could not be re-verified independently because the article is paywalled; they are the single most load-bearing body-level numbers in this review. This pass is a source, claim and scoring-logic audit, not a blinded recalibration, and nothing in it constitutes clinical advice.

Approval: Following adversarial review and the corrections listed above, the revised document is approved. -- Claude Opus 5.5 (Anthropic), 2026-10-14.

References

1. Radjabzadeh et al., Nature Communications 2022. Gut microbiome-wide association study of depressive symptoms. DOI: 10.1038/s41467-022-34502-3; PMID 36473852. Primary full text; results, Mendelian randomization section and limitations read.
2. Fehse et al., Gut Microbes 2026. From gut to brain: data-driven evidence suggests causal contribution of gut-microbiota to major depressive disorder in humans. DOI: 10.1080/19490976.2026.2718589; PMID 42627212. Primary full text; methods, medication analysis and limitations read.
3. Kelly et al., Journal of Psychiatric Research 2016. Transferring the blues: Depression-associated gut microbiota induces neurobehavioural changes in the rat. DOI: 10.1016/j.jpsychires.2016.07.019; PMID 27491067. Primary full text (university repository PDF); donor selection, recipient preparation and behavioral results read.
4. Siopi et al., Molecular Psychiatry 2023. Gut microbiota changes require vagus nerve integrity to promote depressive-like behaviors in mice. DOI: 10.1038/s41380-023-02071-6; PMID 37131071. Primary full text; transfer, vagotomy, brain and behavior results read.
5. Pinto-Sanchez et al., Gastroenterology 2017. Probiotic Bifidobacterium longum NCC3001 Reduces Depression Scores and Alters Brain Activity: A Pilot Study in Patients With Irritable Bowel Syndrome. DOI: 10.1053/j.gastro.2017.05.003; PMID 28483500. Primary full text; prespecified endpoint, microbiota and fMRI results read.
6. Kazemi et al., Clinical Nutrition 2019. Effect of probiotic and prebiotic vs placebo on psychological outcomes in patients with major depressive disorder: A randomized clinical trial. DOI: 10.1016/j.clnu.2018.04.010; PMID 29731182. Primary abstract only; no open-access full text located on 2026-09-27.
7. Romijn et al., Australian and New Zealand Journal of Psychiatry 2017. A double-blind, randomized, placebo-controlled trial of Lactobacillus helveticus and Bifidobacterium longum for the symptoms of depression. DOI: 10.1177/0004867416686694; PMID 28068788. Primary full text; prespecified measures, ITT and null outcomes read.
8. Schaub et al., Translational Psychiatry 2022. Clinical, gut microbial and neural effects of a probiotic add-on therapy in depressed patients: a randomized controlled trial. DOI: 10.1038/s41398-022-01977-z; PMID 35654766. Primary full text; enrollment, ITT/mITT results, taxa and imaging read.
9. Nikolova et al., JAMA Psychiatry 2023. Acceptability, Tolerability, and Estimates of Putative Treatment Effects of Probiotics as Adjunctive Treatment in Patients With Depression: A Randomized Clinical Trial. DOI: 10.1001/jamapsychiatry.2023.1817; PMID 37314797. Primary full text; pilot design, ITT effect estimates and masking read.
10. Green et al., Canadian Journal of Psychiatry 2023. Feasibility, Acceptability, and Safety of Faecal Microbiota Transplantation in the Treatment of Major Depressive Disorder: A Pilot Randomized Controlled Trial. DOI: 10.1177/07067437221150508; PMID 36637229. Primary full text at the publisher; mental-health outcome tables (MADRS, DASS) and blinding assessment read.
11. Gawlik-Kotelnicka et al., Nutrients 2024. Metabolic Status Influences Probiotic Efficacy for Depression-PRO-DEMET Randomized Clinical Trial Results. DOI: 10.3390/nu16091389; PMID 38732635. Primary full text; prespecified MADRS outcome and primary versus secondary results read.
12. Borkent et al., Schizophrenia Bulletin 2025. Probiotic Formulation for Patients With Bipolar or Schizophrenia Spectrum Disorder: A Double-Blind, Randomized Placebo-Controlled Trial. DOI: 10.1093/schbul/sbae188; PMID 39504580. Primary full text; randomization, BPRS primary result and gut-marker results read.
13. Mörkl et al., Gut Microbes 2025. Multi-species probiotic supplement enhances vagal nerve function - results of a randomized controlled trial in patients with depression and healthy controls. DOI: 10.1080/19490976.2025.2492377; PMID 40298641. Primary full text; HRV endpoints, depression scores and microbiome results read.
14. Elahinejad et al., Food Science & Nutrition 2025. Effects of Probiotics as Adjunctive Therapy to Fluoxetine on Depression Severity and Serum Brain-Derived Neurotrophic Factor, Cortisol, and Adrenocorticotropic Hormone in Patients With Major Depressive Disorder: A Randomized, Double-Blind, Placebo-Controlled Trial. DOI: 10.1002/fsn3.4698; PMID 40177327. Primary full text; randomized design, completion, HDRS-24 and blood biomarkers read.
15. Li et al., Cell Host & Microbe 2026. Adjunctive fecal microbiota transplantation for major depressive disorder: A randomized, double-blind, placebo-controlled trial. DOI: 10.1016/j.chom.2026.05.017; PMID 42309058. Primary full text (authors' institutional copy); endpoint hierarchy, masking check, engraftment and mediation model read.
16. Shekarriz et al., Canadian Journal of Psychiatry 2026. The Safety, Efficacy, and Feasibility of Fecal Microbiota Transplantation in a Population With Bipolar Disorder During Depressive Episodes: A Pilot Parallel Arm Randomized Controlled Trial. DOI: 10.1177/07067437261420877; PMID 41785480. Primary full text; 35-person design, MADRS contrast and donor similarity read.
17. Wang et al., Scientific Reports 2026. A study on the efficacy and safety of fecal microbiota transplantation as an adjunctive therapy for treating depressive episodes. DOI: 10.1038/s41598-026-41801-y; PMID 41826399. Primary full text; participant flow, two-week HAMD-24 results and ten-target flora analysis read.
18. Zheng et al., Molecular Psychiatry 2016. Gut microbiome remodeling induces depressive-like behaviors through a pathway mediated by the host's metabolism. DOI: 10.1038/mp.2016.44; PMID 27067014. Primary abstract only; no open-access full text located on 2026-09-27. (added in the red-team pass)
19. Valles-Colomer et al., Nature Microbiology 2019. The neuroactive potential of the human gut microbiota in quality of life and depression. DOI: 10.1038/s41564-018-0337-x; PMID 30718848. Primary abstract only; publisher full text paywalled. (added in the red-team pass)
20. Chen et al., Journal of Affective Disorders 2022. Gut microbiota and major depressive disorder: A bidirectional Mendelian randomization. DOI: 10.1016/j.jad.2022.08.012; PMID 35961601. Primary abstract only; publisher full text paywalled. (added in the red-team pass)
21. Zhuang et al., Journal of Neuroinflammation 2020. Associations between gut microbiota and Alzheimer's disease, major depressive disorder, and schizophrenia. DOI: 10.1186/s12974-020-01961-8; PMID 33008395. Primary abstract only. (added in the red-team pass)
22. Asad et al., Nutrition Reviews 2025. Effects of Prebiotics and Probiotics on Symptoms of Depression and Anxiety in Clinically Diagnosed Samples: Systematic Review and Meta-analysis of Randomized Controlled Trials. DOI: 10.1093/nutrit/nuae177; PMID 39731509. Tier B; abstract read. Field framing and divergence basis only; no figure imported. (added in the red-team pass)
23. Moshfeghinia et al., Journal of Psychiatric Research 2025. The impact of probiotics, prebiotics, and synbiotics on depression and anxiety symptoms of patients with depression: A systematic review and meta-analysis. DOI: 10.1016/j.jpsychires.2025.05.053; PMID 40440772. Tier B; abstract read. Field framing and divergence basis only; no figure imported. (added in the red-team pass)

Scope qualifiers

No independence-basis table is required because no scored subclaim reaches Ζ.

Documentary record

Every grade on this site is reproducible from its record. The files below are the posted edition exactly as assessed.

• ADVERSARIAL_REVIEW.md
• SHA256.txt
• adversarial-findings.json
• codex-correction-provenance.json
• consensus-addendum.md
• consensus-validation.json
• edition-approval-audit.json
• gut-brain-psychiatric.review.md
• gut-brain-psychiatric.scorecard.json
• research-update-2026-10-05.bib
• research-update-2026-10-05.html
• research-update-2026-10-05.json
• research-update-2026-10-05.md
• research-update-2026-10-05.ris
• validation-report.json
• validation-round2-2026-10-14.json
• validation.txt
• video-insert-2026-10-05.md

Sources: 23 · load-bearing claims: 23 · verified: 23 · unverified: 0 · contradicted: 0 · schema 5.4