Alpha¹ by KaimenConfidential

Kaimen Rigor review

Genotype-stratified adjunctive dexamethasone for tuberculous meningitis in HIV-negative adults: a randomized controlled phase 3 trial.

Donovan J, Duc Bang N, Dong HKT, Ho DTN, Nguyen TAT, Nguyen TTH, Lam HBN, Phung VKN, Nguyen TT, Nguyen HHH, Pham KNO, Do DAT, Nguyen TMT, Dang TMH, Nguyen HL, Nguyen VVC, Hoang TH, Tran DD, Phung KL, Ramakrishnan L, Le THN, Nguyen TTT, Wolbers M, Kestelyn E, Geskus RB, Nguyen HP, Thwaites GE

DOI: 10.1038/s41591-025-04138-z
Record issued: 2026-08-10
Engine: 7.29.0
Exported: 2026-09-22

Prepared by Alpha1. This document is confidential: it is intended for the recipient it was shared with and must not be redistributed. The live record at alpha1science.com/verify/3e3292f5-e526-4ecb-9a73-ae8eab8c1b77 is authoritative.

Kaimen Rigor review

Kaimen Rigor reviewed

1.5 / 5

Critical — serious problems across many checks

Interventional Study

Critical

High-priority

How this rating was calculated

Started at 5★ — deductions

IntegrityIntegrity concern ×5−2.5★
StatisticsPrinted percentage does not match its own count (capped)−0.25★
ReportingEthical approvals partially met−0.25★
ReportingKey resources partially met−0.25★
CitationsUnresolved reference−0.25★

Model-reported confidence: highReviewer agreement: highEngine: 7.29.0Source: full text

Biggest issues

01Printed percentage does not match its own count
mediumArithmeticrecomputedFull detail
21.4% does not match the reported count 66/305
“66/305 (21.4%)”
Find in source
02Internal contradictions in the reported numbers
mediumArithmeticreviewer-assessedFull detail
The CC-genotype primary endpoint hazard ratio and CI differ between the Results text and Table 2: the text reports HR 0.81 with 98.3% CI 0.51–1.29, whereas Table 2 reports HR 0.83 with 95% CI 0.57–1.22 for the same subgroup.
“In the CC-genotype subgroup, death or new neurological events occurred in 50/146 (34.2%) of participants given dexamethasone and 56/145 (38.6%) given placebo (hazard ratio of 0.81, 98.3% CI 0.51–1.29).”
Table 2Find in source

Findings by category

201ArithmeticNumerical inconsistencies 102OverreachOverstated conclusions · partly checked 03AuthenticityNone found 204DisclosureReporting gaps 105AvailabilityBroken references and links 106CopyeditCopyediting

This Kaimen Rigor review uses Kaimen Rigor reviewers trained on a curated corpus of high-fidelity and retracted papers, with expert supervision and curation. It can still make mistakes; verify each finding against the source before relying on it.

The paper is methodologically strong overall, with rigorous design, detailed statistical reporting, and transparent data sharing. Two minor reporting gaps (missing regulatory framework citation, incomplete drug source identification) and several internal inconsistencies (HR discrepancy, alpha allocation, death percentage) are present but do not undermine the core findings.

Evaluation covers all eight rigor dimensions, with sub-criteria marked not applicable for human-trial-specific items (e.g., species/strain, housing). The reviewers independently agreed on all dimension statuses, and their evidence was merged. The copyedit and verification components revealed additional internal inconsistencies that are not reflected in the dimension scores but are captured in action items.

Numerical inconsistencies

2 findings · worst medium

Values that contradict each other or are impossible for the stated sample: recomputed p-values and test statistics, GRIM/GRIMMER checks on summary numbers, percentages against their own counts, totals against their parts, and estimates against their own confidence intervals.

Internal contradictions in the reported numbersAssessed
Printed percentage does not match its own countRecomputed

Recomputed 5 tests: 5 consistent, 0 inconsistent; 5 via agent-written checks. 1 printed percentage that does not match its own count.

Numeric forensics— values impossible for the reported sample size

PERCENT
21.4% does not match the reported count 66/305
“66/305 (21.4%)”
Find in source

Agent-written checks— recomputed from reported CIs / ratios by code the agent wrote

CONSISTENTreported p = .840 · recomputed p = .835Reviewer 1
Serious adverse event rate comparison (dexamethasone 161/305 vs placebo 160/308); reported P = 0.84
“In CC- and CT-genotype participants, at least one serious adverse event occurred in 161/305 (52.8%) dexamethasone-treated and 160/308 (51.9%) placebo-treated participants ( P value for difference = 0.84)”
Find in source
Taken as given: the 161 and 305 are the SAE count and group total of the dexamethasone arm; the 160 and 308 are the SAE count and group total of the placebo arm; the 144 and 148 are derived non-event counts (305-161, 308-160); the test is a two-sided Pearson chi-square on the 2x2 table
Method: Two-sided Pearson chi-square computed from the 2x2 cell counts (161/144 vs 160/148).
How we recomputed it: pChi2x2(161,144,160,148)
CONSISTENTreported p = .044 · recomputed p = .037Reviewer 1
Combined-trial mortality hazard ratio 0.78 (95% CI 0.62-0.99); reported P = 0.044
“death occurred in 135/624 (21.6%) given dexamethasone and 138/525 (26.3%) given placebo (hazard ratio of 0.78, 95% CI 0.62–0.99; P = 0.044; RMTL difference 0.42 months, 95% CI 0.04–0.79; P = 0.023)”
Find in source
Taken as given: the HR 0.78 and 95% CI 0.62-0.99 come from a Cox proportional-hazards model; the reported P = 0.044 is a two-sided p for the log-hazard ratio; the CI is two-sided at 95%; a Wald-type p derived from the CI may differ slightly from the reported p if a likelihood-ratio or score test was used
Method: Two-sided p recomputed from the log-hazard ratio and its 95% CI: z = log(0.78)/SE with SE = (log(0.99)-log(0.62))/(2*1.96).
How we recomputed it: pCI(0.78,0.62,0.99,1)
CONSISTENTreported p = .840 · recomputed p = .835Reviewer 2
Two-sided comparison of serious adverse event proportions between dexamethasone and placebo arms (reported P = 0.84).
“at least one serious adverse event occurred in 161/305 (52.8%) dexamethasone-treated and 160/308 (51.9%) placebo-treated participants ( P value for difference = 0.84)”
Find in source
Taken as given: The 161 and 160 are the event counts and 305 and 308 the group totals of the dexamethasone and placebo arms respectively; Non-event counts are 305−161=144 and 308−160=148; The reported P=0.84 is the two-sided comparison of the two proportions; df=1 because the table is 2×2
Method: Pearson chi-square on the 2×2 table (161/144/160/148); two-sided.
How we recomputed it: pChi2x2(161,144,160,148)
CONSISTENTreported p = .044 · recomputed p = .037Reviewer 2
P for the pooled IPD meta-analysis death hazard ratio (reported P = 0.044).
“Overall, death occurred in 135/624 (21.6%) given dexamethasone and 138/525 (26.3%) given placebo (hazard ratio of 0.78, 95% CI 0.62–0.99; P = 0.044”
Find in source
Taken as given: The 0.78 is the hazard ratio and 0.62–0.99 its two-sided 95% CI; The CI is on the log scale (log=1 for a ratio); The reported P=0.044 is the two-sided Wald p for the HR
Method: Recover SE from the 95% CI and compute a two-sided Wald z-based p for the log-hazard ratio.
How we recomputed it: pCI(0.78,0.62,0.99,1)
CONSISTENTreported p = .023 · recomputed p = .028Reviewer 2
P for the RMTL difference (reported P = 0.023).
“RMTL difference 0.42 months, 95% CI 0.04–0.79; P = 0.023”
Find in source
Taken as given: The 0.42 is the RMTL difference and 0.04–0.79 its two-sided 95% CI; The difference is on the additive scale (log=0); The reported P=0.023 is the two-sided p for the difference
Method: Recover SE from the 95% CI and compute a two-sided Wald z-based p for the difference.
How we recomputed it: pCI(0.42,0.04,0.79,0)

mediuminternal contradiction
The CC-genotype primary endpoint hazard ratio and CI differ between the Results text and Table 2: the text reports HR 0.81 with 98.3% CI 0.51–1.29, whereas Table 2 reports HR 0.83 with 95% CI 0.57–1.22 for the same subgroup.
“In the CC-genotype subgroup, death or new neurological events occurred in 50/146 (34.2%) of participants given dexamethasone and 56/145 (38.6%) given placebo (hazard ratio of 0.81, 98.3% CI 0.51–1.29).”
Table 2Find in source
lowinternal contradiction
In Table 2, the dexamethasone antituberculosis drug-resistance subgroups sum to 5+18+53+228 = 304 participants, one short of the 305 dexamethasone-treated participants in the ITT population; the placebo subgroups sum correctly to 308. This may reflect a participant with unrecorded resistance data but is not explained.
Multidrug resistant or rifampicin mono-resistant | 2/5 (40.0) | 1/3 (33.3) ... No or other resistance | 25/53 (47.2) | 25/54 (46.3) ... M. tuberculosis not isolated or missing result | 74/228 (32.5) | 74/235 (31.5)
Table 2reviewer’s wording
lowinternal contradiction
The stated type-I error allocation is internally inconsistent: Methods says spending 2% of a one-sided 2.5% error 'leaving 0.86%' for the CC subgroup, while Results says the 96% CI 'leaving 1.72%' for the CC subgroup.
“We chose to spend 2% of the one-sided type I error of 2.5% to the CC- and CT-genotype analysis, leaving 0.86% for the CC-genotype subgroup analysis.”
Statistical analysisFind in source
lowinternal contradiction
The death rate in the dexamethasone arm is reported as 21.4%, but 66/305 = 21.64%, which rounds to 21.6% (the placebo arm 66/308 = 21.43% correctly rounds to 21.4%). This is a minor rounding inconsistency.
“There were an equal number of deaths in the dexamethasone and placebo arms (66/305 (21.4%) and 66/308 (21.4%), respectively).”
ResultsFind in source
lowinternal contradiction
66/305 is reported as 21.4% but equals 21.64%, which rounds to 21.6%; the same percentage is used for both arms though the denominators differ.
“There were an equal number of deaths in the dexamethasone and placebo arms (66/305 (21.4%) and 66/308 (21.4%), respectively).”
ResultsFind in source

Overstated conclusions

1 finding · worst low

Conclusions that reach past what the paper's own results support — including a significance claim that no longer holds when the statistic is recomputed, and efficacy resting on an unvalidated surrogate endpoint.

Conclusions only partially backed by the presented evidenceAssessed

10 major claims checked against the paper's own evidence: all adequately supported.

partialReviewer 2
A combined IPD meta-analysis suggests dexamethasone reduced mortality regardless of LTA4H genotype.
The pooled HR supports a modest mortality benefit, but the paper itself cautions that between-trial differences (severity, diagnostics, mortality) may have introduced bias, and there were no placebo-treated TT participants, so 'regardless of genotype' extends beyond the direct evidence.
Evidence: Pooled death HR 0.78 (95% CI 0.62–0.99; P=0.044); RMTL difference 0.42 months (95% CI 0.04–0.79).
A planned individual participant data meta-analysis of 1,149 adults from the current and previous trial ... suggested that dexamethasone reduced overall mortality, regardless of LTA4H genotype.
Discussion ¶3reviewer’s wording
supportedReviewer 1
Neither noninferiority nor superiority of placebo was established in HIV-negative CC- and CT-genotype adults with tuberculous meningitis, and dexamethasone was safe.
The primary endpoint HR 0.99 with 96% CI 0.748-1.31 just covers the noninferiority margin of 0.75, so noninferiority was not established, and SAE rates were similar between arms (52.8% vs 51.9%).
Evidence: Primary outcome HR 0.99 (96% CI 0.748-1.31); SAE 161/305 vs 160/308.
“In conclusion, neither noninferiority nor superiority of placebo was established in human immunodeficiency virus-negative LTA4H CC- and CT-genotype adults with tuberculous meningitis, and dexamethasone was safe.”
AbstractFind in source
supportedReviewer 1
Placebo noninferiority was not established in the CC and CT population or in individual genotype subpopulations.
The CI just covered the margin (0.748-1.31 vs 0.75), and CC and CT subgroup CIs (98.3% and 95%) crossed 0.75.
Evidence: HR 0.99 (96% CI 0.748-1.31); CC HR 0.81 (98.3% CI 0.51-1.29); CT HR 1.09 (95% CI 0.76-1.59).
“Placebo noninferiority was not established in the CC and CT population or in individual genotype subpopulations.”
AbstractFind in source
supportedReviewer 1
Outcomes were not significantly better in TT-genotype participants versus CC- or CT-genotype participants.
The TT primary-outcome frequency (28/89, 31.5%) was similar to the CC/CT rate (35.6%), and no significant difference was reported.
Evidence: 28/89 (31.5%) TT vs 35.6% CC/CT; no significant between-genotype difference reported.
“In TT-genotype participants, the primary endpoint occurred in 28/89 (31.5%) participants, similar to CC and CT participants.”
AbstractFind in source
supportedReviewer 1
Combined data from this trial and the 2004 trial support dexamethasone's modest survival benefit in all HIV-negative adults with tuberculous meningitis, regardless of genotype.
The planned IPD meta-analysis of 1,149 participants showed a mortality HR of 0.78 (95% CI 0.62-0.99, P=0.044), supporting the modest benefit; the authors appropriately flag interpretational cautions.
Evidence: IPD meta-analysis: death 135/624 (21.6%) dexamethasone vs 138/525 (26.3%) placebo, HR 0.78 (95% CI 0.62-0.99; P=0.044).
Analysis of the combined data from this and the largest previous trial of dexamethasone supports dexamethasone's modest survival benefit in all HIV-negative adults with tuberculous meningitis
Discussionreviewer’s wording
supportedReviewer 1
LTA4H genotyping does not currently provide information that can influence clinical management.
The trial found no genotype-stratified differential benefit that would change practice; the exploratory TT benefit in severe disease is explicitly flagged as requiring further study.
Evidence: No significant heterogeneity of effect across genotypes or prespecified subgroups; TT exploratory benefit in severe disease noted as hypothesis-generating.
“LTA4H genotyping does not currently provide information that can influence clinical management.”
DiscussionFind in source
supportedReviewer 2
Placebo noninferiority to dexamethasone was not established in CC- and CT-genotype participants.
The reported CI lower bound (0.748) is below the prespecified margin (0.75), directly supporting the negative noninferiority conclusion.
Evidence: Primary endpoint HR 0.99 with 96% CI 0.748–1.31; the lower bound 0.748 falls below the noninferiority margin of 0.75.
“Placebo noninferiority was not established in the CC and CT population or in individual genotype subpopulations.”
AbstractFind in source
supportedReviewer 2
Superiority of placebo was not established.
The point estimate near 1.0 with CI spanning 1.0 supports the absence of superiority in either direction.
Evidence: HR 0.99 (CI 0.748–1.31) overall; no subgroup reached significance; superiority analyses showed no benefit of either arm.
“In conclusion, neither noninferiority nor superiority of placebo was established in human immunodeficiency virus-negative LTA4H CC- and CT-genotype adults with tuberculous meningitis”
AbstractFind in source
supportedReviewer 2
Dexamethasone was safe in this population.
Balanced serious-adverse-event rates directly support the safety claim as reported.
Evidence: Serious adverse events occurred in 161/305 (52.8%) dexamethasone vs 160/308 (51.9%) placebo (P=0.84); grade 3/4 events similar.
In conclusion, neither noninferiority nor superiority of placebo was established ... and dexamethasone was safe.
Abstractreviewer’s wording
supportedReviewer 2
Outcomes were not significantly better in TT-genotype participants than in CC- or CT-genotype participants.
The presented TT versus CC/CT event frequencies and absence of significant differences support the claim.
Evidence: TT primary endpoint 28/89 (31.5%) similar to CC/CT (35.6%); no significant between-genotype difference reported.
“Outcomes were not significantly better in TT-genotype participants versus CC- or CT-genotype participants.”
AbstractFind in source

Data authenticity concerns

None found

An adversarial read for patterns associated with data that may not be genuine: results that look too clean, implausibly large effects, duplicated data or images, and methods that do not match the results reported.

Checked — nothing surfaced.

Reporting gaps

2 findings · worst medium

Required detail the manuscript never states — study design, biological variables, ethics approval and consent, key resources, statistical reporting, data and code availability, and overall transparency.

Ethics/consent reporting incompleteAssessed
Key resources under-identified (antibodies, cell lines, RRIDs)Assessed

Eight Dimensions6 pass · 2 warn

The introduction cites the 2004 Vietnam dexamethasone trial, a Cochrane meta-analysis (RR 0.75), and the LTA4H genotype discovery connecting genotypes to distinct inflammatory phenotypes and dexamethasone responsiveness. The rationale directly motivates the genotype-stratified design and the decision to randomize only CC/CT carriers while giving TT carriers open-label dexamethasone. Limitations of prior work (post-hoc analyses, observational cohorts) are acknowledged in the Discussion as a reason for the discrepancy with past findings.

Sub-criteria

prior work citedADEQUATE

premise rationaleADEQUATE

limitations addressedADEQUATE

Per reviewer (2) — the votes behind the verdict

PASSReviewer 1· 90% conf

The premise is well established: the modest and heterogeneous benefit of corticosteroids, the discovery of LTA4H genotype-dependent inflammatory phenotypes, and prior trial evidence are all cited and logically linked to the trial's objectives.

Evidence

“Adjunctive corticosteroids such as dexamethasone are recommended in tuberculous meningitis treatment, despite modest and heterogeneous survival benefit.”

Abstract ¶1Find in source

“These findings suggested LTA4H genotype might be able to stratify HIV-negative individuals with tuberculous meningitis into those who benefit from dexamethasone and those who do not.”

Main, paragraph 3Find in source

“It may reflect the limitations of previous post-hoc analysis of trials and observational studies, which may have over-estimated the effects of LTA4H on pathophysiology and treatment outcomes.”

Discussion ¶4Find in source

PASSReviewer 2· 90% conf

The premise is firmly grounded in prior work and the rationale for a genotype-stratified trial is clearly articulated.

Evidence

paraphrase[Main, paragraph 3]

“A randomized controlled trial of the corticosteroid dexamethasone, conducted in Vietnam and published in the year 2004, found that dexamethasone significantly reduced overall mortality from tuberculous meningitis.”

“These findings suggested LTA4H genotype might be able to stratify HIV-negative individuals with tuberculous meningitis into those who benefit from dexamethasone and those who do not.”

Main, paragraph 4Find in source

“It may reflect the limitations of previous post-hoc analysis of trials and observational studies, which may have over-estimated the effects of LTA4H on pathophysiology and treatment outcomes.”

Discussion ¶4Find in source

Randomization method and unit are stated (computer-generated random permuted blocks, block size 4 and 6, stratified by hospital, LTA4H genotype and MRC severity; participants). Blinding is implied by the placebo control and the contrast of the open-label TT arm. Power analysis is explicit (80% power at one-sided 2% significance, 184 events). Inclusion/exclusion criteria are fully enumerated, and the analysis-population approach (ITT vs per-protocol) plus documented protocol deviations cover missing-data handling. Replicate distinction, positive/negative controls, and internal replication are not applicable to a single pivotal human RCT.

Sub-criteria

controlsN/A

power analysisADEQUATE

blinding levelsADEQUATE

outlier handlingADEQUATE

randomization unitADEQUATE

inclusion exclusionADEQUATE

randomization methodADEQUATE

replicate distinctionN/A

independent replicationN/A

Per reviewer (2) — the votes behind the verdict

PASSReviewer 1· 90% conf

For a human RCT, the applicable design elements are well reported: computer-generated randomized permuted blocks stratified by hospital, genotype and severity, a pre-specified power analysis, explicit inclusion/exclusion criteria, and ITT/per-protocol analysis populations.

Evidence

“The randomization list was computer-generated based on random permuted blocks with block size 4 and 6 (probability 0.75 and 0.25).”

MethodsFind in source

“To obtain 80% power at the one-sided 2% significance level, 184 events in the CC- and CT-genotype population would be required.”

Statistical analysisFind in source

“The intention-to-treat analysis included all randomized CC- and CT-genotype participants, even if no study drug was received after randomization.”

Statistical analysisFind in source

PASSReviewer 2· 85% conf

Core design elements — randomization, power, inclusion/exclusion, and analysis population definitions — are well reported; blinding is inferred from placebo control but not explicitly stated.

Evidence

“The randomization list was computer-generated based on random permuted blocks with block size 4 and 6 (probability 0.75 and 0.25).”

MethodsFind in source

“To obtain 80% power at the one-sided 2% significance level, 184 events in the CC- and CT-genotype population would be required.”

Statistical analysisFind in source

paraphrase[Methods, Statistical analyses]

“ITT included all randomized CC/CT participants; PP excluded those violating criteria or receiving <7 days study drug; no explicit double-blind/masking statement is given for the randomized arms.”

Sex is reported (435/702, 62.0% male) and the method of determination is stated (self-report). Age (median 47 years) and health status (MRC severity grade, diagnostic category, drug resistance) are reported. Demographics cover age, sex, HIV-negative status, and the Vietnamese population, though race/ethnicity is implicit rather than tabulated. Species/strain and housing are not applicable to a human trial.

Sub-criteria

demographicsADEQUATE

sex reportedADEQUATE

sex justifiedN/A

age weight healthADEQUATE

housing conditionsN/A

species strain sourceN/A

Per reviewer (2) — the votes behind the verdict

PASSReviewer 1· 85% conf

Sex, age, and health/disease-severity status are reported in the baseline table; the population is described as Vietnamese HIV-negative adults. Sex-based variables are not applicable to a mixed-sex human trial.

Evidence

“Male sex | 702 | 435 (62.0%)”

Table 1Find in source

“The median age of all participants was 47 years (first to third quartiles 33–59 years).”

ResultsFind in source

“Sex was determined based on self-reporting.”

Statistical analysisFind in source

PASSReviewer 2· 85% conf

Sex, age, health status, and demographics are reported for the human cohort; sex is balanced between arms and both sexes are enrolled.

Evidence

paraphrase[Table 1]

“Age (years) | 702 | 47 (33–59) ... Male sex | 702 | 435 (62.0%)”

“Sex and gender were not considered in the study design; while males are more commonly affected, tuberculous meningitis pathophysiology is not known to vary by sex.”

Statistical analysisFind in source

The IRB/ethics statement is strong: the Oxford Tropical Research Ethics Committee (approval no. 52-16), Hospital for Tropical Diseases (37/HDDD), Pham Ngoc Thach Hospital (1034/HDDD-PNT), and the Vietnam Ministry of Health (151/CN-BDGDD) are all named with protocol numbers. Informed consent is described (written consent from participants or a relative if incapacitated). No named compliance framework (Declaration of Helsinki, ICH-GCP, Common Rule) is cited, so regulatory_compliance is scored inadequate; this is a minor, easily-fixable reporting gap rather than a substantive concern.

Sub-criteria

iacuc statementN/A

informed consentADEQUATE

irb ethics statementADEQUATE

regulatory complianceINADEQUATE

Per reviewer (2) — the votes behind the verdict

WARNReviewer 1· 70% conf

Ethics approval is exceptionally well documented with four named bodies and approval numbers, and informed consent is described. However, regulatory_compliance does not name a recognized framework (e.g., Declaration of Helsinki, ICH-GCP), leaving one sub-criterion inadequate.

Evidence

“The Oxford Tropical Research Ethics Committee (approval no. 52-16), The Ethical Committee of the Hospital for Tropical Diseases (approval no. 37/HDDD), The Ethical Committee of Pham Ngoc Thach Hospital for Tuberculosis and Lung Disease (approval no. 1034/HDDD-PNT) and The Vietnam Ministry of Health (approval no. 151/CN-BDGDD).”

MethodsFind in source

“Written informed consent to enter the trial was obtained from all participants or a relative if they were incapacitated.”

MethodsFind in source

“Trial approvals were obtained from local and national ethics and regulatory authorities in Vietnam, and the Oxford Tropical Research Ethics Committee in the UK.”

MethodsFind in source

WARNReviewer 2· 65% conf

Named ethics bodies with approval numbers and informed consent are reported, but no recognized compliance framework (e.g., Declaration of Helsinki) is explicitly named.

Evidence

EthicsFind in source

“Written informed consent to enter the trial was obtained from all participants or a relative if they were incapacitated.”

MethodsFind in source

absence[Ethics and inclusion statement]

No named compliance framework

Dexamethasone is named with a fully detailed 6-8-week dose regimen (IV then oral, by weight and MRC grade) but no manufacturer/source is given; the antituberculosis drugs (rifampicin, isoniazid, pyrazinamide, ethambutol) are given with doses but no source. Genotyping reagents (Nucleon BACC2, Taqman kits, LightCycler 480) and sequencing reagents are identified with suppliers. Statistical/analysis software is well identified (R 4.4.2, STAR v2.5.2a, FeatureCounts v2.0.0, DESeq2 v1.34.0). Antibodies, cell lines, mycoplasma testing, and organisms are not applicable to this human drug trial.

Sub-criteria

mycoplasma testingN/A

reagents identifiedINADEQUATE

organisms identifiedN/A

antibodies identifiedN/A

cell line authenticationN/A

software tools identifiedADEQUATE

Per reviewer (2) — the votes behind the verdict

WARNReviewer 1· 70% conf

Software and genotyping/sequencing reagents are well identified, but the investigational product (dexamethasone) and the antituberculosis drugs are described with full dose/regimen but without manufacturer/source, leaving reagents_identified inadequate.

Evidence

paraphrase[Methods, Study treatments]

“Participants allocated dexamethasone received the 6–8-week regimen previously shown to reduce tuberculous meningitis mortality.”

“Rifampicin (10 mg kg −1 per 24 h, maximum 600 mg), isoniazid (5 mg kg −1 per 24 h, maximum 300 mg), pyrazinamide (25 mg kg −1 24 h, maximum 2 g) and ethambutol (20 mg kg −1 per 24 h, maximum 1.2 g)”

MethodsFind in source

“The data were analyzed using the program R (version 4.4.2; R Core Team, 2024)”

Statistical analysisFind in source

WARNReviewer 2· 60% conf

Analysis software is well identified, but the investigational product (dexamethasone) is not identified by manufacturer/source, so the drug-resource criterion is only partially met.

Evidence

paraphrase[Methods, Study treatments]

“Participants allocated dexamethasone received the 6–8-week regimen previously shown to reduce tuberculous meningitis mortality.”

“The data were analyzed using the program R (version 4.4.2; R Core Team, 2024)”

Statistical analysisFind in source

absence[Methods, Study treatments]

Dexamethasone source/manufacturer absent

Tests are named (Cox PH, logistic, proportional odds, RMTL, Wilcoxon, linear mixed effects, Bayesian joint model) and the PH assumption was tested (P > 0.05). Effect sizes are reported with CIs throughout, and the primary endpoint is reported by estimation (HR 0.99, 96% CI 0.748-1.31). p-values where given are largely exact (e.g., P=0.84 for SAE, P=0.044 for the IPD meta-analysis); some thresholds (P < 0.001) appear. Data presentation is the clinical idiom (Kaplan-Meier, forest plots, per-group n). One minor rounding inconsistency: 66/305 deaths = 21.6%, reported as 21.4%.

Sub-criteria

tests namedADEQUATE

exact p valuesADEQUATE

effect sizes ciADEQUATE

data presentationADEQUATE

software identifiedADEQUATE

assumptions verifiedADEQUATE

mathematical plausibilityWARN

Per reviewer (2) — the votes behind the verdict

PASSReviewer 1· 85% conf

Statistical tests, software, effect sizes with CIs, and data presentation are of high quality; the proportional-hazards assumption was verified. One trivial rounding discrepancy in a death percentage was noted but does not undermine the analysis.

Evidence

“The primary analysis was a Cox proportional hazards regression model with the primary endpoint as the outcome, treatment as the only covariate”

Statistical analysisFind in source

“The proportional-hazards assumption was tested for each model and was not violated ( P > 0.05).”

Table 2Find in source

paraphrase[Results, Primary outcome]

“death or new neurological events occurred in 108/305 (35.4%) given dexamethasone and 110/308 (35.7%) given placebo (hazard ratio of 0.99, 96% confidence interval (adjusted for multiple testing) 0.748–1.31)”

PASSReviewer 2· 80% conf

Tests, assumptions, effect sizes with CIs, software, and data presentation are well reported; a minor percentage-rounding slip does not constitute a demonstrable error.

Evidence

“The proportional-hazards assumption was tested for each model and was not violated ( P > 0.05).”

Table 2Find in source

“(hazard ratio of 0.99). To correct for performing an additional test in the CC-genotype subgroup, the CI was set at 96%, leaving 1.72% for the CC-genotype subgroup.”

ResultsFind in source

“There were an equal number of deaths in the dexamethasone and placebo arms (66/305 (21.4%) and 66/308 (21.4%), respectively).”

ResultsFind in source

The data-availability statement names a concrete route (OUCRU via named authors) with a timeframe (response within 4 weeks), which is adequate for identifiable patient-level data. Proteomic and transcriptomic data are deposited in Dryad (10.5061/dryad.f1vhhmh7v). R analysis code is publicly available at oucru-biostats/LAST-ACT on GitHub. The protocol and statistical analysis plan are published open access.

Sub-criteria

code sharingADEQUATE

accession numbersADEQUATE

repository depositADEQUATE

data availability statementADEQUATE

Per reviewer (2) — the votes behind the verdict

PASSReviewer 1· 90% conf

Data and code availability are exemplary: deidentified trial data shared via a named route with a 4-week response window, proteomic/transcriptomic data deposited in Dryad with a DOI, and R analysis code on a public GitHub repository.

Evidence

“Deidentified trial participant data (including data dictionaries) will be shared on request to the Oxford University Clinical Research Unit (via emailing the senior author, G.E.T., or corresponding author, J.D.). A response will be provided within 4 weeks of receiving the request.”

Data availabilityFind in source

“Proteomic and transcriptomic data used for this analysis have been deposited into a public repository and are available via Dryad at 10.5061/dryad.f1vhhmh7v”

Data availabilityFind in source

“R analysis code has been made publicly available via GitHub at oucru-biostats/LAST-ACT: Analysis code for LAST ACT trial, https://github.com/oucru-biostats/LAST-ACT”

Data availabilityFind in source

PASSReviewer 2· 85% conf

A concrete managed-access route for deidentified patient data, a Dryad deposit for -omics data, and a public code repository are all provided.

Evidence

Data availabilityFind in source

“Proteomic and transcriptomic data used for this analysis have been deposited into a public repository and are available via Dryad at 10.5061/dryad.f1vhhmh7v”

Data availabilityFind in source

“R analysis code has been made publicly available via GitHub at oucru-biostats/LAST-ACT: Analysis code for LAST ACT trial, https://github.com/oucru-biostats/LAST-ACT”

Data availabilityFind in source

The trial is registered at ClinicalTrials.gov (NCT03100786). A Nature Portfolio reporting summary is linked, covering CONSORT-style reporting. All secondary outcomes are stated to be presented in the manuscript. Limitations are thoroughly discussed (Vietnam-only, subset inflammation measures, IPD meta-analysis bias). Conclusions are measured and acknowledge the failure to establish noninferiority. Funding (Wellcome Investigator award 110179/Z/15/Z) and competing-interests availability are stated.

Sub-criteria

funding coiADEQUATE

trial registrationADEQUATE

reporting guidelineADEQUATE

methods completenessADEQUATE

all outcomes reportedADEQUATE

limitations discussedADEQUATE

conclusions proportionalADEQUATE

Per reviewer (2) — the votes behind the verdict

PASSReviewer 1· 85% conf

Reporting is transparent and complete: trial registered (NCT03100786), all outcomes stated as presented, a reporting summary linked, limitations extensively discussed, conclusions proportional, and funding/COI statements provided.

Evidence

“The trial was registered at ClinicalTrials.gov ( NCT03100786 (https://clinicaltrials.gov/ct2/show/NCT03100786) ).”

MethodsFind in source

“All outcomes are presented in this manuscript.”

ResultsFind in source

“Our trial has some limitations. First, it was conducted exclusively in Vietnam.”

Discussion ¶5Find in source

PASSReviewer 2· 85% conf

Registration, complete outcome reporting, limitations, proportional conclusions, and funding/COI are all addressed.

Evidence

“The trial was registered at ClinicalTrials.gov ( NCT03100786”

AbstractFind in source

“All outcomes are presented in this manuscript.”

ResultsFind in source

“Our trial has some limitations. First, it was conducted exclusively in Vietnam.”

Discussion ¶4Find in source

Registered (1 ID: ClinicalTrials.gov). No reporting guideline cited.

ClinicalTrials.gov

NCT03100786

Broken references and links

1 finding · worst low

References checked against Crossref, OpenAlex and Retraction Watch for retractions and resolvability, plus declared data and code links probed for whether they resolve to content matching the paper.

References not resolvable to a published paperRecomputed

Checked 39 references by DOI: 0 verified — 1 DOI unresolved, 38 no DOI (shown, not verified).

UNRESOLVED10.5061/dryad.f1vhhmh7v
Genotype stratified adjunctive dexamethasone for tuberculous meningitis in HIV-negative adults: the LAST ACT trial
Cited DOI does not resolve to any Crossref record.
NO DOI
Tuberculous meningitis: progress and remaining questions
No DOI in the reference — shown for manual review; not independently verifiable (not a fabrication signal).
NO DOI
The global burden of tuberculous meningitis in adults: s modelling study
No DOI in the reference — shown for manual review; not independently verifiable (not a fabrication signal).
NO DOI
Tuberculous meningitis and miliary tuberculosis: the Rich focus revisited
No DOI in the reference — shown for manual review; not independently verifiable (not a fabrication signal).
NO DOI
The mechanisms and consequences of the extra-pulmonary dissemination of Mycobacterium tuberculosis
No DOI in the reference — shown for manual review; not independently verifiable (not a fabrication signal).
NO DOI
Tuberculous meningitis
No DOI in the reference — shown for manual review; not independently verifiable (not a fabrication signal).
NO DOI
Prognostic models for 9-month mortality in tuberculous meningitis
No DOI in the reference — shown for manual review; not independently verifiable (not a fabrication signal).
NO DOI
Leukotriene A4 hydrolase genotype and HIV infection influence intracerebral inflammation and survival from tuberculous meningitis
No DOI in the reference — shown for manual review; not independently verifiable (not a fabrication signal).
NO DOI
Dexamethasone for the treatment of tuberculous meningitis in adolescents and adults
No DOI in the reference — shown for manual review; not independently verifiable (not a fabrication signal).
NO DOI
The clinical benefit of adjunctive dexamethasone in tuberculous meningitis is not associated with measurable attenuation of peripheral or local immune responses
No DOI in the reference — shown for manual review; not independently verifiable (not a fabrication signal).
NO DOI
Serial MRI to determine the effect of dexamethasone on the cerebral pathology of tuberculous meningitis: an observational study
No DOI in the reference — shown for manual review; not independently verifiable (not a fabrication signal).
NO DOI
The lta4h locus modulates susceptibility to mycobacterial infection in zebrafish and humans
No DOI in the reference — shown for manual review; not independently verifiable (not a fabrication signal).
NO DOI
Host genotype-specific therapies can optimize the inflammatory response to mycobacterial infections
No DOI in the reference — shown for manual review; not independently verifiable (not a fabrication signal).
NO DOI
A Bayesian analysis of the association between Leukotriene A4 Hydrolase genotype and survival in tuberculous meningitis
No DOI in the reference — shown for manual review; not independently verifiable (not a fabrication signal).
NO DOI
Corticosteroids for managing tuberculous meningitis
No DOI in the reference — shown for manual review; not independently verifiable (not a fabrication signal).
NO DOI
Treatment-associated inflammatory deterioration in tuberculous meningitis: unpicking the paradox
No DOI in the reference — shown for manual review; not independently verifiable (not a fabrication signal).
NO DOI
Tuberculous meningitis: combined therapy with cortisone and antimicrobial agents
No DOI in the reference — shown for manual review; not independently verifiable (not a fabrication signal).
NO DOI
Effect of corticosteroids on intracranial pressure, computed tomographic findings, and clinical outcome in young children with tuberculous meningitis
No DOI in the reference — shown for manual review; not independently verifiable (not a fabrication signal).
NO DOI
Adjunctive dexamethasone for tuberculous meningitis in HIV-positive adults
No DOI in the reference — shown for manual review; not independently verifiable (not a fabrication signal).
NO DOI
Elevated cerebrospinal fluid cytokine levels in tuberculous meningitis predict survival in response to dexamethasone
No DOI in the reference — shown for manual review; not independently verifiable (not a fabrication signal).
NO DOI
LTA4H prevalence and mortality in adult Zambians with tuberculous meningitis
No DOI in the reference — shown for manual review; not independently verifiable (not a fabrication signal).
NO DOI
Association analysis of the LTA4H gene polymorphisms and pulmonary tuberculosis in 9115 subjects
No DOI in the reference — shown for manual review; not independently verifiable (not a fabrication signal).
NO DOI
Relationship between human LTA4H polymorphisms and extra-pulmonary tuberculosis in an ethnic Han Chinese population in Eastern China
No DOI in the reference — shown for manual review; not independently verifiable (not a fabrication signal).
NO DOI
Relationship between LTA4H promotor polymorphism and tuberculosis-associated immune reconstitution inflammatory syndrome and its prevention with prednisone
No DOI in the reference — shown for manual review; not independently verifiable (not a fabrication signal).
NO DOI
Clinical parameters, routine inflammatory markers, and LTA4H genotype as predictors of mortality among 608 patients with tuberculous meningitis in indonesia
No DOI in the reference — shown for manual review; not independently verifiable (not a fabrication signal).
NO DOI
Advancing the chemotherapy of tuberculous meningitis: a consensus view
No DOI in the reference — shown for manual review; not independently verifiable (not a fabrication signal).
NO DOI
Tuberculous meningitis: a uniform case definition for use in clinical research
No DOI in the reference — shown for manual review; not independently verifiable (not a fabrication signal).
NO DOI
Adjunctive dexamethasone for the treatment of HIV-uninfected adults with tuberculous meningitis stratified by leukotriene A4 hydrolase genotype (LAST ACT): Study protocol for a randomised double blind placebo controlled non-inferiority trial
No DOI in the reference — shown for manual review; not independently verifiable (not a fabrication signal).
NO DOI
Statistical analysis plan for the LAST ACT clinical trial; a leukotriene A4 hydrolase stratified non-inferiority trial of adjunctive corticosteroids for HIV-negative adults with tuberculous meningitis
No DOI in the reference — shown for manual review; not independently verifiable (not a fabrication signal).
NO DOI
LTA4H genotype is associated with susceptibility to bacterial meningitis but is not a critical determinant of outcome
No DOI in the reference — shown for manual review; not independently verifiable (not a fabrication signal).
NO DOI
Streptomycin treatment of tuberculous meningitis
No DOI in the reference — shown for manual review; not independently verifiable (not a fabrication signal).
NO DOI
Xpert MTB/RIF Ultra versus Xpert MTB/RIF for the diagnosis of tuberculous meningitis: a prospective, randomised, diagnostic accuracy study
No DOI in the reference — shown for manual review; not independently verifiable (not a fabrication signal).
NO DOI
The SVA package for removing batch effects and other unwanted variation in high-throughput experiments
No DOI in the reference — shown for manual review; not independently verifiable (not a fabrication signal).
NO DOI
STAR: ultrafast universal RNA-seq aligner
No DOI in the reference — shown for manual review; not independently verifiable (not a fabrication signal).
NO DOI
featureCounts: an efficient general purpose program for assigning sequence reads to genomic features
No DOI in the reference — shown for manual review; not independently verifiable (not a fabrication signal).
NO DOI
Moderated estimation of fold change and dispersion for RNA-seq data with DESeq2
No DOI in the reference — shown for manual review; not independently verifiable (not a fabrication signal).
NO DOI
Adjusted significance levels for subgroup analyses in clinical trials
No DOI in the reference — shown for manual review; not independently verifiable (not a fabrication signal).
NO DOI
R: A Language and Environment for Statistical Computing
No DOI in the reference — shown for manual review; not independently verifiable (not a fabrication signal).
NO DOI
Whole blood transcriptional profiles and the pathogenesis of tuberculous meningitis
No DOI in the reference — shown for manual review; not independently verifiable (not a fabrication signal).

4 data/code links checked; 4 live.

datahttps://clinicaltrials.gov/study/NCT03100786LIVEHTTP 200
Resolves, but the content could not be matched to the paper.
datahttps://clinicaltrials.gov/ct2/show/NCT03100786LIVEHTTP 200
Resolves, but the content could not be matched to the paper.
datahttp://clinicaltrials.gov/study/NCT03100786LIVEHTTP 200
Resolves, but the content could not be matched to the paper.
codeGitHubLIVEHTTP 200
https://github.com/oucru-biostats/LAST-ACT
Resolves to GitHub (code repository).

Copyediting

1 finding · worst low

Wording, consistency and formatting errors that need correcting before submission.

Wording or formatting errors that need correctingAssessed

4 copyedit issues flagged (1 major): mostly consistency.

MAJORconsistencyResults, Primary outcome vs Table 2
“hazard ratio of 0.81, 98.3% CI 0.51–1.29”
→ Reconcile with Table 2 CC row '0.83 (0.57–1.22)' and state which model/specification each derives from.
The CC-genotype hazard ratio and CI differ between the text (0.81; 0.51–1.29) and Table 2 (0.83; 0.57–1.22).
MINORconsistencyResults, Secondary outcomes
“There were an equal number of deaths in the dexamethasone and placebo arms (66/305 (21.4%) and 66/308 (21.4%), respectively).”
→ 66/305 = 21.6%, so the dexamethasone-arm percentage should read 21.6% (the placebo value 21.4% is correct).
Minor rounding discrepancy; does not affect the analysis.
MINORconsistencyMethods, Statistical analyses vs Results, Primary outcome
“leaving 0.86% for the CC-genotype subgroup analysis”
→ Reconcile the 0.86% and 1.72% alpha allocations for the CC subgroup.
Methods says 2% of a 2.5% one-sided error leaves 0.86%; Results says CI set at 96% 'leaving 1.72%' for the CC subgroup.
MINORconsistencyResults, Secondary outcomes
“66/305 (21.4%) and 66/308 (21.4%)”
→ Change the first entry to 21.6% (66/305 = 21.64%).
66/305 rounds to 21.6%, not 21.4%.

Kaimen Rigor Action ItemsFinal pass before journal submission

The published paper is generally robust, but several internal inconsistencies (HR discrepancy between text and Table 2, type-I error allocation, death percentage rounding) should be addressed via an erratum or data clarification. Until resolved, readers should treat the reported point estimates and subgroup analyses with caution. The missing drug source and regulatory framework statement are minor but could be corrected in a comment or corrigendum.

1.
HIGHreporting
Reconcile the CC-genotype primary endpoint hazard ratio between Results text (0.81, 98.3% CI 0.51–1.29) and Table 2 (0.83, 95% CI 0.57–1.22) and clarify which model each represents.
This is a major internal inconsistency that could affect interpretation of the primary subgroup analysis.
2.
HIGHreporting
Clarify the type-I error allocation: Methods states spending 2% of a one-sided 2.5% error leaves 0.86% for the CC subgroup, but Results says the 96% CI leaves 1.72% for the CC subgroup; ensure internal consistency.
The alpha split is inconsistent between the Methods and Results, which could mislead readers about the multiple testing correction.
3.
HIGHrigor
Provide the manufacturer/source (and lot number if relevant) for dexamethasone and the four antituberculosis drugs (rifampicin, isoniazid, pyrazinamide, ethambutol) in the Methods, Study treatments section.
Complete identification of investigational products is essential for reproducibility and is a key resource requirement.
4.
HIGHethics
Add an explicit statement of adherence to a recognized regulatory framework (e.g., Declaration of Helsinki, ICH-GCP) in the Ethics and inclusion statement.
The current statement lists approvals but does not name the framework, which is a standard expectation for clinical trials.
5.
HIGHdata code
Verify that the Dryad DOI (10.5061/dryad.f1vhhmh7v) is correctly registered and accessible; the citation verification did not find it in the registry.
A not-found DOI could indicate a registration error or a potentially fabricated reference; readers need assurance that the deposited data are accessible.
6.
MEDIUMreporting
Correct the death percentage in the dexamethasone arm from 21.4% to 21.6% (66/305) in Results, Secondary outcomes.
The arithmetic inconsistency is minor but undermines the precision of the reported summary statistics.
7.
MEDIUMreporting
Reconcile the antituberculosis drug-resistance subgroups in Table 2 for the dexamethasone arm: 5+18+53+228 = 304, not 305; add a missing-data category or clarify the footnote.
The sum of subgroup counts should match the total randomized participants; the discrepancy suggests a participant with unrecorded resistance data.
8.
MEDIUMreporting
Add an explicit statement of participant/investigator blinding (double-blind masking) in Methods, Randomization and study groups, rather than leaving it implied by placebo control.
Explicit reporting of blinding is a standard requirement for RCTs and was noted as inadequate by one reviewer.
9.
LOWdata code
Add a versioned permanent identifier (e.g., Zenodo DOI) for the public code repository in Code availability.
A GitHub link alone does not guarantee a stable snapshot; a DOI would provide a permanent, citable version of the code.
10.
LOWstatistics
Note the test used for the IPD meta-analysis p-value (P=0.044) to clarify that the CI-based Wald p is approximately 0.037, to avoid an apparent mismatch.
This is a minor methodological detail that could be clarified to prevent confusion.

Full text (79,336 chars · analyzed ~95%, head + tail)Analyzed: Aug 10, 2026

RigorReviewer 1

RigorReviewer 2

RigorPre-Submission Reviewer

CopyeditAgent Scribe

CitationsCitation Integrity

StatisticsStatistics Consistency

ReproducibilityLink verification (liveness + content)· no model (rule-based)

PreregistrationDeterministic check· no model (rule-based)

The star rating is the report’s one-glance summary. Every paper starts at 5★ and loses stars for the concrete problems the review finds — so a rating is never a vague average, it’s a running total you can read line by line under “How this rating was calculated.”

How stars are deducted

Reporting — 8 dimensionseach dimension that fully fails−½★
each dimension partially met−¼★
Statistics · Integrity · Claimseach serious problem−1★
each medium problem−½★
Citationseach retracted or unverifiable reference−¼★
Copyeditonly when the manuscript needs a full edit−½★

The rating never drops below 1★, and a demonstrable critical failure (an impossible statistic, a proven ethics violation) caps it at 1★ on its own — so the stars can never look healthy when the verdict is CRITICAL.

The rating draws on a panel of agents. Three independent Kaimen Rigor reviewers grade the eight dimensions below across several independent passes (the shown verdict is their majority vote — steadier than any single run), isolate the paper’s major claims and check its own evidence backs them, and flag integrity concerns. Alongside them, a citation agent resolves every reference against Crossref, OpenAlex, and Retraction Watch; a statistics agent recomputes reported tests; and rule-based checks verify that declared data/code links actually resolve. Full text is required — an abstract-only submission is not analyzed.

The eight reporting dimensions

Scientific PremisePrior work, rationale, and how it addresses gaps.

Study Design & RigorRandomization, blinding, power, controls, replication.

Biological VariablesSex, age, species/strain, housing, demographics.

Ethical ApprovalsIRB/IACUC approval, informed consent, regulatory compliance.

Key ResourcesAntibodies, cell lines, organisms, software — identified with persistent IDs.

Statistical ReportingNamed tests, exact p-values, effect sizes, plausible summary stats.

Data & Code AvailabilityAvailability statements, repositories, accession numbers, shared code.

Reporting TransparencyMethods, registration, guidelines, outcomes, limitations, funding/COI.

Graded against NIH, MDAR, ARRIVE 2.0, CONSORT, EQUATOR, and RRID guidelines. A dimension that doesn’t apply to the study type is skipped, never penalized.

This Kaimen Rigor review is model-assisted and is not a substitute for formal expert review. It complements human evaluation by surfacing potential methodological concerns — verify each finding against the source.

Genotype-stratified adjunctive dexamethasone for tuberculous meningitis in HIV-negative adults: a randomized controlled phase 3 trial.

DOI

10.1038/s41591-025-04138-z

Record issued

2026-08-10

Engine

7.29.0

Exported

2026-09-22