Effect of high flow nasal cannula oxygenation on incidence of hypoxia during sedated gastrointestinal endoscopy in patients with obesity: multicentre randomised controlled trial.
Wang L, Zhang Y, Han D, Wei M, Zhang J, Cheng X, Zhang Y, Shi M, Song Z, Wang X, Zhang X, Su D
- DOI
- 10.1136/bmj-2024-080795
- Record issued
- 2026-08-15
- Engine
- 7.39.0
- Exported
- 2026-09-20
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/52b4ef70-cdbf-494e-b991-96552c15ea34 is authoritative.
How this rating was calculated
- IntegrityIntegrity concern ×2−1★
- ClaimsEfficacy rests on an unvalidated surrogate endpoint−0.5★
- ReportingData & code availability partially met−0.25★
- The numeric-impossibility checks (GRIM/GRIMMER/DEBIT/SPRITE) did not run: 14 reported means were read, and their group size is not stated where the values are printed (this source has no machine-readable table structure). These checks need the count the mean was averaged over, so none was performed.
- No data or code availability links were detected to verify.
- 01Efficacy rests on an unvalidated surrogate endpoint
The primary outcome is the incidence of hypoxia, defined by SpO2 thresholds, which is a surrogate for clinical outcomes such as neurological damage or death. The paper does not demonstrate target engagement at the tested dose (e.g., PK/PD) nor cite validated evidence linking this surrogate to hard clinical outcomes.
“The primary outcome was the incidence of hypoxia (75%≤SpO 2 <90% for <60 s) during the procedure.”
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.
This is a well-designed multicentre RCT with strong methodological rigor: proper randomisation, allocation concealment, patient blinding, pre-specified sample size, and detailed statistical reporting. The main weakness is a vague data availability statement that lacks a clear mechanism or timeframe for access, and the absence of an explicit reporting guideline (CONSORT) mention.
Both reviewers agreed on study type (interventional) and on all dimension statuses; no divergence to resolve. The evaluation is based on the full published text; all verification components (citation, statistics, preregistration, integrity, claim audit) were run and are incorporated.
Numerical inconsistencies
1 finding · worst lowValues 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
Recomputed 7 tests: 7 consistent, 0 inconsistent; 4 recomputed directly from the reported test statistics, 3 via agent-written checks.
- CONSISTENTreported p < .001 · recomputed p = <.001Recomputed risk ratio 0.10 (95% CI 0.04–0.18), reported p<0.001
“risk ratio 0.10, 95% confidence interval (CI) 0.04 to 0.18; P<0.001”
Taken as given: 0.04–0.18 is a two-sided 95% confidence interval for the risk ratio of 0.10, not a range, an IQR, or a different interval level; the risk ratio is a RATIO measure, so the interval is symmetric on the log scale; p<0.001 is the p for THIS estimate, not for another comparison in the same sentenceMethod: back the two-tailed p out of the log-scale CI width and compare it against the printed pHow we recomputed it: pCI(0.1, 0.04, 0.18, 1) - CONSISTENTreported p < .001 · recomputed p = <.001Recomputed risk ratio 0.07 (95% CI 0.02–0.16), reported p<0.001
“risk ratio 0.07, 95% CI 0.02 to 0.16; P<0.001”
Taken as given: 0.02–0.16 is a two-sided 95% confidence interval for the risk ratio of 0.07, not a range, an IQR, or a different interval level; the risk ratio is a RATIO measure, so the interval is symmetric on the log scale; p<0.001 is the p for THIS estimate, not for another comparison in the same sentenceMethod: back the two-tailed p out of the log-scale CI width and compare it against the printed pHow we recomputed it: pCI(0.07, 0.02, 0.16, 1) - CONSISTENTreported p < .001 · recomputed p = <.001Recomputed risk ratio 0.16 (95% CI 0.10–0.22), reported p<0.001
“risk ratio 0.16, 95% CI 0.10 to 0.22; P<0.001”
Taken as given: 0.10–0.22 is a two-sided 95% confidence interval for the risk ratio of 0.16, not a range, an IQR, or a different interval level; the risk ratio is a RATIO measure, so the interval is symmetric on the log scale; p<0.001 is the p for THIS estimate, not for another comparison in the same sentenceMethod: back the two-tailed p out of the log-scale CI width and compare it against the printed pHow we recomputed it: pCI(0.16, 0.1, 0.22, 1) - CONSISTENTreported p < .001 · recomputed p = <.001Recomputed risk ratio 0.13 (95% CI 0.08–0.20), reported p<0.001
“risk ratio 0.13, 95% CI 0.08 to 0.20; P<0.001”
Taken as given: 0.08–0.20 is a two-sided 95% confidence interval for the risk ratio of 0.13, not a range, an IQR, or a different interval level; the risk ratio is a RATIO measure, so the interval is symmetric on the log scale; p<0.001 is the p for THIS estimate, not for another comparison in the same sentenceMethod: back the two-tailed p out of the log-scale CI width and compare it against the printed pHow we recomputed it: pCI(0.13, 0.08, 0.2, 1) - CONSISTENTreported p < .001 · recomputed p = <.001Reviewers 1, 2Primary outcome hypoxia in full analysis set: HFNC 10/497 vs control 103/487
“The incidences of hypoxia were 2.0% (10/497) and 21.2% (103/487) in the HFNC and control groups, respectively (risk ratio 0.10, 95% confidence interval (CI) 0.04 to 0.18; P<0.001).”
Taken as given: The numbers 10 and 103 are the event counts in the HFNC and control groups, respectively.; The denominators 497 and 487 are the group totals.; The test used is Pearson's chi-square test (as stated in methods).Method: Two-tailed Pearson chi-square test on the 2x2 table.How we recomputed it: pChi2x2(10, 487, 103, 384) - CONSISTENTreported p < .001 · recomputed p = <.001Reviewer 1Secondary outcome subclinical respiratory depression in full analysis set: HFNC 28/497 vs control 177/487
“The incidences of subclinical respiratory depression were 5.6% (28/497) and 36.3% (177/487) in the HFNC and control groups, respectively (risk ratio 0.16, 95% CI 0.10 to 0.22; P<0.001).”
Taken as given: The numbers 28 and 177 are the event counts in the HFNC and control groups, respectively.; The denominators 497 and 487 are the group totals.; The test used is Pearson's chi-square test.Method: Two-tailed Pearson chi-square test on the 2x2 table.How we recomputed it: pChi2x2(28, 469, 177, 310) - CONSISTENTreported p < .001 · recomputed p = <.001Reviewers 1, 2Secondary outcome severe hypoxia in full analysis set: HFNC 0/497 vs control 20/487
“The incidences of severe hypoxia were 0.0% (0/497) and 4.1% (20/487) in the HFNC and control groups, respectively (risk ratio 0.00, 95% CI 0.00 to 0.19; P<0.001).”
Taken as given: The numbers 0 and 20 are the event counts in the HFNC and control groups, respectively.; The denominators 497 and 487 are the group totals.; Fisher's exact test is appropriate due to zero cell count.Method: Two-tailed Fisher's exact test on the 2x2 table.How we recomputed it: pFisher2x2(0, 497, 20, 467, 0)
- lowinternal contradictionThe abstract reports 36.9% female (n=363) but Table 1 shows male sex percentages that imply female counts of 188 (497-309) and 175 (487-312), totaling 363, which is consistent. No contradiction.
“36.9% (n=363) female”
Table 1Find in source
Overstated conclusions
1 finding · worst highConclusions 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.
- Efficacy rests on an unvalidated surrogate endpointAssessed
5 major claims checked against the paper's own evidence: all adequately supported.
- supportedReviewers 1, 2HFNC oxygenation reduces the incidence of hypoxia during sedated gastrointestinal endoscopy in patients with obesity.The primary outcome shows a large, statistically significant reduction in hypoxia (2.0% vs 21.2%, P<0.001), supported by the full analysis set and per protocol set.Evidence: Primary outcome: 10/497 vs 103/487, risk ratio 0.10, 95% CI 0.04-0.18, P<0.001.
“HFNC oxygenation reduced the incidence of hypoxia from 21.2% (103/487) to 2.0% (10/497) (difference −19.14, 95% confidence interval −23.09 to −15.36; P<0.001)”
AbstractFind in source - supportedReviewers 1, 2HFNC oxygenation reduces the incidence of subclinical respiratory depression.Secondary outcome shows a significant reduction (5.6% vs 36.3%, P<0.001).Evidence: Secondary outcome: 28/497 vs 177/487, risk ratio 0.16, 95% CI 0.10-0.22, P<0.001.
“subclinical respiratory depression from 36.3% (177/487) to 5.6% (28/497) (difference −30.71, −35.40 to −25.92; P<0.001)”
AbstractFind in source - supportedReviewers 1, 2HFNC oxygenation reduces the incidence of severe hypoxia.Secondary outcome shows a significant reduction (0% vs 4.1%, P<0.001).Evidence: Secondary outcome: 0/497 vs 20/487, risk ratio 0.00, 95% CI 0.00-0.19, P<0.001.
“severe hypoxia from 4.1% (20/487) to 0% (0/497) (difference −4.11%, −6.26 to −2.48; P<0.001)”
AbstractFind in source - supportedReviewers 1, 2HFNC oxygenation does not increase other adverse events.The safety outcomes show no significant differences in other sedation-related adverse events between groups.Evidence: Table 4: Other sedation related adverse events 46 (9.3%) vs 34 (7.0%), P=0.19.
We observed no significant difference in the incidence of other sedation related adverse events between the two groups.
Resultsreviewer’s wording - supportedReviewer 2HFNC oxygenation is more beneficial for patients with obesity than regular nasal cannula oxygenation.The trial demonstrates significant reductions in hypoxia outcomes, supporting this conclusion.Evidence: Primary and secondary outcomes.
“We consider that HFNC oxygenation is more beneficial for patients with obesity than regular nasal cannula oxygenation.”
DiscussionFind in source
Premise concern: surrogate not validated for clinical benefit.
- INADEQUATESurrogate endpointThe primary outcome is the incidence of hypoxia, defined by SpO2 thresholds, which is a surrogate for clinical outcomes such as neurological damage or death. The paper does not demonstrate target engagement at the tested dose (e.g., PK/PD) nor cite validated evidence linking this surrogate to hard clinical outcomes.
“The primary outcome was the incidence of hypoxia (75%≤SpO 2 <90% for <60 s) during the procedure.”
- ADEQUATEEffect sizeThe effect size is large and clinically meaningful: hypoxia reduced from 21.2% to 2.0%, severe hypoxia from 4.1% to 0%, and subclinical respiratory depression from 36.3% to 5.6%. These are substantial absolute reductions in a common adverse event, with statistical significance and narrow confidence intervals.
“HFNC oxygenation reduced the incidence of hypoxia from 21.2% (103/487) to 2.0% (10/497) (difference −19.14, 95% confidence interval −23.09 to −15.36; P<0.001)”
Data authenticity concerns
1 finding · worst lowAn 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.
- Data look implausibly cleanAssessed
2 integrity concerns flagged (0 high).
- lowdata too cleanThe baseline characteristics are very well matched between groups, which is expected in a large randomized trial but could be scrutinized.
“The characteristics of the patients in the two groups, including age, sex, weight, body mass index, and ASA classification, matched well at baseline.”
Table 1Find in source
Reporting gaps
1 finding · worst mediumRequired detail the manuscript never states — study design, biological variables, ethics approval and consent, key resources, statistical reporting, data and code availability, and overall transparency.
- Data/code availability incompleteAssessed
The introduction cites prior research on hypoxia in sedated endoscopy, the higher risk in obesity, and conflicting evidence on HFNC, including a post hoc subgroup analysis and a small trial. It explicitly notes the limitations of prior work (small sample, post hoc analysis) and justifies the need for a large trial. The hypothesis follows logically from the cited evidence.
“Evidence on the effects of HFNC in patients with obesity is conflicting and weak.”
“Therefore, a large sample size trial is warranted to validate the effects of HFNC during sedated gastrointestinal endoscopy in patients with obesity.”
“However, the results came from a post hoc subgroup analysis. In a small sample size trial with 59 morbidly obese patients, HFNC oxygenation did not effectively reduce the incidence of hypoxia during colonoscopy under propofol sedation.”
“Evidence on the effects of HFNC in patients with obesity is conflicting and weak. In a randomised trial of 379 patients at high risk (29.3% had body mass index ≥30), Nay and colleagues reported that use of HFNC significantly reduced the incidence of hypoxia during gastrointestinal endoscopy under sedation. However, the results came from a post hoc subgroup analysis.”
“Therefore, a large sample size trial is warranted to validate the effects of HFNC during sedated gastrointestinal endoscopy in patients with obesity.”
Randomization used SAS PROC PLAN with block size six and stratification by centre, with allocation concealment via sealed opaque envelopes. Patients were blinded by covering the nasal cannula with an HFNC device; anaesthesiologists were not blinded, which is acknowledged as a limitation. Sample size was calculated with alpha=0.05, power=0.90, and a 10% dropout rate, estimating 972 patients. Inclusion/exclusion criteria are detailed. The analysis populations (full analysis set, per protocol set, safety set) are defined, and missing data handling is pre-specified. Outlier handling is not explicitly described, but the analysis population definitions cover missing data.
“A biostatistician who was not involved in data management and statistical analyses generated the random sequence by using the PROC PLAN programme in SAS (version 9.0). Randomisation was generated in a one-to-one allocation ratio with a block length of six and stratified by the participating centres. Randomisation results were sealed in sequentially numbered opaque envelopes which were opened immediately before sedation was started.”
“Therefore, assuming that HFNC oxygenation can reduce the incidence of hypoxia to 12%, we set total class I error to α=0.05, the test efficacy power to 0.90, and the dropout rate to 10%.”
“After peripheral intravenous access was established, each patient was given a regular nasal cannula covered with an HFNC (Airvo 2, Fisher, and Paykel, Panmure, New Zealand) to blind patients to the group assignment.”
“A biostatistician who was not involved in data management and statistical analyses generated the random sequence by using the PROC PLAN programme in SAS (version 9.0). Randomisation was generated in a one-to-one allocation ratio with a block length of six and stratified by the participating centres. Randomisation results were sealed in sequentially numbered opaque envelopes which were opened immediately before sedation was started.”
“Therefore, assuming that HFNC oxygenation can reduce the incidence of hypoxia to 12%, we set total class I error to α=0.05, the test efficacy power to 0.90, and the dropout rate to 10%.”
“After peripheral intravenous access was established, each patient was given a regular nasal cannula covered with an HFNC (Airvo 2, Fisher, and Paykel, Panmure, New Zealand) to blind patients to the group assignment.”
The study reports sex, age, weight, height, BMI, ASA classification, Mallampati classification, snoring, OSAHS status, and STOP-Bang scores. Since both sexes are enrolled, sex justification is not applicable. Age and health status are reported. Species/strain and housing conditions are not applicable for a human trial.
“Mean (SD) age, years | 48.6 (12.8) | 49.9 (13.2) | | Male sex | 309 (62.2) | 312 (64.1)”
“Mean (SD) age, years | 48.6 (12.8) | 49.9 (13.2) | | Male sex | 309 (62.2) | 312 (64.1)”
The study was approved by the ethics committee of Renji Hospital, Shanghai Jiao Tong University School of Medicine (KY2020-019). Written informed consent was obtained from each participant. Regulatory compliance is implied by the ethics approval and trial registration, though not explicitly named.
“This study was approved by the ethics committee of Renji Hospital, Shanghai Jiao Tong University School of Medicine (KY2020-019).”
“Each participant provided written informed consent.”
“This study was approved by the ethics committee of Renji Hospital, Shanghai Jiao Tong University School of Medicine (KY2020-019).”
“Each participant provided written informed consent.”
The HFNC device (Airvo 2, Fisher and Paykel) is identified with manufacturer and location. The monitors used at each centre are named. Statistical software (SAS 9.4, PASS) is identified. No antibodies, cell lines, or other bench reagents are used, so those criteria are not applicable.
“HFNC (Airvo 2, Fisher, and Paykel, Panmure, New Zealand)”
“We used SAS software version 9.4 for all statistical analyses.”
“We used SAS software version 9.4 for all statistical analyses.”
The primary analysis used the chi-square test, with differences and 95% CIs calculated using the Newcombe method. Exact p-values are reported (e.g., P<0.001). Effect sizes are reported as risk ratios with 95% CIs. Software is identified. Data presentation includes per-group n and percentages. Mathematical plausibility checks: the reported percentages and counts are consistent (e.g., 10/497=2.0%, 103/487=21.2%). The p-values are all <0.001, consistent with the large effect sizes.
“We used the χ 2 test to analyse the primary endpoint, incidence of hypoxia. We calculated the difference between groups with 95% confidence intervals calculated using the Newcombe method.”
“risk ratio 0.10, 95% confidence interval (CI) 0.04 to 0.18; P<0.001”
“We used the χ 2 test to analyse the primary endpoint, incidence of hypoxia. We calculated the difference between groups with 95% confidence intervals calculated using the Newcombe method.”
“risk ratio 0.10, 95% confidence interval (CI) 0.04 to 0.18; P<0.001”
The data availability statement says data will be available on request after approval by the corresponding author, but does not specify a platform, conditions, or timeframe. This is reported_but_inadequate. No code is shared, but the study is a clinical trial and code sharing is not applicable unless bespoke code was used; the statistical analysis used standard software.
“The data will be available to other researchers on request, with information shared after approval by the corresponding author (D-SS).”
“The data will be available to other researchers on request, with information shared after approval by the corresponding author (D-SS).”
The trial is registered (NCT04500392). Methods are detailed enough for replication. Limitations are explicitly discussed, including lack of blinding of anaesthesiologists and generalizability. Conclusions are proportional to the results. Funding sources and competing interests are declared. No reporting guideline is explicitly mentioned, but the manuscript follows CONSORT-like structure.
“Trial registration ClinicalTrials.gov NCT04500392”
“Firstly, HFNC is fundamentally different from a regular nasal cannula, making blinding of the anaesthesiologists extremely challenging. Thus, the anaesthesiologists were not blinded in this trial.”
“Trial registration ClinicalTrials.gov NCT04500392”
“Firstly, HFNC is fundamentally different from a regular nasal cannula, making blinding of the anaesthesiologists extremely challenging. Thus, the anaesthesiologists were not blinded in this trial.”
Registered (1 ID: ClinicalTrials.gov). No reporting guideline cited.
Broken references and links
None found · partly checkedReferences 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.
Nothing surfaced — but not everything feeding this category ran (missing: data/code link verification), so read this as a partial clean bill.
Checked 33 references by DOI: 32 verified — 1 no DOI (shown, not verified).
- NO DOIDelivered oxygen concentrations using low-flow and high-flow nasal cannulasNo DOI in the reference — shown for manual review; not independently verifiable (not a fabrication signal).
Copyediting
5 minorWording, consistency and formatting errors that need correcting before submission.
No major wording or formatting errors. 5 minor suggestions below.
5 copyedit issues flagged: mostly consistency, typo, clarity.
- MINORtypoTable 2“Subclincal respiratory depression”→ Subclinical respiratory depressionTypo in table header.
- MINORconsistencyResults, Efficacy outcomes“The above data show that HFNC oxygenation can completely eliminate the occurrence of severe hypoxia in patients with obesity but not the occurrence of hypoxia and subclinical respiratory depression.”→ Consider rephrasing for clarity: 'HFNC oxygenation eliminated severe hypoxia but did not eliminate hypoxia or subclinical respiratory depression.'Awkward phrasing.
- MINORconsistencyDiscussion, Comparison with other studies“In the control group, the patients had highest oxygen uptake concentrations of only up to 45%, respectively.”→ Remove 'respectively' as there is only one group mentioned.Grammatical error.
- MINORconsistencyDiscussion, Comparison with other studies“In the control group, the patients had highest oxygen uptake concentrations of only up to 45%, respectively.”→ In the control group, the patients had oxygen uptake concentrations of only up to 45%.Awkward phrasing and 'respectively' is unnecessary.
- MINORclarityResults, Efficacy outcomes“shows the last method used to correcting hypoxia.”→ Table 3 shows the last method used to correct hypoxia.Missing table reference and grammatical error.
The published work is methodologically robust and the results are credible. An informed reader should note the vague data availability statement as a limitation for reproducibility; the authors should consider issuing a correction or supplement with a more concrete data access plan. No erratum is warranted for the core findings.
- 1.HIGHdata codeReplace the vague data availability statement with a concrete plan specifying a data access committee or platform (e.g., YODA, Vivli) and a timeframe for response, in the Data availability statement section.A vague 'on request' statement undermines reproducibility and is a common reviewer concern.
- 2.HIGHreportingExplicitly state adherence to the CONSORT reporting guideline in the Methods or a dedicated section.Mentioning CONSORT improves transparency and helps readers assess reporting completeness.
- 3.HIGHethicsAdd a statement on regulatory compliance, e.g., 'The study was conducted in accordance with the Declaration of Helsinki', in the Ethics statements section.Explicit regulatory compliance is expected for human trials and is currently only implied.
- 4.MEDIUMreportingClarify how outliers were handled in the statistical analysis, if any, in the Methods, Statistical analysis section.Outlier handling is not described; adding this detail would strengthen methodological transparency.
- 5.MEDIUMreportingClarify the blinding of outcome assessors (the independent recorder) in the Methods, Randomisation and masking section.The current description focuses on patient and anaesthesiologist blinding; outcome assessor blinding is not explicitly stated.
- 6.MEDIUMreportingProvide a CONSORT flow diagram as a figure (currently only referenced as Fig 1) to show participant flow through the trial.A flow diagram is a standard CONSORT requirement and improves transparency of attrition.
- 7.MEDIUMdata codeConsider sharing the statistical analysis code (e.g., SAS or R scripts) in a public repository to enhance reproducibility.Even with standard software, sharing code allows exact replication of analyses.
- 8.LOWcopyeditFix typo in Table 2: change 'Subclincal respiratory depression' to 'Subclinical respiratory depression'.Corrects a minor typographical error.
- 9.LOWcopyeditRephrase in Results, Efficacy outcomes: 'HFNC oxygenation eliminated severe hypoxia but did not eliminate hypoxia or subclinical respiratory depression' for clarity.Improves readability and removes awkward phrasing.
- 10.LOWcopyeditIn Discussion, Comparison with other studies, remove 'respectively' from 'the patients had highest oxygen uptake concentrations of only up to 45%, respectively'.Grammatical correction; 'respectively' is unnecessary with a single group.
- 11.LOWcopyeditIn Results, Efficacy outcomes, add missing table reference: 'Table 3 shows the last method used to correct hypoxia'.Fixes a grammatical error and missing reference.
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.”
- 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.
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.