Randomized Trial of Targeted Indoor Spraying to Prevent Aedes-Borne Diseases.
Dean NE, Crisp AM, Che-Mendoza A, Kirstein OD, Barrera-Fuentes GA, Earnest JT, Puerta-Guardo HN, Collins MH, Pavia-Ruz N, Ayora-Talavera G, González-Olvera G, Medina-Barreiro A, Bibiano-Marín W, Jabbarzadeh S, Halloran ME, Longini IM Jr, Lenhart A, Waller LA, Correa-Morales F, Palacio-Vargas J, Gomez-Dantes H, Manrique-Saide P, Vazquez-Prokopec GM
- DOI
- 10.1056/NEJMoa2501069
- Record issued
- 2026-08-15
- Engine
- 7.39.0
- Exported
- 2026-09-21
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/2228dcbe-2f64-4c4d-9689-cd63f57421bc is authoritative.
How this rating was calculated
- IntegrityIntegrity concern−0.5★
- ClaimsEfficacy rests on an unvalidated surrogate endpoint−0.5★
- ReportingData & code availability partially met−0.25★
- 01Efficacy rests on an unvalidated surrogate endpoint
The primary endpoint is laboratory-confirmed symptomatic Aedes-borne disease (ABD), which is a clinical outcome, not a surrogate. However, the paper also reports entomological indices (Ae. aegypti density) as a secondary endpoint and uses them to support efficacy. The entomological reduction (59%) is presented as evidence of impact, but the paper does not provide validated evidence linking this entomological surrogate to clinical ABD reduction, nor does it demonstrate target engagement at the tested dose for the surrogate. The primary clinical endpoint itself is adequate, but the surrogate claim for entomological indices is inadequate.
“TIRS reduced indoor Ae. aegypti density by 59% (95% CI: 51%, 65%).”
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-conducted cluster-randomized trial with rigorous design, clear reporting of methods, and appropriate ethical approvals. The main weakness is the absence of a data availability statement and code sharing, which is a notable gap for a clinical trial. Minor reporting issues include lack of explicit reporting guideline adherence and some copyedit errors.
Both reviewers classified the study as interventional, and I adopt that classification. The evaluation covered all eight dimensions; several sub-criteria were marked not applicable (e.g., cell line authentication, housing conditions) due to the human field trial nature. The statistics verification recomputed only a subset of tests (1 test), so the absence of errors does not confirm overall statistical correctness.
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 1 test: 1 consistent, 0 inconsistent; 1 via agent-written checks.
- CONSISTENTreported p < .050 · recomputed p = .003Reviewer 1Check p-value for community impact estimate from CI
“estimated community impact was 24.0% (95% CI: 6.0%, 38.6%).”
Taken as given: The CI is a 95% confidence interval for the incidence rate ratio (log scale).; The estimate is the log rate ratio.Method: Using pCI function to derive p-value from estimate and CI, assuming log scale.How we recomputed it: pCI(0.24, 0.06, 0.386, 1)
- lowinternal contradictionThe abstract reports 4461 children monitored, but the results section reports 4461 in the entire cluster cohort at start of Year 1, which is consistent. However, the trial registration number in the abstract (30041-105) differs from the ClinicalTrials.gov identifier (NCT04343521) in the funding section.
The trial was registered in ClinicalTrials.gov (http://ClinicalTrials.gov) (identifier 30041-105; registered on April 13, 2020) ... Funded by National Institutes of Health and others; ClinicalTrials.gov (http://ClinicalTrials.gov) NCT04343521
Abstractreviewer’s wording
Overstated conclusions
2 findings · 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
- Conclusions only partially backed by the presented evidenceAssessed
8 major claims checked against the paper's own evidence: 1 only partially supported (evidence backs part of the claim; gaps or caveats remain); the rest adequately supported.
- partialReviewer 2Deploying one TIRS application preventively can have a measurable public health value when considering its entomological and community impacts.The claim is partially supported by the entomological and community impact results, but the primary analysis did not show a statistically significant reduction in the cohort, and the community impact analysis has limitations (e.g., potential confounding by age, lack of individual-level denominator).Evidence: Entomological efficacy (59% reduction) and community impact (24% reduction) are presented, but the primary cohort analysis was null.
“The combined trial results suggest that deploying one TIRS application preventively can have a measurable public health value when considering its entomological and community impacts.”
DiscussionFind in source - supportedReviewers 1, 2TIRS reduced indoor Ae. aegypti density by 59%.The claim is directly supported by the entomological efficacy analysis.Evidence: The paper reports a 59% reduction (95% CI: 51%, 65%) in Ae. aegypti density.
“TIRS reduced indoor Ae. aegypti density by 59% (95% CI: 51%, 65%).”
AbstractFind in source - supportedReviewers 1, 2No statistically significant reduction in ABD was detected among cohort children.The primary and secondary analyses show confidence intervals crossing zero, supporting the claim.Evidence: Primary efficacy estimate -12.8% (95% CI: -60.7%, 23.0%); ITT estimate 3.9% (95% CI: -28.1%, 26.7%).
“Despite reductions in entomological indices, no statistically significant reduction in ABD was detected among cohort children”
AbstractFind in source - supportedReviewer 1A community impact was quantified.The community impact estimate from national surveillance data shows a statistically significant reduction.Evidence: Community impact estimate 24.0% (95% CI: 6.0%, 38.6%).
“estimated community impact was 24.0% (95% CI: 6.0%, 38.6%).”
AbstractFind in source - supportedReviewer 1The trial did not replicate the high estimated effectiveness from an observational analysis in Australia.The paper acknowledges the difference and discusses possible reasons.Evidence: Discussion states the trial did not replicate the Australian effectiveness.
“Yet the trial did not replicate the high estimated effectiveness from an observational analysis in Australia.”
Discussion ¶2Find in source - supportedReviewer 2A community impact was quantified (24.0% reduction, 95% CI: 6.0%, 38.6%).The claim is supported by the pre-specified secondary analysis of national surveillance data, which shows a statistically significant reduction.Evidence: Community impact estimate of 24.0% (95% CI: 6.0%, 38.6%) from a Poisson regression of geolocated surveillance cases.
“estimated community impact was 24.0% (95% CI: 6.0%, 38.6%).”
AbstractFind in source - supportedReviewer 2Two multi-symptom adverse event cases were associated with TIRS.The claim is supported by the report of two adverse events possibly related to TIRS, with symptoms described.Evidence: Results section states: 'Two residents reported adverse events possibly related to the TIRS intervention, including colic/abdominal pain, nausea, watery eyes, and runny nose.'
Two residents reported adverse events possibly related to the TIRS intervention, including colic/abdominal pain, nausea, watery eyes, and runny nose.
Resultsreviewer’s wording - supportedReviewer 2The trial results highlight the need to better understand how entomological efficacy translates into epidemiological impact.This is a reasonable interpretation of the trial results, which showed entomological efficacy but not epidemiological efficacy in the primary analysis.Evidence: Discussion states: 'The trial results highlight the need to better understand how entomological efficacy translates into epidemiological impact'
“The trial results highlight the need to better understand how entomological efficacy translates into epidemiological impact”
Discussion ¶1Find in source
Premise concern: surrogate not validated for clinical benefit.
- INADEQUATESurrogate endpointThe primary endpoint is laboratory-confirmed symptomatic Aedes-borne disease (ABD), which is a clinical outcome, not a surrogate. However, the paper also reports entomological indices (Ae. aegypti density) as a secondary endpoint and uses them to support efficacy. The entomological reduction (59%) is presented as evidence of impact, but the paper does not provide validated evidence linking this entomological surrogate to clinical ABD reduction, nor does it demonstrate target engagement at the tested dose for the surrogate. The primary clinical endpoint itself is adequate, but the surrogate claim for entomological indices is inadequate.
“TIRS reduced indoor Ae. aegypti density by 59% (95% CI: 51%, 65%).”
- ADEQUATEEffect sizeThe primary efficacy analysis shows no statistically significant reduction in ABD (efficacy -12.8%, 95% CI -60.7% to 23.0%). The effect size is not clinically meaningful as the confidence interval includes zero and negative efficacy. However, the community impact analysis shows a 24.0% reduction (95% CI 6.0% to 38.6%), which is statistically significant and anchored to a clinical outcome (dengue cases from national surveillance). The paper does not claim a positive effect for the primary endpoint, and the community impact is presented as a secondary analysis with a modest but statistically significant effect.
“estimated community impact was 24.0% (95% CI: 6.0%, 38.6%)”
Data authenticity concerns
None foundAn 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
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 the growing public health impact of dengue, chikungunya, and Zika, and notes that classic control approaches have not succeeded. It references prior observational evidence for TIRS from Australia and mathematical models predicting its impact, and explicitly states the need for rigorous randomized trial evidence to inform policy. The limitations of prior work (observational design, reactive deployment) are acknowledged, and the trial is designed to address these gaps with a pre-seasonal, randomized design.
“In Australia, TIRS deployed to houses during a dengue outbreak prevented up to 86% of cases.”
“rigorous evidence of an intervention’s epidemiological impact is required to inform policy.”
“Classic control approaches against the primary ABD vector, Aedes aegypti (e.g., larval control, source reduction and space spraying), have not succeeded in reducing disease burden.”
“Yet the trial did not replicate the high estimated effectiveness from an observational analysis in Australia. In that context, TIRS was deployed reactively to a dengue outbreak via contact tracing to households.”
The trial uses covariate-constrained randomization to balance clusters, with a fair coin flip for allocation. The primary analysis uses a 'fried egg' design with per-protocol analysis of cluster centers, and a secondary ITT analysis of the entire cluster. A power analysis is reported with assumptions on incidence, ICC, loss to follow-up, and target effect size. Inclusion/exclusion criteria are described (children aged 2–15 years, residing in study clusters). Blinding is not reported as infeasible (unblinded design), which is acceptable for a public health intervention. Outlier handling is not explicitly discussed, but the analysis uses time-to-event methods which are robust. Controls are appropriate (routine vector control). Independent replication is not applicable for a single pivotal trial.
“Assuming 4% ABD incidence, an ICC of 0.035, and 20% loss to follow-up, the trial required 92 age-eligible children enrolled per cluster for an overall sample size of 50 clusters and 4600 children to have 80% power to detect a 70% reduction in ABD with a two-sided 0.05 test.”
“We conducted a two-arm, parallel, unblinded, cluster-randomized trial”
“Assuming 4% ABD incidence, an ICC of 0.035, and 20% loss to follow-up, the trial required 92 age-eligible children enrolled per cluster for an overall sample size of 50 clusters and 4600 children to have 80% power to detect a 70% reduction in ABD with a two-sided 0.05 test.”
“We conducted a two-arm, parallel, unblinded, cluster-randomized trial in Mérida, Mexico”
Sex and age at enrollment are reported for the cohort, with 51.5% male and median age 8.7 years in both arms. Baseline seroprevalence for dengue/Zika and chikungunya is reported from a subset. Demographics such as race/ethnicity and comorbidities are not reported, but this is not a major omission for a vector control trial. Species/strain and housing conditions are not applicable as this is a human study.
“Median age was 8.7 years in both trial arms, with 51.5% male.”
“Baseline serum samples from a subset of 1,399 children tested by commercially available ELISAs found that 45.1% were seropositive for dengue (DENV) or Zika (ZIKV) viruses and 24.0% were seropositive for chikungunya virus (CHIKV)”
“Median age was 8.7 years in both trial arms, with 51.5% male.”
“Baseline serum samples from a subset of 1,399 children tested by commercially available ELISAs found that 45.1% were seropositive for dengue (DENV) or Zika (ZIKV) viruses and 24.0% were seropositive for chikungunya virus (CHIKV)”
The paper states that the trial protocol was approved by the institutional review board at all collaborating institutions, and informed consent was obtained from participants or guardians. The trial was registered at ClinicalTrials.gov and complied with WHO ICTRP requirements. An independent external monitor reviewed consents and adverse effects. These statements are sufficient for a human trial.
“The trial protocol was approved by the institutional review board at all collaborating institutions.”
“Informed consent was obtained from all participants or their guardians, as detailed below.”
“The trial was registered in ClinicalTrials.gov (http://ClinicalTrials.gov) (identifier 30041-105; registered on April 13, 2020) and complied with the WHO International Clinical Trials Registry Platform requirements (ICTRP).”
“The trial protocol was approved by the institutional review board at all collaborating institutions.”
“Informed consent was obtained from all participants or their guardians, as detailed below.”
The insecticide pirimiphos-methyl (Actellic 300CS, Syngenta) is identified with manufacturer and formulation. The statistical software R version 4.4.1 is identified. Antibodies, cell lines, mycoplasma testing, and organisms are not applicable as this is a human trial with no wet-lab component. Reagents are covered by the investigational product identification.
“The organophosphate insecticide pirimiphos-methyl (Actellic 300CS, Syngenta) was selected”
“All statistical analyses were conducted with R software, version 4.4.1.”
“The organophosphate insecticide pirimiphos-methyl (Actellic 300CS, Syngenta) was selected”
“All statistical analyses were conducted with R software, version 4.4.1.”
The primary analysis uses an unadjusted Cox model with permutation-based confidence intervals, which is named. Effect sizes are reported with 95% confidence intervals for all efficacy estimates. Statistical software (R 4.4.1) is identified. Data presentation includes Kaplan-Meier curves with confidence bands and per-group Ns. Assumptions for the Cox model (proportional hazards) are not explicitly verified, but this is common for large trials. Exact p-values are not reported; instead, inference is based on confidence intervals, which is a valid approach. Mathematical plausibility checks are not applicable as the data are continuous and model-derived.
“efficacy was estimated as one minus the unadjusted Cox model hazard ratio with a permutation-based confidence interval over the constrained randomization space.”
“TIRS reduced indoor Ae. aegypti density by 59% (95% CI: 51%, 65%).”
“All statistical analyses were conducted with R software, version 4.4.1.”
“For each, efficacy was estimated as one minus the unadjusted Cox model hazard ratio with a permutation-based confidence interval over the constrained randomization space.”
“The estimated TIRS efficacy was −12.8% (95% CI: −60.7%, 23.0%).”
“All statistical analyses were conducted with R software, version 4.4.1.”
The paper does not include a dedicated data availability statement in the main text. The trial registration and protocol are mentioned, but no explicit statement about data access is provided. Code sharing is not mentioned. For a clinical trial, a managed-access statement with a mechanism (e.g., via a data access committee) would be expected for patient-level data. The absence of any data availability statement is a reporting gap.
Methods are detailed enough for replication, including trial design, randomization, intervention, endpoints, and statistical analysis. The trial is registered at ClinicalTrials.gov. Limitations are discussed in the Discussion section, including the impact of mobility, the 2023 outbreak, and the difference between entomological and epidemiological efficacy. Conclusions are proportional, noting the lack of statistical significance in the primary analysis while acknowledging the community impact signal. Funding sources and a conflicts-of-interest statement are provided. A specific reporting guideline (e.g., CONSORT) is not mentioned, but the paper's structure is consistent with such guidelines.
“The trial was registered in ClinicalTrials.gov (http://ClinicalTrials.gov) (identifier 30041-105; registered on April 13, 2020)”
“Mobility analyses further reveal that more time was spent out of the home in 2023 than 2021 and 2022. Out-of-cluster ABD exposure may be a key contributing factor to the reduced efficacy observed in the cohort”
“This study was supported by the National Institutes of Health, National Institute of Allergy and Infectious Diseases (U01AI148069; Vazquez-Prokopec, PI)”
“Mobility analyses further reveal that more time was spent out of the home in 2023 than 2021 and 2022. Out-of-cluster ABD exposure may be a key contributing factor to the reduced efficacy observed in the cohort”
Registered (1 ID: ClinicalTrials.gov). No reporting guideline cited.
Broken references and links
None foundReferences 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.
Checked — nothing surfaced.
Checked 32 references by DOI: 29 verified — 3 no DOI (shown, not verified).
- NO DOISituation Report No 28 - Dengue Epidemiological Situation in the Region of the Americas - Epidemiological Week 28, 2024No DOI in the reference — shown for manual review; not independently verifiable (not a fabrication signal).
- NO DOIManual for Indoor Residual Spraying in Urban Areas for Aedes aegypti ControlNo DOI in the reference — shown for manual review; not independently verifiable (not a fabrication signal).
- NO DOIEnding the neglect to attain the sustainable development goals: a road map for neglected tropical diseases 2021–2030: overviewNo DOI in the reference — shown for manual review; not independently verifiable (not a fabrication signal).
2 data/code links checked; 2 live.
- datahttp://ClinicalTrials.govLIVEHTTP 200Resolves, but the content could not be matched to the paper.
- datahttps://clinicaltrials.gov/ct2/show/NCT04343521LIVEHTTP 200Resolves, but the content could not be matched to the paper.
Copyediting
7 minorWording, consistency and formatting errors that need correcting before submission.
No major wording or formatting errors. 7 minor suggestions below.
7 copyedit issues flagged: mostly typo, consistency, grammar.
- MINORtypoAbstract, Results“There were 422 confirmed ABD cases, primarily dengue in 2023.”→ Consider rephrasing to 'The majority of the 422 confirmed ABD cases were dengue, occurring in 2023.'Clarity improvement.
- MINORconsistencyMethods, Statistical Analysis“All statistical analyses were conducted with R software, version 4.4.1.”→ Ensure the version number is consistent with any supplementary materials.Version consistency check.
- MINORgrammarDiscussion, paragraph 2“It could be that even greater reductions in household mosquitoes at the household, neighborhood, or regional level may be required.”→ Simplify to 'Greater reductions in household mosquitoes may be required.'Redundancy.
- MINORtypoAbstract, Trial registration“identifier 30041-105”→ Check if this is the correct ClinicalTrials.gov identifier; the URL provided is NCT04343521, which is different.The identifier in the text (30041-105) does not match the NCT number (NCT04343521) in the URL. This could be a grant number or internal ID.
- MINORconsistencyMethods, Statistical Analysis“stical Computing&)”→ Remove stray text 'stical Computing&)'.Appears to be a copy-paste artifact from a citation or software reference.
- MINORclarityResults, Table 1“Community impact (Secondary) | 150 | 202 | 24.0% (6.0%, 38.6%)”→ Clarify that the numbers 150 and 202 refer to ABD cases in TIRS and control clusters, respectively, and that the N for this analysis is not applicable (or state the number of clusters).The table header for community impact is ambiguous; it lists ABD cases without a denominator.
- MINORpunctuationMethods, Trial Design and Oversight“Figure 1: Map of study area Clusters selected within the Aedes -borne virus (ABV) hot-spot region of Mérida, Mexico.”→ Add a period after 'area' or restructure the sentence.The figure caption runs into the next sentence without punctuation.
The published work is methodologically robust and transparent, but readers should weigh the absence of a data availability statement and code sharing, which limits reproducibility. The trial registration identifier inconsistency (30041-105 vs NCT04343521) should be clarified, and the lack of explicit reporting guideline adherence is a minor gap. No evidence of statistical errors or retracted citations was found, so the findings appear reliable.
- 1.HIGHdata codeAdd a data availability statement in the Methods or a dedicated section, specifying how de-identified data can be accessed (e.g., through a data access committee or repository) and any conditions.The paper currently lacks any data availability statement, which is a reporting gap for a clinical trial and limits reproducibility.
- 2.HIGHdata codeShare custom analysis code in a public repository (e.g., GitHub, Zenodo) with a DOI, and reference it in the paper.Code sharing is not mentioned, and providing it would enhance transparency and reproducibility.
- 3.HIGHreportingResolve the inconsistency between the trial registration identifier '30041-105' in the abstract and the ClinicalTrials.gov identifier 'NCT04343521' in the funding section; ensure the correct identifier is used consistently.The integrity check flagged this internal contradiction, which could confuse readers and undermine trust in the registration.
- 4.MEDIUMreportingExplicitly reference the CONSORT checklist for cluster-randomized trials in the Methods or as a supplementary file.Adherence to a reporting guideline is not stated, and referencing it would improve reporting completeness.
- 5.MEDIUMstatisticsAdd a brief statement on verification of the proportional hazards assumption for the Cox models used in the primary analysis.Assumptions are not explicitly verified, and reporting this would strengthen the statistical analysis.
- 6.MEDIUMstatisticsInclude a statement on outlier handling or sensitivity analyses for the primary endpoint, even if none were excluded.Outlier handling is not reported, and a statement would clarify the robustness of the results.
- 7.MEDIUMreportingProvide a CONSORT flow diagram for the entire trial, including the number of clusters and participants at each stage.A flow diagram would improve transparency about participant flow and cluster allocation.
- 8.MEDIUMreportingReport the number of clusters analyzed in each arm for the primary and secondary analyses explicitly in the text.Currently inferable from tables, explicit reporting would improve clarity.
- 9.MEDIUMreportingAdd a brief rationale for the unblinded design in the Methods section.The trial is described as unblinded without justification, and a rationale would address a potential reviewer concern.
- 10.MEDIUMreportingSpecify the names and manufacturers of diagnostic ELISA kits used for serology to improve reagent identification.Reagent identification is incomplete, and specifying kits would improve reproducibility.
- 11.MEDIUMreportingProvide more detailed demographic data (e.g., race/ethnicity, socioeconomic status) for the cohort to improve generalizability.Demographics are reported as inadequate, and additional data would help readers assess external validity.
- 12.LOWcopyeditFix the stray text 'stical Computing&)' in the Methods, Statistical Analysis section.This appears to be a copy-paste artifact that should be removed for professionalism.
- 13.LOWcopyeditClarify the table header for 'Community impact (Secondary)' in Table 1, specifying that 150 and 202 refer to ABD cases in TIRS and control clusters, respectively, and state the denominator or number of clusters.The current table is ambiguous and could mislead readers about the analysis.
- 14.LOWcopyeditAdd a period after 'area' in the Figure 1 caption to separate the sentence.The caption runs into the next sentence without punctuation, which is a minor grammar issue.
- 15.LOWcopyeditRephrase the abstract sentence 'There were 422 confirmed ABD cases, primarily dengue in 2023.' to 'The majority of the 422 confirmed ABD cases were dengue, occurring in 2023.'The original phrasing is awkward and could be clearer.
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.