Editorial screen · real, unedited output on a published paper · a methods exercise, not a commentary on its authors

RigorMD Editorial

What the screen said about a paper a randomized trial later contradicted

One case · read blind · nothing about the aftermath supplied

Nurses' Health Study
source: NEJM 1996
2 engines + deterministic

§00 What this is, and what it is not

The author-side concordance benchmark tests the engine against papers with a documented correction or retraction — an adjudicated public record you can be concordant with, scored the same way each time. This page is not that, and it is not a benchmark. It is one case, on a question the field never fully settled, published because it is the most useful thing we can currently show a journal about how the screen reads an observational design.

The paper below has not been retracted. It is a careful analysis from one of the best-run cohorts in epidemiology and it stands as part of the literature. A one-page correction was published in the same volume two months later (N Engl J Med 1996;335:1406); we make no claim about its contents, which are not in the metadata. This is a methods exercise in a validation context — the way the literature routinely reads published work — not a commentary on any author.

§01 The case

Grodstein F, Stampfer MJ, Manson JE, et al. Postmenopausal Estrogen and Progestin Use and the Risk of Cardiovascular Disease. N Engl J Med 1996;335:453-461. The screen received this manuscript and the journal's current public author instructions. Nothing about what came afterward.

What the record went on to say

The Women's Health Initiative — already recruiting when this paper appeared — randomized 16,608 postmenopausal women to estrogen plus progestin or placebo, with coronary heart disease as its primary outcome. Coronary events were more frequent on therapy, not less: hazard ratio 1.29, nominal 95% CI 1.02–1.63, though the multiplicity-adjusted interval crosses 1 (0.85–1.97). The arm was stopped early on the invasive-breast-cancer boundary and an unfavourable global index — not on the coronary finding (JAMA 2002;288:321-333).

What the screen said, from the manuscript alone

Study validity came back a concern, on this ground: “The design is a strong observational cohort, but healthy-user self-selection and residual confounding — which the authors concede — sit at the center of whether the apparent coronary benefit is causal.”

The methodological section, verbatim

“Table 1 shows current users are systematically lower-risk (less diabetes, less smoking, less parental MI), so residual/unmeasured confounding biases toward apparent coronary benefit; the analysis also spans many endpoints and 13 risk-factor subgroups with no stated multiplicity correction, and key combined-therapy inferences rest on single-digit event counts.”

Printed verbatim from the run artifact — the complete, unedited evaluation is below, so every sentence quoted on this page can be read in its context.

What that is, precisely

threat named: residual confounding
direction stated: toward apparent benefit
routing: early editorial assessment
reproduced: 3 independent runs

Named one of the mechanisms later argued — not an adjudication among them

What this does not show

The screen did not predict the trial, and does not claim to

It named a threat to the inference and the direction that threat would push the result. That is not knowing the answer. The discrepancy has at least five candidate explanations in the literature: residual healthy-user confounding; the timing or critical-window hypothesis; prevalent-user versus incident-user design differences between a cohort and a trial; population differences (the trial's mean age was about 63, most participants well past menopause); and differences in regimen, dose, route and outcome ascertainment. The screen named the first. It did not rank them, and neither does this page — the trial published its own years-since-menopause analyses, and they have been argued over ever since.

Nor can we claim the models had never met this literature. They are pretrained, and this is one of the most discussed episodes in modern epidemiology. What we can say is that nothing about the aftermath was supplied to them, that the reading is argued from the manuscript's own sentences, and that the judgments are anchored to text a reader can check against the paper.

And the screen did not simply alarm. It credits the analysis first — extensive covariate adjustment is called appropriate — before naming the limit of what adjustment can do. Its highest-level dimension here is not the confounding at all but a compliance flag, raised because the supplied text carries no visible human-subjects ethics documentation or the other disclosure statements a journal now requires. That is the screen applying a journal's current instructions to a 1996 article, which is what it was given — an administrative gap, not a scientific one. A screen that shouted at a careful cohort study would be worth nothing to an editor.

§02 Why this is the wedge for a journal

The screen's own words for why it matters: “errors a clinician reviewer may read past.” A clinical reviewer reliably catches a wrong clinical claim. Directional confounding in a well-adjusted cohort is not a clinical error — it is a causal-inference one, and it survives clinical peer review precisely because the work looks careful. The evaluation puts the specific instance in its own finding: “Because current users have systematically fewer cardiovascular risk factors, residual confounding plausibly inflates the apparent coronary protection; the authors concede self-selection but the adjustment cannot fully remove unmeasured lifestyle differences.”

Two independent engines read this manuscript blind and disagreed about the route: one returned external review may be reasonable with defined major concerns, the other early editorial assessment recommended before reviewer assignment. The delivered route is early editorial assessment recommended before reviewer assignment, and the disagreement is recorded in the evaluation rather than smoothed away.

See the canonical sample How the screen works Editorial Review The author-side benchmark

§03 The evaluation, in full

Unedited output of the run this page describes — the same renderer a journal receives. Read it against the paper: every judgment is anchored to text you can check.

RigorMD Editorial Evaluation

Postmenopausal Estrogen and Progestin Use and the Risk of Cardiovascular Disease

Editorial decision support · Structured manuscript triage and peer-review support. Editors and editorial teams screening manuscripts.

generated 2026-09-06
format editorial-evaluation-report-2026-07-07.1

Structured editorial decision support. Not an editorial decision, and not a replacement for peer review.

Editor's snapshot

Is this worth sending to a reviewer?

Early editorial assessment recommended before reviewer assignmentFlag

Two engines read this manuscript and named different routes — first engine: External review may be reasonable with defined major concerns; second engine: Early editorial assessment recommended before reviewer assignment. The route above is the one delivered after reconciliation and, where it applies, the numeric layer's protection.

Clear

Audience fit & interest

A large prospective cohort analysis of postmenopausal hormone therapy and cardiovascular disease is squarely within scope for a general-medicine original article.

Concern

Study validity & design

The design is a strong observational cohort, but healthy-user self-selection and residual confounding — which the authors concede — sit at the center of whether the apparent coronary benefit is causal.

Concern

Conclusions vs. evidence

The headline conclusion that added progestin 'does not appear to attenuate' cardioprotection rests on very few combined-therapy events and causal-sounding language drawn from observational data.

Concern

Statistical reporting & rigor

Many subgroups and multiple endpoints are reported without multiplicity correction, some cells are very sparse, and only significance thresholds rather than exact p-values are given.

Flag

Guideline & checklist compliancefixable

Current journal-required ethics, COI, author-contribution, and AI-disclosure statements are not present in the supplied text, although funding is stated.

Clarity & communication

Not assessed: no grounded writing or communication signal in this read. The dimension is reported only when one is.

Contribution typewhere it sits in the literature — not a concern rating

ConfirmatoryEvolutionaryNovelLandmark

This is an evolutionary extension of earlier Nurses' Health Study hormone-CVD analyses into combined estrogen-progestin exposure; the sparseness of prior combined-therapy evidence is asserted by the manuscript and not independently verified here.

Journal and manuscript metadata

Manuscript: NEJM-1996-335-453
Article type: Original research
Journal section: Original Article
Prompt version: editorial-evaluation-prompt-2026-08-29.1

Topic and audience fitClear

Core scope general medicine

The manuscript examines postmenopausal hormone therapy and cardiovascular outcomes in a large cohort, matching the original-research/original-article designation and a general-medicine cardiovascular readership.

Grounding

Postmenopausal Estrogen and Progestin Use and the Risk of
information about the risk of cardiovascular disease associated with combined therapy is sparse.
We now report on the relation between combined hormone therapy and cardiovascular disease.
Our analysis is based on 16 years of follow-up data in 59,337 postmenopausal women participating in the Nurses' Health Study.

Design, data, and conclusions fitConcern

Observational causal overreach

A prospective cohort can establish association, but the conclusion is framed as an effect on 'cardioprotective effects,' and the pivotal combined-therapy estimate rests on few events, so the leap from association to attenuation-of-benefit exceeds what the data cleanly support.

Grounding

The addition of progestin does not appear to attenuate the cardioprotective effects of postmenopausal estrogen therapy.
multivariate adjusted relative risk, 0.39; 95 percent confidence interval, 0.19 to 0.78
Women who take hormones are a self-selected group and usually have healthier lifestyles with fewer risk factors than women who do not take hormones.
Proportional-hazards models were used to calculate relative risks and 95 percent confidence intervals, adjusted for confounding variables.

Potential for improvement with current dataConcern

Reframing and sensitivity analyses feasible

With the existing data the authors can soften causal language, foreground event counts for combined therapy, add a formal multiplicity discussion, and quantify how strong an unmeasured confounder would need to be — much of which they gesture at already.

Grounding

Unknown confounders may have influenced our results, but to explain the apparent benefit on the basis of confounding
in the subgroup of women who reported a visit in each period (accounting for 50 percent of the follow-up time), the relative risk of major coronary disease among the current hormone users was 0.52
For each participant, person-months were allocated to categories of hormone use according to the 1976 data and updated every two years

Statistical and methodological alignmentConcern

The section written for the reader a clinical reviewer is not: threats to the inference that survive clinical peer review — unadjusted confounding and its direction, multiplicity without correction, immortal time, an odds ratio narrated as a risk, a stateable missing-data mechanism, unmodelled clustering, competing risks treated as censoring, a precision claim the sample cannot support, ITT and per-protocol conflated. Where one applies it is named here; the instance it rests on, and the reviewer competence that would settle it, are below.

Confounding and multiplicity central

Table 1 shows current users are systematically lower-risk (less diabetes, less smoking, less parental MI), so residual/unmeasured confounding biases toward apparent coronary benefit; the analysis also spans many endpoints and 13 risk-factor subgroups with no stated multiplicity correction, and key combined-therapy inferences rest on single-digit event counts.

Grounding

but the data were based on only three cases of stroke among women using the combined regimen
The protective effect appeared to be similar in most subgroups.
Current hormone users, regardless of whether they used estrogen alone or with progestin, tended to have a better risk profile than women who had never used hormones (Table 1).
In subsequent analyses, we combined the data on use of estrogen alone and use of estrogen with progestin, since the results were similar for the two types of hormone therapy.
Tests of trends across categories of exposure were calculated by treating the levels of exposure as a continuous, ordinal variable in the regression model.
We performed an analysis to determine whether the effect of hormones on the risk of major coronary disease varied in specific subgroups of women (Table 5).

Statistical/methods follow-up needConcern

Epidemiology biostatistics review advised

An epidemiologist/biostatistician should assess the healthy-user confounding structure, the absence of multiplicity control across endpoints and subgroups, and the reliability of estimates built on very small event counts before external clinical review anchors on the conclusions.

Grounding

Tests of trends across categories of exposure were calculated by treating the levels of exposure as a continuous, ordinal variable in the regression model.
there was a trend toward an increased risk with higher doses (P for trend, 0.047).
relative risk, 0.32; P>0.05
current estrogen users as compared with women who had never used estrogen (P<0.05).

What was read

One further document was supplied with this submission and NOT read by the evaluation (a second manuscript document). Nothing below reflects its contents.

This evaluation read one manuscript document, which is what an editorial evaluation takes — 42,269 characters of extracted text.

The manuscript refers to Figure 1, Table 1, Table 2, Table 3, Table 4 and Table 5. The deterministic layer transcribed 47 structured values from the text and recomputed 49 checks.

Where a table is an image rather than selectable text, its cells cannot be transcribed and the checks that would depend on them do not run. This report does not distinguish a table with nothing to flag from a table it could not read — so a short list of checks is not, on its own, evidence that the numbers are sound.

Recomputed checks

Statistics recomputed from the manuscript's own printed numbers — arithmetic only, run independently of both appraisers. A consistent check means the printed value is reproducible from the values beside it; it does not assess the underlying data. The arithmetic on any one number is exact and repeats; which numbers this layer could pull out of the manuscript does not, so the count below is what this run reached rather than a fixed property of the paper. 49 checks recomputed · 48 consistent · 1 flagged.

Check

Threshold-only p-value reporting

presentation (mild): results report only significance thresholds (2 instances, e.g. "P>0.05") and never an exact p-value — report exact p-values (e.g. p = 0.03); p < 0.001 is the accepted floor

Reported vs recomputed

threshold_only: 2
exact_like: 0
examples: P>0.05, P<0.05

Inconsistent — flagged in findings

Consistent, nothing to report: Subgroup counts vs analytic N (2); Impossible p-values; Over-precise p-values; Cross-location value consistency (4); Effect estimate vs its confidence interval (40).

Non-obvious weaknessesConcern

Sparse events drive headline claim

The central 'progestin does not attenuate benefit' message leans on the estrogen-with-progestin coronary estimate built on only 8 age-adjusted (17 multivariate) cases, and the combined-regimen hemorrhagic-stroke reassurance on three cases; the wide intervals make these fragile foundations for a strong conclusion.

Grounding

multivariate adjusted relative risk, 0.39; 95 percent confidence interval, 0.19 to 0.78
but the data were based on only three cases of stroke among women using the combined regimen

Stroke dose response against safety framing

Table 4 stroke estimates rise monotonically across dose (0.64, 1.24, 1.44, 1.86) with a significant trend, a signal that partly counters the overall reassuring tone and warrants explicit weighting in the conclusions rather than a discussion aside.

Grounding

there was a trend toward an increased risk with higher doses (P for trend, 0.047)

Healthy user confounding direction stated

Because current users have systematically fewer cardiovascular risk factors, residual confounding plausibly inflates the apparent coronary protection; the authors concede self-selection but the adjustment cannot fully remove unmeasured lifestyle differences.

Grounding

Women who take hormones are a self-selected group and usually have healthier lifestyles with fewer risk factors than women who do not take hormones.

Self reported exposure and procedures

Hormone use was not validated and coronary-surgery data came from self-report alone, introducing exposure and outcome misclassification whose net effect on the relative risks is not quantified.

Grounding

Although we did not validate self-reported hormone use, we believe the reports were accurate
Data on coronary-artery surgery were obtained from the study participants' reports alone.

Healthy-user confounding biases toward apparent protection

Current hormone users had healthier baseline profiles and more preventive-medication use; incomplete measurement or residual confounding would be expected to bias the hormone association toward lower apparent coronary risk. This should be assessed by an observational epidemiologist or causal-inference statistician.

Grounding

Current hormone users, regardless of whether they used estrogen alone or with progestin, tended to have a better risk profile than women who had never used hormones (Table 1).
Fewer current users had a parental history of myocardial infarction, had diabetes, or smoked cigarettes.
Current users also took multivitamins, vitamin E, and aspirin more often than women who had never used hormones, were slightly younger and leaner, and drank more alcohol.

Non-attenuation claim lacks direct contrast

The central conclusion compares combined therapy conceptually with estrogen alone, but the stated relative-risk framework compares exposure categories with never-users, and the combined-therapy coronary estimate rests on few cases. A statistician should request a direct estrogen-plus-progestin versus estrogen-alone contrast, interaction test, and uncertainty interval before retaining non-attenuation language.

Grounding

The addition of progestin does not appear to attenuate the cardioprotective effects of postmenopausal estrogen therapy.
Estrogen with progestin 27,161 8 0.22 (0.12–0.41) 0.39 (0.19–0.78) 17

Multiplicity across secondary and subgroup analyses

Dose, duration, stroke subtype, and multiple risk-factor subgroup analyses are interpreted, including a borderline dose trend for stroke, without a stated multiplicity approach. A biostatistician should determine which analyses were prespecified and how secondary findings should be qualified.

Grounding

We performed an analysis to determine whether the effect of hormones on the risk of major coronary disease varied in specific subgroups of women (Table 5).
there was a trend toward an increased risk with higher doses (P for trend, 0.047).
Among current users, a longer period of use was not associated with any apparent trend toward a further reduction in the risk of either major coronary disease or stroke (P for trend, 0.73 for both).

Threshold-only p-value presentation

The supplied text contains threshold-only p-values; while this is not central to the main numerical conclusions, exact p-values should be provided when available or the threshold reporting should be justified.

Grounding

relative risk, 0.32; P>0.05
current estrogen users as compared with women who had never used estrogen (P<0.05).

Journal-specific publishing criteriaFlag

Structured abstract present

The abstract carries the required Background/Methods/Results/Conclusions structure, consistent with the journal's structured-abstract requirement.

Grounding

BACKGROUND
METHODS
RESULTS
CONCLUSIONS

Required statements incomplete

For current submission criteria, the supplied text contains a funding statement but no located IRB/ethics, COI, author-contribution, or AI-disclosure statement. Because this is human-participant cohort research, the missing ethics documentation is an administrative/ethics gate that should be resolved before reviewer assignment.

Grounding

Supported by grants (CA 40356 and HL 30594) from the National Institutes of Health.
The Nurses' Health Study began in 1976, when 121,700 female nurses, 30 to 55 years of age, completed a mailed questionnaire about their use of postmenopausal hormones and their medical history, including cardiovascular disease and associated risk factors.

Tables figures count over cap

Five tables plus Figure 1 total six display items against the journal's stated maximum of five tables/figures; the office should confirm the count and request consolidation if needed.

Grounding

FIGURE 1
TABLE 5
Abstract BACKGROUND

Trial registration not applicable

The registration/protocol/analysis-plan requirements are scoped to clinical trials; this is an observational cohort, so those specific flags do not apply, though the office should confirm the article-type routing.

Grounding

We examined the relation between cardiovascular disease and postmenopausal hormone therapy during up to 16 years

Clinical-trial-specific items not applicable

The study is presented as a prospective observational cohort analysis, not as an interventional clinical trial; trial registration, protocol/SAP submission, and clinical-trial data-sharing requirements therefore do not appear applicable on the supplied text.

Grounding

In this large prospective study, the risk of major coronary disease was substantially decreased among current users of estrogen and progestin, as well as among current users of estrogen alone.

Author-instruction and reporting-guideline complianceWatch

Observational reporting partly met

For a cohort study of the STROBE class, the manuscript reports eligibility, exclusions, adjustment covariates, and missingness for hormone-use time, but a formal reporting-checklist mapping is not evident in the captured text.

Grounding

information was missing for 3.2 percent of the follow-up time
Proportional-hazards models were used to calculate relative risks, with adjustments for age, age at menopause, body- mass index

P value reporting thresholds only

A recomputation check flags that results give only significance thresholds (e.g., P>0.05) rather than exact p-values in at least two instances; this is a presentation issue the office can request corrected.

Grounding

relative risk, 0.32; P>0.05

STROBE-style cohort reporting partially assessable

The manuscript reports cohort assembly, follow-up, outcome definitions, blinding of record review, covariates, and missing exposure time, which supports many STROBE-style elements. A formal checklist and fuller missing-data handling are not supplied, so a reviewer should verify reporting completeness.

Grounding

Follow-up data were available for over 90 percent of the cohort.
All interviews and reviews of medical records were conducted by investigators without knowledge of the category of hormone use.
Information on hormone use was missing for 21,534 person-years, 25 cases of coronary heart disease, and 20 cases of stroke

Place in the literatureWatch

Gap filling observational evidence

The manuscript positions itself as one of few analyses of combined therapy against a large body of estrogen-only data, and explicitly anticipates randomized confirmation, situating it as influential but provisional cohort evidence.

Grounding

The few epidemiologic studies of estrogen with progestin for the most part found an inverse association between
This issue will be addressed more directly in the next decade, when the results of clinical trials, such as the Women's Health Initiative, are known.
few women were taking progestin with estrogen.
The use of progestins combined with estrogen is now common, but information about the risk of cardiovascular disease associated with combined therapy is sparse.

Strengths worth preserving

Large cohort with long follow-up (662,891 person-years) and high retention, giving substantial power for the primary coronary endpoint.
Time-updated exposure classification with exclusion of prevalent disease at each interval, which mitigates reverse causation and immortal-time concerns.
Extensive covariate adjustment and transparent acknowledgment of self-selection and unknown-confounder limitations.
A physician-visit sensitivity analysis directly probing the surveillance/access-to-care confounder.
Large prospective cohort with long follow-up and time-updated hormone exposure classification.
Clear definitions of major coronary disease and stroke outcomes, with medical-record review for many events.
Blinded outcome interviews and record reviews with respect to hormone-use category.
Rich multivariable adjustment for cardiovascular risk factors and additional lifestyle/diet variables in sensitivity analyses.
Presentation of absolute rate differences by risk strata, not only relative risks.
Explicit discussion of healthy-user and residual-confounding concerns.

Suggested reviewer or statistical-review focus

Whether the combined estrogen–progestin conclusion is adequately supported given the single-digit and low-double-digit event counts and correspondingly wide confidence intervals.
The absence of any multiplicity adjustment across multiple endpoints, dose strata, and 13 risk-factor subgroups.
How strong an unmeasured confounder would need to be to explain the coronary association, and whether the physician-visit analysis fully addresses surveillance bias.
Interpretation of the significant stroke dose-response trend relative to the overall reassuring narrative.
Whether current-user versus never-user contrasts, with time-updated exposure, adequately handle prevalent-user and duration effects.
Assign an observational pharmacoepidemiologist or causal-inference statistician to assess residual healthy-user confounding and its likely direction.
Request a direct estrogen-plus-progestin versus estrogen-alone comparison, including interaction or heterogeneity testing and confidence intervals.
Evaluate whether the conclusion implies equivalence or non-inferiority despite sparse combined-therapy events and no stated equivalence margin.
Assess multiplicity across dose, duration, stroke subtype, and Table 5 subgroup analyses, especially the borderline estrogen-dose trend for stroke.
Check time-updated exposure modeling, proportional-hazards assumptions, and whether exposure changes after disease symptoms could introduce bias.
Confirm how missing hormone-use person-time was handled and whether complete-case assumptions are defensible.
Verify current required ethics, COI, author-contribution, AI-disclosure, and funding statements before reviewer assignment.

Highest-yield author queries

Please report the exact event counts and exact p-values underlying the estrogen-with-progestin coronary and stroke estimates and soften causal language accordingly.
Please provide a sensitivity/E-value-type analysis quantifying the confounding strength needed to nullify the coronary association.
Please state whether any correction for multiple endpoints and subgroups was applied, and if not, address the interpretive implications.
Please supply the required disclosure, author-contribution, AI-use, and ethics/IRB statements, and confirm the table/figure count against the journal cap.
Please reconcile the significant stroke dose-response trend with the conclusion that hormone therapy does not substantially affect stroke risk.
Provide IRB/ethics approval or exemption details, consent procedures if applicable, COI disclosures, author-contribution statement, AI-disclosure statement, and confirmation that trial registration is not applicable.
Report a direct estrogen-plus-progestin versus estrogen-alone analysis for major coronary disease and stroke, with confidence intervals and an interaction or heterogeneity test.
Clarify the prespecified primary endpoint and primary exposure comparison, and identify which dose, duration, stroke subtype, and subgroup analyses were prespecified.
Provide a multiplicity strategy or revise secondary/subgroup findings as exploratory, especially for the estrogen-dose trend in stroke.
Add sensitivity analyses or quantitative bias discussion for healthy-user and physician-contact confounding.
Replace threshold-only p-values with exact p-values where available.
Confirm whether the published correction changes any values in Tables 1 to 5 or the interpretation.

Limitations of this evaluation

The source is a web/HTML capture with advertising and navigation text interleaved; front-matter statements (COI, author contributions, AI, IRB) may exist in the full article but are not present in the captured text, so their absence cannot be confirmed.
A published correction (N Engl J Med 1996;335:1406) is referenced but not included, so specific numeric values could not be checked against the corrected version.
Body word count, reference formatting, and figure content could not be verified from the supplied capture.
Supplementary duration-of-use data (NAPS document) referenced in the text were not available for review.
No editor notes or private substantive review criteria beyond public author instructions were supplied, limiting journal-specific tailoring.
This evaluation used the supplied OCR text and table transcription, which include web-navigation and advertising artifacts.
The source states that the article has a published correction, but the correction content was not supplied.
No protocol, statistical analysis plan, data dictionary, or individual participant data were supplied, so model estimates could not be independently recomputed.
No submission forms, IRB documentation, COI forms, author-contribution forms, or AI-disclosure forms were supplied.
The supplied private journal criteria appear contemporary and may not reflect the journal's 1996 requirements.
Exact body word count could not be assessed reliably from the supplied PDF/OCR capture.

Scope Legal and scope limits

Scope. Structured editorial decision support; not an editorial decision and not a replacement for peer review. Not medical, legal, or regulatory advice. Limited to supplied materials. Terms | Privacy | Security