Public research example · NHANES
Has blood pressure control changed among cancer survivors?
Using public NHANES 2015–2018 and 2021–2023 data, compare changes in uncontrolled blood pressure prevalence between cancer survivors and people without cancer among US adults receiving antihypertensive treatment.
Frozen research question
Among US adults receiving antihypertensive treatment, did the change in uncontrolled blood pressure prevalence from 2015–2018 to 2021–2023 differ between cancer survivors and people without cancer?
Example process
What steps does this example include?
- 01
Check official CDC/NCHS NHANES public files, variable documentation, data structure, and direct-identification risks.
- 02
Freeze the population, two survey periods, cancer history, uncontrolled blood pressure outcome, and eligibility rules.
- 03
Prespecify complex-survey weights, strata, PSUs, interaction terms, and marginal standardization.
- 04
Run survey-weighted logistic regression, same-device sensitivity analyses, and blood pressure device bridging.
- 05
Organize tables, scientific figures, and the manuscript; perform consistency checks, software quality review, and bounded revision.
Main result
Period interaction OR 0.91
- Interval estimate
- 95% CI 0.64–1.30
- Two-sided test
- p=0.597
There was no statistical evidence that change across periods differed between cancer survivors and people without cancer. The difference in changes in standardized prevalence was −2.24 percentage points (95% CI −10.72–6.24). The interval still permits clinically meaningful differences in either direction, so the results establish neither a difference nor equivalence between groups.
Why is NHANES not an ordinary spreadsheet analysis?
NHANES uses a complex, multistage probability sample. Omitting weights, strata, and PSUs removes the design basis for population estimates, standard errors, and confidence intervals. This example fixed survey weights before analysis and used Taylor linearization variance estimation with PSUs nested within strata.
The main model was survey-weighted logistic regression including cancer survivor status, period, and their interaction, adjusted for age, sex, and race/ethnicity. Main absolute effects were reported using marginally standardized prevalences and the difference in their changes.
Who was included in the comparison?
The analysis included 4,816 US adults taking antihypertensive medication: 2,446 people without cancer and 518 cancer survivors in 2015–2018, and 1,433 people without cancer and 419 cancer survivors in 2021–2023.
The outcome was uncontrolled blood pressure defined using prespecified thresholds, with 2,950 events. Cancer history was self-reported. Cancer type, stage, and treatment details were unavailable, so between-group differences cannot be attributed to a particular cancer treatment.
What can and cannot be concluded?
Standardized uncontrolled blood pressure prevalence changed from 59.0% to 59.9% among people without cancer and from 58.5% to 57.1% among cancer survivors. Confidence intervals for both the interaction estimate and the absolute difference in changes crossed their null values. The study did not identify a clear cancer-survivor-specific change over periods.
Blood pressure measurement protocols changed between periods, so the main results must be interpreted with same-device sensitivity analyses and paired-device bridging. This serial cross-sectional association study cannot identify individual longitudinal change, pandemic effects, or cancer treatment effects.
Demonstration materials
What materials are recorded in this example?
- Complete revised manuscript with DOCX and PDF renderings
- 5 research tables and 3 scientific figures based on actual data
- 13 verified publications and claim-to-source mappings
- Internally reviewable analysis and reproducibility materials, not publicly downloadable
- Software quality checks, simulated peer review, and revision records
- Editable graphical abstract files for internal evaluation only
Quality checks
Software checks do not mean submission-ready
- The internal delivery contains 45 files and passed independent checks of the file inventory, formats, and cross-material bindings.
- Recalculation in a clean environment compared 895 floating-point values; the maximum absolute difference was below the prespecified 1e−12 tolerance.
- The software publication-readiness score was 93/100. This score is not an impact factor, acceptance probability, or human expert judgment.
- Internal status: PASS_WITH_WARNINGS / major_revision / submission_allowed=false.
Public-use boundaries
How should this example not be interpreted?
- This is an internal workflow demonstration using public data, not a real user task in the production Web product or a commitment about the product's current deliverables.
- This serial cross-sectional association study cannot establish individual change over time or infer that cancer, treatment, or the pandemic caused changes in blood pressure control.
- Blood pressure measurement protocols changed across periods. Interpret the main results alongside device bridging and same-device sensitivity analyses.
- Submission governance facts, including authors, affiliations, ethics or exemption grounds, funding, competing interests, and author contributions, have not been supplied.
- No human experts or external peer reviewers were involved. All review conclusions on this page come from software quality processes.
- Source redistribution requirements have not yet been bound. This page provides no raw data or reproducibility-package downloads; it displays only software-checked aggregate results and process descriptions.
Sources
Verified sources
- CDC/NCHS National Health and Nutrition Examination SurveyOfficial NHANES program page
- NHANES 2015–2016One official access point for the earlier period
- NHANES 2017–2018Another official access point for the earlier period
- NHANES August 2021–August 2023Official access point for the recent period
Public-data demonstration