Decomposing the Study Aim

BMI, central adiposity, and mortality in MASLD

Example paper: Kueh et al. BMJ Open. 2026;16:e113719

Teaching focus: Move from a broad published aim to a clearly defined research question — and see exactly which NHANES variables it rests on.

What did the authors say they wanted to study?

The paper has two related aims:

1. Examine the clinical correlates of four groups defined by BMI and waist-to-height ratio.
2. Assess the prognostic value of this classification for mortality.

These are related, but they are not the same research question.

The aim contains two questions

A single paper can contain more than one research question.

First: how NHANES is organised

• Six two-year cycles, 2007–2018. Each cycle has its own file suffix: E F G H I J.
• Files are .XPT (SAS transport), read in R with haven::read_xpt().
• Each file is one row per participant for one topic — not one big table.
• SEQN is the key. Every component is joined on it.

Where the measurements and labs come from

Anthropometry, blood pressure and laboratory values are measured/assayed. Age, sex and race/ethnicity come from the demographic interview/file.

Where the questionnaire, design and outcome come from

Five NHANES traps — each one has bitten this project

• ALQ130 is drinks per drinking day, not drinks per day. Applying a “>2 a day” rule to it directly over-excluded about 2,000 people here — an average intake needs frequency × quantity.
• 777 = refused, 999 = don’t know. They are codes, not counts. Left inside a threshold rule they reclassified nine refusals as heavy drinkers. Two traps on one variable, a factor of two hundred apart.
• Do not automatically recode diastolic BP = 0 as missing. NHANES explicitly allows a diastolic value of 0; failure to obtain a reading is handled separately. Treating every 0 as missing would also be an error.
• RIDRETH3 does not exist before 2011. It is what identifies Asian participants for the lower BMI threshold, so the first two cycles cannot apply it.
• Fasting labs carry their own weight. LBXTR and LBXGLU come from the fasting subsample, which is weighted by WTSAF2YR, not WTMEC2YR.

Survey design: paper versus reproduction

NHANES is a complex, multistage probability sample.

For analyses using fasting triglycerides, CDC provides the fasting-subsample weight WTSAF2YR, together with strata and PSU variables.

Survey design: the distinction that matters

• The published paper's Statistical analysis section does not report using NHANES survey weights, strata or PSUs.
• Our reproduction pipeline includes survey-design variables and fasting-subsample weights.
• Therefore, a survey-weighted reproduction is a design-informed reanalysis, not necessarily a literal reproduction of every published analytic choice.

Population: who is being studied?

Adults with MASLD in NHANES, 2007–2018

MASLD = metabolic dysfunction-associated steatotic liver disease: excess fat in the liver together with metabolic dysfunction.

The study population is not the general population. Everyone in the main analysis already meets the study definition of MASLD.

How MASLD was actually ascertained

Hepatic steatosis was estimated using the Fatty Liver Index (Bedogni 2006), thresholded at FLI ≥ 60:

Plus at least one of five cardiometabolic criteria: adiposity, raised glucose/HbA1c or diabetes, raised blood pressure or treatment, raised triglycerides or treatment, or low HDL.

So “MASLD” here is a derived variable, not a diagnosis. No imaging, no biopsy.

Notice what the eligibility formula contains

We only name it today. You will dissect what it does to the estimate at M1.

What is the main exposure?

The exposure is not one measurement. It is built in two steps:

BMXWAIST + BMXHT → waist-to-height ratio BMXBMI → obesity category BMI category × WHtR category → four phenotype groups

Exposure component 1: BMI

• Weight and height were measured in the exam, not self-reported (BMXBMI).
• BMI marks overall body size.
• Obesity defined as BMI ≥ 30; for Asian participants BMI ≥ 25 (RIDRETH3 == 6).

BMI does not measure body fat, or where fat is stored.

Exposure component 2: waist-to-height ratio

A person 170 cm tall with a 102 cm waist has WHtR = 102/170 = 0.60.

• High central adiposity is defined as WHtR ≥ 0.60.
• WHtR is a simple marker of abdominal adiposity.
• It is not a CT or MRI measurement of visceral fat.

The exposure is the combination

Every hazard ratio in the paper is measured against Group I. A ratio without its reference group means nothing.

Exposure and comparison

There is no single “treated” group and “control” group.

• The four-category phenotype is the exposure / index factor.
• Mortality is compared across the four categories.
• In the Cox model, Group I is the reference; the hazard in Groups II, III and IV is expressed relative to it.

Outcome and follow-up

Primary outcome: all-cause mortality — death from any cause (MORTSTAT).

• Participants entered NHANES between 2007 and 2018.
• Mortality follow-up ran through 31 December 2019, via the NCHS linked mortality file — the date is stated in the NCHS 2019 Linked Mortality File linkage methodology document, not inferred from the filename.
• Follow-up time is PERMTH_EXM — months from exam to death or censoring.
• Participants had different lengths of follow-up and censoring, so time-to-event methods are appropriate. A risk difference could still be estimated at a fixed time horizon using methods that account for censoring.

Our reproduction: how 59,842 people became 6,048

What the funnel reveals

Reproducibility note: the published paper reports 6,300 participants with MASLD, whereas this pipeline yields 6,048 mortality-linked participants. The two numbers should not be presented as if they were the same cohort.

The paper also states eligibility as age >18 years, while this reproduction flow uses age ≥18. Small implementation differences like this should be recorded rather than silently harmonized.

PECOT decomposition

Reconstructed mortality question

This wording makes the population, exposure, comparison, outcome and time structure explicit — and it is now honest about how the population was defined.

What does “prognostic value” mean here?

A prognostic question asks: does knowing the phenotype tell us something about what is likely to happen later?

After multivariable adjustment: does phenotype remain associated with mortality among people who are similar on the other variables in the model?

Same variables, different targets

Same variables, different analyses

Same variables, different evidence of success

HR + CI: what it is, and is not, evidence for

HR + CI can be enough to report an adjusted association or prognostic-factor effect.

It is not evidence that a prediction model is good, and it is not by itself evidence that a causal effect has been identified.

What does this paper most directly support?

It directly supports:

• description of how phenotype groups differ clinically;
• association between phenotype and subsequent mortality;
• an adjusted prognostic association.

It does not, by itself, establish:

• that central adiposity causes the mortality difference;
• that reducing WHtR would reduce mortality;
• that adding WHtR meaningfully improves individual prediction.

Take-home sequence

1. Define the population — and state how membership was ascertained, not just what it means.
2. Define exactly how the exposure is measured — down to the variable name and the file it came from.
3. Check whether the two overlap. Here they do.
4. Define outcome and time — all-cause mortality to 2019.
5. State the research question so that all five are explicit.

Methods references for these distinctions

• Prognostic factors: Riley RD, et al. BMJ 2019;364:k4597 — adjusted prognostic effects and value over and above other prognostic factors.
• Prediction models: Moons KGM, et al. BMJ 2009;338:b375; Riley RD, et al. BMJ 2024;386:e078276 — development, validation, calibration and discrimination.
• Causal inference: Hernán MA, Robins JM. J Epidemiol Community Health 2006 — estimating causal effects from epidemiologic data; Hernán MA, Dahabreh IJ. Ann Intern Med 2025 — target-trial framing and explicit causal estimands.

Reference

Kueh MTW, Intaran MAU, Goh R, et al. Body weight categories and fat distribution in relation to all-cause mortality among adults with metabolic dysfunction-associated steatotic liver disease: a population-based analysis of NHANES, 2007–2018. BMJ Open. 2026;16:e113719.

doi: 10.1136/bmjopen-2025-113719

Cohort counts and variable names on these slides come from this repository’s own reproduction pipeline, not from the paper’s prose.