WHtR captures hidden fatty liver risk in adults below BMI 24 kg/m²
Waist-to-height ratio WHtR exposed a concentrated burden of ultrasound-detected fatty liver in adults whose BMI remained below the Chinese overweight threshold, but the measure did not meaningfully sharpen risk prediction once BMI and the triglyceride–glucose index TyG were already in the model. The distinction is clinically familiar: central adiposity can coexist with a body size that looks unremarkable on paper, yet still track with ectopic fat deposition in the liver and insulin resistance in metabolic tissues.
Ultrasound-detected fatty liver was present in 839 of 3,469 adults, a prevalence of 24.2%. Among the 1,933 participants with BMI below 24 kg/m², those with WHtR at or above 0.5 represented 11.0% of that BMI stratum but accounted for 31.4% of its fatty liver cases. In practical terms, a relatively small subgroup carried a disproportionately large share of disease. Their fatty liver prevalence reached 23.0%, compared with 6.2% in people with both low BMI and low WHtR.
Central adiposity carried more signal than weight alone
The pattern fits the biology of fatty liver better than BMI alone ever can. BMI is a blunt measure of overall mass, while WHtR is a rougher but more direct proxy for abdominal fat distribution. Visceral and truncal adiposity are linked to increased free fatty acid flux to the liver, hepatic insulin resistance, and intrahepatic lipid accumulation. A person can remain below an overweight BMI threshold and still carry the metabolic burden that matters most for steatosis.
After adjustment for age, sex, and TyG, the low-BMI/high-WHtR phenotype carried a prevalence ratio of 2.63, with a 95% confidence interval of 1.90 to 3.63. That separation is large enough to matter at the bedside, especially in routine health examinations where a normal-looking BMI can falsely reassure clinicians and patients. The study also showed that prevalence rose gradually across the joint BMI-WHtR distribution, with no abrupt step at BMI 24 kg/m² or WHtR 0.5. Disease risk behaved like a continuum rather than a cliff.
TyG improved discrimination; WHtR added almost nothing on top
The incremental modeling results were much less dramatic. Adding TyG to BMI increased the cross-validated ROC AUC from 0.837 to 0.865. Adding WHtR on top of BMI plus TyG increased the ROC AUC only to 0.867, with a paired difference of 0.00163 and a confidence interval that crossed zero at the lower end. Precision-recall AUC and Brier score changed by essentially nothing in a clinically persuasive direction.
The message is straightforward: WHtR identified a phenotype with higher disease prevalence, but it did not materially improve individual-level classification once metabolic information was already available. In diagnostic terms, WHtR helped describe who was at risk, but it did not meaningfully refine who should be labeled high risk after BMI and TyG had already spoken. For screening systems that prize simplicity, that matters.
Why the liver signal appears before BMI crosses the overweight line
The data align with a basic metabolic sequence. Fat accumulation in the abdomen tends to precede overt weight-based obesity in many adults, especially in populations where body habitus and metabolic risk do not map neatly onto BMI thresholds. Waist circumference relative to height may therefore expose early central adiposity before conventional weight cutoffs do. Once metabolic dysfunction is already summarized by TyG, however, WHtR may be redundant rather than additive.
That redundancy is exactly what the prediction metrics suggest. The cross-validated Brier score changed by -0.00053 with WHtR added to BMI plus TyG, and the precision-recall AUC difference was -0.00008, both too small to support a meaningful improvement in practical prediction. The composite TyG-derived anthropometric indices also failed to show a clear overall advantage over the simpler BMI-plus-TyG model.
County hospital screening may benefit from phenotype recognition, not a new cutoff
The setting matters. These were retrospective routine health-examination data from a single county hospital, the kind of clinical environment where low-cost measurement has real operational value and where liver ultrasound cannot be deployed indiscriminately. Waist and height are easy to obtain, inexpensive, and familiar. They may help clinicians recognize a central-adiposity phenotype that deserves closer metabolic attention even when BMI appears reassuring.
What the data do not support is a stand-alone WHtR threshold as a trigger for liver ultrasound referral. A binary cutoff of 0.5 would miss the continuous way fatty liver risk accumulates across BMI and waist-to-height combinations, and it would not outperform a model that already includes BMI and TyG. In other words, WHtR is useful as a descriptive lens, not as a replacement rule.
The study’s limitations are concrete. It was cross-sectional, so it cannot determine whether WHtR predicts incident fatty liver or progression over time. It relied on ultrasound rather than histology, which is routine and pragmatic but less specific than biopsy for characterizing hepatic fat and inflammation. It also came from one center, so performance may differ in other populations, care settings, and ethnic groups.
Even with those constraints, the clinical pattern is hard to ignore. Abdominal adiposity can hide behind a non-overweight BMI, and fatty liver can already be present when body weight alone still looks ordinary. The question for future work is not whether waist-to-height ratio can identify that phenotype; it clearly can. The more useful question is whether that extra phenotype labeling changes management enough to justify adding another routine measurement when BMI and TyG are already in hand.