Table 2.
Threshold effects of ZJU index on hepatic steatosis and fibrosis.
| Parameter | Hepatic steatosis (CAP, dB/m) | Liver fibrosis (LSM, kPa) | P-value |
|---|---|---|---|
| Chinese MASLD cohort (n = 2,156) | |||
| Inflection point (95% CI) | 46.38 (45.50–49.44) | 49.49 (34.93–50.15) | — |
| Pre-threshold slope β (95% CI)† | 2.164 (1.734–2.595) | 0.075 (0.018–0.133) | < 0.001/0.010 |
| Post-threshold slope β (95% CI)† | 0.017 (−0.367 to 0.401) | 0.002 (−0.062 to 0.067) | 0.930/0.945 |
| Slope difference (Δβ) | 2.147 | 0.073 | < 0.001/0.035 |
| Model fit (adjusted R2) | 0.371 | 0.191 | — |
| Bootstrap success rate | 1,000/1,000 (100%) | 100/100 (100%) | — |
| US NAFLD cohort-comparison (n = 2,122)* | |||
| Inflection point | 60.56 | 51.27 | — |
| Pre-threshold slope β (95% CI)† | 2.881 (2.625–3.137) | 0.060 (0.024–0.096) | < 0.001 |
| Post-threshold slope β (95% CI) | 0.389 (−0.379 to 1.157) | 0.296 (0.249–0.344) | NS/ < 0.001 |
| Ethnic difference (Chinese vs. US) | |||
| Threshold difference (absolute) | −14.18 | −1.78 | — |
| Threshold difference (relative %) | −23.4% | −3.5% | — |
| Pre-threshold slope difference | −0.717 (−24.9%) | +0.015 (+25.0%) | — |
| Post-threshold pattern | Plateau in both | Plateau (CN) vs. continued rise (US) | — |
| Quartile analysis (fully adjusted model 3) | |||
| Q1 (< 39.1)-reference | — | — | — |
| Q2 (39.1–43.6) | +8.33 (5.25–11.40) | +0.03 (−0.42 to 0.49) | < 0.001/0.883 |
| Q3 (43.6–49.7) | +17.29 (13.29–21.28) | +0.78 (0.19–1.36) | < 0.001/0.010 |
| Q4 (≥49.7) | +17.19 (10.54–23.83) | +0.95 (−0.02 to 1.93) | < 0.001/0.056 |
| P for trend | < 0.001 | 0.007 | — |
Piecewise linear regression with bootstrap validation in Chinese MASLD patients.
CAP, controlled attenuation parameter; CI, confidence interval; CN, Chinese; LSM, liver stiffness measurement; MASLD, metabolic dysfunction-associated steatotic liver disease; NS, not significant; US, United States; ZJU, Zhejiang University index.
Statistical methods: Piecewise linear regression with automatic inflection point detection using segmented regression. All models adjusted for age, sex, education, occupation, income level, waist circumference (WC), hip circumference (HC), waist-to-hip ratio (WHR), body fat percentage (BFP), laboratory parameters [alkaline phosphatase (ALP), gamma-glutamyl transferase (GGT), total cholesterol (TC), high-density lipoprotein cholesterol (HDL-C), low-density lipoprotein cholesterol (LDL-C), uric acid (UA)], dietary habits, taste preference, exercise frequency, daily routine (Model 3 specification).
†Bootstrap confidence intervals based on 1,000 iterations with bias-corrected percentile method.
LSM analysis used 100 bootstrap iterations for computational efficiency; validation confirmed threshold stability.
Critical divergence: US cohort demonstrates continued LSM increase post-threshold (β = 0.296, P < 0.001), indicating ongoing metabolic-driven fibrogenesis. Chinese cohort shows plateau (β = 0.002, P = 0.945), suggesting earlier transition to autonomous fibrosis phase independent of ZJU.
*US data from: Luo et al. (21). Correlation between ZJU index and hepatic steatosis and liver fibrosis in American adults with NAFLD.
Complete model specifications, alternative model comparisons (linear, restricted cubic spline), detailed bootstrap validation results, and sensitivity analyses are provided in Supplementary Tables S1–S6.