Results
Baseline characteristics and liver-related measurements by quartiles of the study group are presented in table 1. Higher quartiles of LAI were more often male and of higher BMI. Current alcohol usage and drinks per week were consistent across quartiles of LAI, except the lowest quartile, which had one alcoholic drink per week, compared with two drinks per week for the remaining quartiles. An average of 221.7±164.9 cubic centimeters (CC) of liver was analyzed in our study. On average, males had higher liver volume in the scanned image, higher LAI, and higher US-FLI scores than females (p<0.0001). Hispanic participants had higher LAI scores than non-Hispanic White, Chinese, and non-Hispanic Black participants (p<0.001) (figure 2). In univariate analysis, cases with incident CVD events had higher LAI scores than those who did not (42.5% (95% CI 41.7% to 43.4%) vs 40% (39.8% to 40.5%), respectively, p<0.0001). These differences remained statistically significant after adjustment for BMI (p<0.0001) and all CVD risk factors (p=0.03). Additionally, individuals with type 2 diabetes at baseline had higher LAI scores than those without (47.5% vs 39.4%, p<0.0001) (figure 3).
Liver attenuation index by race/ethnicity. HU, Hounsfield units.
Liver attenuation index in individuals with diabetes versus without diabetes at baseline examination. HU, Hounsfield units.
•
Baseline characteristics of MESA-1 participants overall and by quartiles of LAI
LAI scores had moderate correlation with log-transformed US FLI scores (R=0.57, p<0.0001) (figure 4). Among CVD risk factors, LAI scores had moderate correlation with BMI (R=0.56, p<0.0001) and insulin (R=0.47, p<0.0001), and weak correlation with HDL (r=−0.30, p<0.0001), glucose (R=0.31, p<0.0001), triglycerides (R=0.24, p<0.0001), and IL-6 (R=0.28, p<0.0001. Little correlation was seen with age (R=−0.03, p=0.0128).
Spearman’s correlation between artificial intelligence-cardiovascular disease (AI-CVD) liver measurements and CVD risk factors. BMI, body mass index; BP, blood pressure; HDL, high-density lipoprotein; HU, Hounsfield units; LAI, liver attenuation index; LDL, low-density lipoprotein.
Over a mean follow-up of 12.4 years, 751 total CVD events occurred (1.1 per 100 person-years), including 499 CHD events (0.7 per 100 person-years), 236 stroke events (0.3 per 100 person-years), and 1343 all-cause mortality events (1.9 per 100 person-years). Of these, 457 (60.9%) CVD events occurred in males and 294 (39.1%) in females, 330 (66.1%) CHD events occurred in males and 169 (33.9%) in females, 122 (51.7%) stroke events occurred in males and 112 (48.3%) in females, and 747 (55.6%) mortality events in males and 596 (44.4%) mortality events in females. Participants in the highest LAI quartile (>50%) had significantly higher incident CVD over 15 years than those in the lowest quartile (<30%): 19.2% (95% CI 17.1% to 21.5%) vs 11.7% (9.5% to 14.3%), respectively, p<0.0001 (figure 5a). These differences remained even after adjustment for BMI (figure 5b) and diabetes (figure 5c). Participants in the highest quartile of both LAI and CAC score (n=386) experienced 37.3% (32.3% to 42.7%) incidence of all CVD over 15 years (figure 6).
(a–c) 15-year cardiovascular disease (CVD) incidence by liver attenuation index (LAI), adjusted for body mass index (BMI), diabetes, interleukin (IL)-6, and insulin.
15-year cardiovascular disease (CVD) incidence by liver attenuation index (LAI) and Agatston coronary artery calcium (CAC) score.
Analysis of quartile-based HRs
In fully adjusted models, individuals in the highest quartile of LAI (Q4) had higher risks of CVD (HR 1.43, 95% CI 1.08 to 1.89), stroke (HR 1.77, 95% CI 1.09 to 2.88), and all-cause mortality (HR 1.36, 95% CI 1.10 to 1.67) compared with those in the lowest quartile (Q1), independent of CVD risk factors and the Agatston CAC score. However, the association with CHD was not statistically significant (HR 1.29, 95% CI 0.91 to 1.83) (table 2).
•
Minimally and fully adjusted HRs of incident CVD by the fourth quartile (Q4) vs first quartile (Q1) of LAI
In BMI-adjusted models, females in the highest quartile of LAI (Q4) compared with the lowest quartile (Q1) showed a higher risk of CVD (HR 1.42, 95% CI 0.93 to 2.16) and CHD (HR 1.20, 95% CI 0.66 to 2.17), although these associations were not statistically significant (table 2). The association between LAI Q4 and all-cause mortality showed significantly increased risk compared with Q1 (HR 1.51, 95% CI 1.12 to 2.03). Females in the third quartile (Q3) also exhibited a significantly elevated risk of mortality compared with Q1 (HR 1.44, 95% CI 1.12 to 1.86) (online supplemental table 1). After further adjustment by all CVD risk factors and Agatston CAC score, the risk of mortality remained elevated for women in LAI Q3 vs Q1 (HR 1.34, 95% CI 1.04 to 1.74). However, the association with CVD for females in Q4 vs Q1 remained insignificant (HR 1.18, 95% CI 0.76 to 1.82) along with Q3 vs Q1 and Q2 vs Q1.
In BMI-adjusted models, males in the highest quartile of LAI (Q4) had a significantly increased risk of CVD (HR 1.54, 95% CI 1.08 to 2.20) and stroke (HR 2.59, 95% CI 1.37 to 4.90) compared with Q1 (table 2). Elevated risk of mortality was observed (HR 1.29, 95% CI 0.99 to 1.68) but was not statistically significant, as the CI crossed 1. The third quartile (Q3) also showed elevated risks for CVD (HR 1.46, 95% CI 1.08 to 1.99), stroke (HR 2.11, 95% CI 1.21 to 3.67), and mortality (HR 1.41, 95% CI 1.12 to 1.78) (online supplemental table 2). In fully adjusted models, the highest quartile of LAI in males remained significantly associated with stroke risk (HR 2.05, 95% CI 1.09 to 3.86) compared with Q1 and Q3 showed a borderline significant stroke risk compared with Q1 (HR 1.76, 95% CI 1.00 to 3.09). The association with mortality was no longer significant (HR 1.07, 95% CI 0.82 to 1.41).
Analysis of continuous HRs
In males, each SD increase in LAI was associated with an HR of 1.14 (95% CI 1.01 to 1.28) for CVD events, independent of CVD risk factors plus Agatston CAC score. Each SD decrease of liver mean attenuation was associated with similar, but slightly lower risk (HR 0.89, 95% CI 0.78 to 1.0). For stroke events, the HR per SD increase of LAI was 1.31 (95% CI 1.04 to 1.66), independent of CVD risk factors plus Agatston CAC score. Each SD decrease in liver mean attenuation was not significant for stroke events (HR 0.81, 95% CI 0.64 to 1.02, p=0.16). In females, the HR per SD increase of LAI for mortality was 1.16 (95% CI 1.04 to 1.29) after adjustment for other CVD risk factors plus CAC score. Each SD decrease of liver mean attenuation was associated with similar risk (HR 0.83, 95% CI 0.72 to 0.95). No significant association with incident CVD was observed in females after adjustment for CVD risk factors plus Agatston CAC score.
Among non-Hispanic White individuals, each SD increase in LAI was significantly associated with CVD events (HR 1.17, 95% CI 1.0 to 1.36) independently of other CVD risk factors plus CAC score. In non-Hispanic Black individuals, a significant association was observed for overall mortality (HR 1.17, 95% CI 1.03 to 1.33); similarly, in Hispanic individuals, LAI was significantly associated with higher overall mortality (HR 1.23, 95% CI 1.04 to 1.46). No significant associations were found in the Chinese subgroup independent of CVD risk factors plus CAC score.
Leave a Reply