Automated AI-enabled CT Biomarkers for Population-Based Opportunistic Detection of Patients at Risk for Unsuspected Advanced Hepatic Fibrosis and Cirrhosis

Statistical analysis

Categorical variables were summarized as counts and percentages, and continuous variables were summarized as median (interquartile range) and range. The Kruskal-Wallis rank sum test was used to compare continuous variables across groups. The Spearman rank correlation coefficient was calculated to assess the correlation between FIB-4 and age. The significance level was set at 0.05. All statistical analyses were performed using R version 4.2.2 (R Foundation for Statistical Computing, Vienna, Austria).

Patient characteristics

Patient characteristics are summarized in Table 1.

Table 1: Patient characteristics

Characteristic	Overall, N = 31,2891	F, N = 16,6261	M, N = 14,6631
Age (years)	52 (45, 58)	52 (44, 58)	53 (46, 59)
FIB-4	0.94 (0.65, 1.48)	0.86 (0.60, 1.28)	1.05 (0.72, 1.74)
FIB-4 Category
Low Risk (< 1.3)	21,771 (70%)	12,572 (76%)	9,199 (63%)
Intermediate Risk (1.3-2.67)	5,759 (18%)	2,563 (15%)	3,196 (22%)
High Risk (≥2.67)	3,759 (12%)	1,491 (9.0%)	2,268 (15%)
LSVR	0.30 (0.25, 0.36)	0.31 (0.26, 0.37)	0.28 (0.23, 0.34)
LSVR Category
LSVR<.347	22,212 (71%)	10,999 (66%)	11,213 (76%)
LSVR≥.347	9,077 (29%)	5,627 (34%)	3,450 (24%)
Spleen Volume (mL)	247 (170, 365)	209 (150, 300)	299 (208, 438)
Spleen Volume Category
SV<315 mL	20,779 (66%)	12,921 (78%)	7,858 (54%)
SV≥315 mL	10,510 (34%)	3,705 (22%)	6,805 (46%)
1 Median (IQR); n (%)

FIB-4 distribution across LSVR and SV groups

Table 2 and Figure 1 show the distribution of FIB-4 score across LSVR and SV groups by sex.

Table 2: FIB-4 distribution across LSVR and SV groups by sex

Characteristic	LSVR<.347, SV<315 mL, N = 14,8761	LSVR≥.347, SV<315 mL, N = 5,9031	LSVR<.347, SV≥315 mL, N = 7,3361	LSVR≥.347, SV≥315 mL, N = 3,1741	p-value2
Overall	0.86 (0.61, 1.22)	0.89 (0.62, 1.28)	1.07 (0.70, 1.83)	2.08 (0.94, 5.21)	<0.001
1 Median (IQR)
2 Kruskal-Wallis rank sum test

Characteristic	LSVR<.347, SV<315 mL, N = 8,6631	LSVR≥.347, SV<315 mL, N = 4,2581	LSVR<.347, SV≥315 mL, N = 2,3361	LSVR≥.347, SV≥315 mL, N = 1,3691	p-value2
Female	0.82 (0.58, 1.15)	0.84 (0.59, 1.21)	0.97 (0.62, 1.67)	1.64 (0.80, 4.66)	<0.001
1 Median (IQR)
2 Kruskal-Wallis rank sum test

Characteristic	LSVR<.347, SV<315 mL, N = 6,2131	LSVR≥.347, SV<315 mL, N = 1,6451	LSVR<.347, SV≥315 mL, N = 5,0001	LSVR≥.347, SV≥315 mL, N = 1,8051	p-value2
Male	0.93 (0.66, 1.32)	1.02 (0.72, 1.58)	1.11 (0.75, 1.91)	2.47 (1.08, 5.61)	<0.001
1 Median (IQR)
2 Kruskal-Wallis rank sum test

Figure 1: Boxplots of FIB-4 distribution across LSVR and SV groups by sex

FIB-4 vs age

Relationship between FIB-4 and age is plotted in Figure 2, which shows a general correlation in each sex and overall.

Figure 2: Correlation between FIB-4 and age by sex

Time between CT and diagnosis

The cohort included 7913 patients with diagnosed liver disease. Their time interval between CT scan and diagnosis of liver disease ranged from -22.05 years to 32.76 years, with a median of 0.02 years.

Exploratory analyses show:

• No clear variations in markers (FIB-4, LSVR, SV) among those with negative CT-diagnosis time;
• Minor differences between time \(0-0.4\) years and \(>0.4\) years groups, depending on the marker.

Table 3: Biomarker values by time interval categories

Characteristic	< 0 years, N = 2,3551	0 to 0.4 years, N = 2,7731	> 0.4 years, N = 2,7851	p-value2
fib4	0.9 (0.6, 1.6)	2.3 (0.9, 5.7)	1.6 (0.9, 4.1)	<0.001
LSVR	0.30 (0.24, 0.36)	0.32 (0.25, 0.42)	0.32 (0.25, 0.41)	<0.001
Unknown	118	162	152
SV	271 (187, 393)	406 (243, 677)	388 (239, 651)	<0.001
Unknown	188	140	209
1 Median (IQR)
2 Kruskal-Wallis rank sum test

Figure 3: Boxplots of biomarker values by time interval categories