Longitudinal computed tomography-based delta-radiomics of visceral adipose tissue predicts infliximab secondary loss of response in Crohn’s disease patients

Table 1 Characteristics of patients in the training and validation cohorts, n (%)

Variables	Total (n = 161)	Training (n = 112)	Validation (n = 49)	P value
Age (years)	29 (22, 38)	27 (21, 38)	31 (24, 38)	0.308
Male gender	129 (80.1)	90 (80.4)	39 (79.6)	0.999
BMI (kg/m²)	18.70 (16.60, 21.50)	19.10 (16.60, 21.50)	17.60 (16.60, 20.10)	0.282
Smoking	49 (30.4)	33 (29.5)	16 (32.7)	0.827
Platelet count (× 109 cells/L)	299 (299, 368)	298 (227, 377)	311 (234, 358)	0.834
Albumin (g/dL)	35.2 (32.3, 40.1)	35.2 (32.4, 39.9)	35.0 (31.7, 40.1)	0.827
ESR (mm/hour)	40.0 (22.0, 74.0)	42.0 (22.0,75.0)	37.0 (23.0,70.0)	0.887
CRP (mg/dL)	20.2 (5.0, 50.0)	21.6 (5.4, 49.9)	17.6 (5.0, 51.2)	0.779
Cr (μmol/L)	67 (59, 76)	68 (61, 77)	66 (58, 74)	0.309
NLR	4.3 (2.7, 5.8)	4.3 (2.7, 5.7)	4.0 (2.7, 6.0)	0.977
CDAI	159.0 (107.9, 234.2)	158.7 (99.5, 235.4)	159.0 (118.0, 230.0)	0.786
Montreal location classification				0.380
L1 (ileal disease)	58 (36.0)	41 (36.6)	17 (34.7)
L2 (colonic disease)	15 (9.3)	13 (11.6)	2 (4.1)
L3 (ileocolonic disease)	88 (54.7)	58 (51.8)	30 (61.2)
Montreal behaviour classification				0.554
B1 (non stricturing, non-penetrating)	72 (44.7)	47 (42.0)	25 (51.0)
B2 (stricturing)	79 (49.1)	58 (51.8)	21 (42.9)
B3 (penetrating)	10 (6.2)	7 (6.2)	3 (6.1)
Pre-IFX treatment
SAT area (cm2/m2)	20.0 (7.9, 33.4)	21.7 (8.3, 33.4)	19.2 (7.2, 33.4)	0.511
VAT area (cm2/m2)	8.1 (3.2, 25.9)	7.4 (3.2, 24.5)	9.7 (3.1, 28.0)	0.790
SAT volume (cm3/m3)	423.2 (166.0, 773.3)	438.2 (170.7, 758.5)	359.8 (155.1, 806.2)	0.901
VAT volume (cm3/m3)	187.3 (83.1, 421.4)	169.6 (79.15, 410.3)	214.7 (93.6, 468.1)	0.866
VAT/SAT area ratio (≥ 1)	45 (28.0)	28 (25.0)	17 (34.7)	0.210
VAT/SAT volume ratio (≥ 1)	27 (16.8)	19 (17.0)	8 (16.3)	0.921
Post-IFX treatment
SAT area (cm2/m2)	22.7 (12.9, 35.6)	23.9 (13.7, 39.7)	19.8 (12.1, 31.7)	0.068
VAT area (cm2/m2)	11.0 (5.2, 24.9)	11.7 (6.1, 26.7)	9.8 (4.2, 20.3)	0.219
SAT volume (cm3/m3)	310.5 (166.1, 596.4)	321.5 (181.8, 627.6)	297.6 (156.8, 535.2)	0.368
VAT volume (cm3/m3)	436.2 (244.5, 759.9)	469.1 (272.9, 824.0)	359.6 (214.8, 626.8)	0.131
VAT/SAT area ratio (≥ 1)	37 (23.0)	25 (22.3)	12 (24.5)	0.765
VAT/SAT volume ratio (≥ 1)	18 (11.2)	12 (10.7)	6 (12.2)	0.778

Data are expressed as median (interquartile range), (SD) or number (percentage). BMI: Body mass index; ESR: Erythrocyte sedimentation rate; CRP: C-reactive protein; Cr: Creatinine; NLR: Neutrophil-lymphocyte ratio; CDAI: Crohn’s disease activity index; IFX: Infliximab; SAT: Subcutaneous adipose tissue; VAT: Visceral adipose tissue.

Table 2 Univariable and multivariable logistic regression analysis for prediction of secondary loss of response to infliximab in Crohn’s disease patients in the training cohort, n (%)

Variables	Non-SLR (n = 74)	SLR (n = 38)	Univariable analysis		Multivariable analysis
			OR (95CI)	P value	OR (95%CI)	P value
Age (years)	27 (21, 38)	27 (21, 38)	1.004 (0.677-1.489)	0.990
Male gender	61 (82.4)	29 (76.3)	1.456 (0.323-6.573)	0.603
BMI (kg/m2)	19.25 (16.62, 21.10)	18.95 (16.65, 23.02)	0.968 (0.292-3.210)	0.794
Smoking	23 (31.1)	10 (26.3)	1.263 (0.453-3.521)	0.760
Platelet count (× 109 cells/L)	291 (197, 368)	323 (272, 406)	0.997 (0.021-46.938)	0.049	1.005 (1.001-1.008)	0.011
Albumin (g/dL)	35.7 (32.7, 40.3)	34.3 (32.3, 39.7)	1.020 (0.289-3.593)	0.480
ESR (mm/hour)	39.5 (24.0, 74.8)	47.0 (18.3, 75.5)	1.000 (0.972-1.029)	0.703
CRP (mg/dL)	22.1 (6.7, 51.2)	16.1 (5.0, 44.4)	1.006 (0.116-8.735)	0.296
Cr (μmol/L)	69 (61,78)	65 (58, 74)	1.020 (0.078-13.332)	0.167
NLR	4.3 (2.7, 5.7)	4.3 (3.1, 5.8)	1.033 (0.360-2.964)	0.825
CDAI	156.5 (98.5, 242.4)	166.4 (124.8, 233.5)	1.001 (0.590-1.695)	0.890
Montreal location classification			0.786 (0.090-6.874)	0.401
L1 (ileal disease)	29 (39.2)	12 (31.5)
L2 (colonic disease)	8 (10.8)	5 (13.2)
L3 (ileocolonic disease)	37 (50.5)	21 (55.3)
Montreal behaviour classification			0.472 (0.007-32.379)	0.031	2.307 (1.108-4.803)	0.025
B1 (non stricturing, non-penetrating)	36 (48.6)	11 (29.0)
B2 (stricturing)	35 (47.3)	23 (60.5)
B3 (penetrating)	3 (4.1)	4 (10.5)
Pre-IFX treatment
SAT area (cm2/m2)	21.2 (7.5, 33.4)	21.7 (9.2, 33.4)	0.999 (0.873-1.143)	0.98
VAT area (cm2/m2)	6.9 (3.0, 21.6)	9.7 (3.6, 26.9)	0.994 (0.320-3.088)	0.405
SAT volume (cm3/m3)	453.3 (165.2, 761.2)	409.8 (179.3, 745.3)	1.000 (0.840-1.190)	0.815
VAT volume (cm3/m3)	168.5 (75.5, 394.8)	234.1 (97.3, 415.2)	0.999 (0.210-4.761)	0.424
VAT/SAT area ratio (≥ 1)	17 (57.0)	11 (28.9)	1.366 (0.353-5.282)	0.491
VAT/SAT volume ratio (≥ 1)	8 (10.8)	11 (28.9)	3.361 (0.034-330.690)	0.016	3.977 (1.320-11.980)	0.014
Post-IFX treatment
SAT area (cm2/m2)	26.4 (14.3, 39.8)	20.2 (12.9, 39.3)	1.007 (0.272-3.731)	0.386
VAT area (cm2/m2)	11.8 (6.1, 28.6)	11.1 (6.2, 23.0)	1.009 (0.222-4.577)	0.681
SAT volume (cm3/m3)	481.9 (289.0, 782.3)	408.1 (271.2, 836.1)	1.000 (0.226-4.427)	0.435
VAT volume (cm3/m3)	319.2 (187.9, 624.6)	434.9 (170.6, 612.7)	1.000 (0.544-1.838)	0.975
VAT/SAT area ratio (≥ 1)	18 (24.3)	7 (18.4)	0.703 (0.175-2.815)	0.479
VAT/SAT volume ratio (≥ 1)	6 (8.1)	6 (15.8)	2.125 (0.193-23.398)	0.215

Data are expressed as median (interquartile range), (SD) or number (percentage). IFX: Infliximab; SLR: Secondary loss of response; BMI: Body mass index; ESR: Erythrocyte sedimentation rate; CRP: C-reactive protein; Cr: Creatinine; NLR: Neutrophil-lymphocyte ratio; CDAI: Crohn’s disease activity index; SAT: Subcutaneous adipose tissue; VAT: Visceral adipose tissue; OR: Odds ratio.

Table 3 Performance of models for predicting secondary loss of response in the training and validation cohorts

	AUC (95%CI)	Sensitivity	Specificity	PPV	NPV	P alue1	Accuracy
Training cohort (n = 112)
Clinical model	0.767 (0.678-0.857)	85.1% (63/74)	36.8% (14/38)	72.4% (63/87)	56.0% (14/25)	-	68.8% (77/112)
Radiomics model	0.816 (0.737-0.896)	86.5% (64/74)	47.4% (18/38)	76.2% (64/84)	64.3% (18/28)	0.133	73.2% (82/112)
Combined model	0.853 (0.786-0.921)	85.1% (63/74)	52.6% (20/38)	77.8% (63/81)	64.5% (20/31)	0.023	74.1% (83/112)
Validation cohort (n = 49)
Clinical model	0.707 (0.539-0.876)	96.97% (32/33)	31.3% (5/16)	74.4% (32/43)	83.3% (5/6)	-	79.9% (37/49)
Radiomics model	0.750 (0.605-0.895)	84.8% (28/33)	50.0% (8/16)	77.8% (28/36)	61.5% (8/13)	0.204	73.5% (36/49)
Combined model	0.812 (0.677-0.948)	93.9% (31/33)	50.0% (8/16)	79.5% (31/39)	80.0% (8/10)	0.048	79.6% (39/49)

¹P value indicates the significance level of the comparison of areas under the curves with the clinical model as the reference using DeLong test in the training and validation cohort, respectively.

SLR: Secondary loss of response AUC: Area under the curve NPV: Negative predictive value PPV: Positive predictive value.