Can the computed tomography texture analysis of colorectal liver metastases predict the response to first-line cytotoxic chemotherapy?

Table 1 Computed tomography acquisition parameters in study cohort

CT scanner model	Number of study cases (n)	Detector collimation	Spiral pitch factor	Rotation time (s)	Voltage (kVp)	Tube current-time product at level of liver metastases (mAs)	Noise Index	Reconstruction kernel	Slice thickness/reconstruction interval (mm)	Field of view (cm)	Matrix size (pixels)
GE Lightspeed RT16	16	16 × 1.25 mm	1.375:1	0.8 (n = 14); 0.9 (n = 1); 1.0 (n = 1)	120 (n = 14); 140 (n = 2)	96-300	11.5	SOFT	1.25/1.25	50	512 × 512
GE Optima CT540	11	16 × 1.25 mm	1.375:1 (n = 9); 0.938:1 (n = 2)	0.7 (n = 6); 0.8 (n = 4); 0.9 (n = 1)	120	109.6-277.2	11.5 -13	SOFT	1.25/1.25	50	512 × 512
GE Discovery IQ	2	16 × 1.25 mm	0.938:1	0.8	120	155.2 and 209.6	11-11.5	SOFT	1.25/1.25	50	512 × 512

CT: Computed tomography.

Table 2 Characteristics of the patients, overall and by response to first-line chemotherapy

	All		Responders		Non-responders		P valuea
	n	%	n	%	n	%
Age at diagnosis (yr)1	59	(52-73)	60	(52-74)	59	(54-71)	0.79
Sex							1.00
Female	11	(38)	6	(40)	5	(36)
Male	18	(62)	9	(60)	9	(64)
Position of colorectal Tumor							0.71
Colon (incl. rectosigmoid)	18	(62)	10	(67)	8	(57)
Rectum	11	(38)	5	(33)	6	(43)
T-stage of the primary tumor							1.00
T3	22	(81)	11	(85)	11	(79)
T4	5	(19)	2	(15)	3	(21)
Unknown	2		2		0
N-stage of the primary tumor							0.85
N0	5	(22)	2	(20)	3	(23)
N1	5	(22)	3	(30)	2	(15)
N2	13	(57)	5	(50)	8	(62)
Unknown	6		5		1
Primary CRC grade							0.60
Moderate	25	(86)	12	(80)	13	(93)
Poor	4	(14)	3	(20)	1	(7)
M-stage of the primary tumor							1.00
M0	3	(10)	2	(13)	1	(7)
M1	26	(90)	13	(87)	13	(93)
KRAS mutation status							1.00
Wild type	5	(42)	2	(40)	3	(43)
Mutant	7	(58)	3	(60)	4	(57)
Unknown	17		10		7
Extent of metastatic disease							0.71
Liver only	18	(62)	10	(67)	8	(57)
Liver and extrahepatic	11	(38)	5	(33)	6	(43)
CRLM timing
Synchronous	26	(90)	13	(87)	13	(93)	1.00
Metachronous	3	(10)	2	(13)	1	(7)
Number of metastases							0.05
≤ 5	10	(34)	2	(13)	8	(57)
6-10	7	(24)	5	(33)	2	(14)
> 10	12	(41)	8	(53)	4	(29)
Maximum size of metastases (mm)							0.49
< 30	8	(28)	4	(27)	4	(29)
30-70	15	(52)	9	(60)	6	(43)
> 70	6	(21)	2	(13)	4	(29)
Target liver metastases
Baseline maximum transverse diameter (cm)1	2.7	(2.0-3.3)	2.9	(2.6-3.4)	2.4	(1.8-3.0)	0.14
Baseline lesion volume (cm³)1	7.7	(3.6-12.7)	8.3	(6.2-12.2)	5.2	(3.1-12.3)	0.32
CEA (ng/mL)1	107	(10-171)	130	(28-239)	51	(11-136)	0.24
CA19-9 (IU/mL)1,2	127	(37-377)	136	(40-327)	59	(21-773)	0.77

^aFisher’s exact test for categorical variables and Kruskal-Wallis test for continuous variables.

¹Median (interquartile range).

²Number of missing data n = 12.

CRC: Colorectal cancer; KRAS: Kirsten rat sarcoma viral oncogene homolog; CRLMs: Colorectal liver metastases; CEA: Carcinoembryonic antigen.

Table 3 Summary of Response Evaluation Criteria in Solid Tumors response and chemotherapy regimes in response and non-response group

	Response group	Non-response group
RECIST response
CR	0
PR	15
SD		7
PD		7
Chemotherapy regimen
FOLFOX	4	4
FOLFIRI	4	3
FOLFOXIRI	2	0
CAPE-OX	2	3
CAPE-IRI	2	3
Capecitabine	1	1
Number of chemotherapy cycles between baseline and follow-up scan
Range in cycles (median)	3-12 (8)	3-12 (6)
Time between baseline and follow-up scan
Range in d (median)	72-203 (141)	76-198(111)

RECIST: Response Evaluation Criteria in Solid Tumors.

Table 4 Radiomic features associated with response to chemotherapy

	Univariable models			Multiple model
	OR (95%CI)	P value1	AUC	OR (95%CI)	P value1	AUC
Minimum histogram gradient intensity	3.82 (1.26-15.3)	0.02	0.74	3.24 (1.05-12.00)	0.04	0.80
Discretized intensity skewness	0.33 (0.11-0.86)	0.02	0.73
Skewness	0.33 (0.11-0.86)	0.02	0.73
Long run low grey level emphasis	3.01 (1.16-9.26)	0.02	0.73	2.84 (0.98-10.09)	0.05
Low grey level count emphasis	3.01 (1.16-9.26)	0.02	0.73
Low grey level run emphasis	3.01 (1.16-9.26)	0.02	0.73
Volume at intensity fraction 10%	0.33 (0.11-0.86)	0.02	0.73
Short run low grey level emphasis	2.83 (1.08-8.81)	0.03	0.71

¹Likelihood ratio test. Results from univariable and multiple logistic regression models. Radiomic features were included as pseudo-continuous tertiles. OR: Odds ratio; CI: Confidence interval; AUC: Area under the receiver operating characteristic curve.