The Role of Magnetic Resonance Elastography in the Diagnosis of Noncirrhotic Portal Hypertension

Portal hypertension (PH) is defined as abnormal elevation of portal venous pressure with cirrhosis accounting for 90% of cases and 10% of cases classified as noncirrhotic PH (NCPH).^1,2 The differentiation of cirrhotic PH (CPH) from NCPH is difficult (Supplementary Figure 1), with recent research efforts focusing on noninvasive evidence of increased hepatic stiffness.^3,4 Magnetic resonance elastography (MRE) is an established imaging technique in the assessment of hepatic stiffness, and is now the most efficacious, noninvasive method to assess for hepatic fibrosis.^5–8 The aim of this study was to assess the ability of magnetic resonance imaging (MRI) and MRE to differentiate between CPH and NCPH.

Methods

We searched the institutional database for NCPH cases with histological confirmation of absence of cirrhosis including 11 patients with nodular regenerative hyperplasia reported previously.⁴ Subjects with an obvious noncirrhotic cause such as portal venous thrombosis were also included. A cohort of age- and sex-matched patients with CPH was also created. Clinical information and laboratory values were abstracted.

MRI and MRE images were attained using the standard clinical liver MRI technique described previously.⁶ Mean liver stiffness measurement (LSM) and mean splenic stiffness measurement (SSM) were calculated by drawing freehand region of interest by a single board-certified radiologist blinded to the diagnosis (Supplementary Figure 1). MRI images were reviewed by a board-certified consultant abdominal radiologist (15 years of experience) blinded to MRE data for features of cirrhosis and PH (see Supplementary Material). Statistical analysis are also detailed in supplement (see Supplementary Material)

Results

Forty-one subjects with NCPH (17 had nodular regenerative hyperplasia, 7 had portal venous thrombosis, 6 had hepatoportal sclerosis, 2 had hepatic sarcoidosis, and 9 had idiopathic PH) were compared with 41 age- and sex-matched subjects with CPH. Clinical features, MRI features, and laboratory test results are summarized in Table 1. Jaundice and variceal bleed was significantly higher in the CPH group. The mean serum aspartate aminotransferase level and total bilirubin levels were significantly higher whereas serum albumin level was significantly lower in the CPH group compared with the NCPH group.

Table 1.

Comparison of Stiffness Values, Imaging, and Clinical Features in the Diagnosis of NCPH vs CPH

	Features	NCPH (n = 41)	Cirrhosis (n = 41)	P value	Area under the curve		Odds Ratio (CI)	Sensitivity (%)	Specificity (%)	Positive likelihood ratio	Negative likelihood ratio	Positive predictive value (%)	Negative predictive value (%)
Stiffness values	Liver stiffness, kPaa	3.4 ± 1.0	8.7 ± 2.6	<.01	0.99	LSM ≤4.7	1311 (61–28,178)	95	100	—	0.05	100	95
	Splenic stiffness, kPa	8.2 ± 4.9	7.7 ± 1.8	.57	0.47	SSM ≥11.1	13.0 (1.5–110.5)	22	98	8.8	0.8	88	61
	SSM/LSMa	2.7 ± 1.7	0.9 ± 0.3	<.01	0.93	SSM/LSM ≥1.23	55 (13–239)	91	85	6.04	0.11	83	92
Imaging characteristicsb	Liver size, mm	152 ± 32.9	163 ± 32.7	.13	0.60	Size >157	0.9 (0.4–2.1)	52	46	0.97	1.04	65	33
	Surface nodularity	26 (63)	28 (68)	.64	0.47		1.2 (0.5–3.1)	68	37	1.1	0.9	52	54
	Heterogeneous parenchyma	20 (49)	28 (68)	.09	0.40		2.2 (0.9–5.3)	68	50	1.4	0.6	58	61
	Presence of fibrous bands	9 (22)	7 (17)	.65	0.52		0.8 (0.3–2.3)	18	78	0.8	1.1	44	50
	Caudate lobe hypertrophy	17 (41)	20 (49)	.51	0.47		1.3 (0.6–3.2)	49	59	1.2	0.9	54	53
	Left lateral segment hypertrophya	21 (51)	30 (73)	.04	0.41		2.6 (1.0–6.5)	73	49	1.4	0.6	59	65
	Enlarged gallbladder fossa sign	4 (10)	1 (2)	.14	0.55		0.2 (0.1–2.1)	2	90	0.2	1.1	20	47
	Enlarged periportal space	12 (29)	11 (27)	.81	0.54ee		0.9 (0.3–2.3)	27	71	0.9	1	48	49
	Posterior hepatic notch signa	0 (0)	6 (15)	<.01	0.42		8.2 (1.0–70.0)	15	100	-	0.9	100	54
	Splenomegaly	31 (76)	34 (83)	.41	0.48		1.6 (0.5–4.6)	83	24	1.1	0.7	52	59
	Esophageal varicesa	14 (34)	27 (66)	<.01	0.32		4.3 (1.7–11.1)	66	69	2.1	0.5	69	66
	Perisplenic varices	29 (71)	22 (54)	.14	0.61		0.5 (0.2–1.3)	55	29	0.8	1.5	43	40
	Portal cavernoma	2 (5)	1 (2)	.59	0.50		0.5 (0.1–5.9)	3	95	0.5	1	33	51
Laboratory values	Alanine aminotransferase, U/L	37.1 ± 19.3	66.9 ± 95.2	.07
	Aspartate aminotransferase, U/La	41.7 ± 26.2	89.2 ± 84.6	<.01
	Total bilirubin, mg/dLa	1.1 ± 1.1	2.7 ± 3.0	<.01
	Albumin, g/dLa	3.9 ± 0.5	3.6 ± 0.5	<.01
	International normalized ratio	1.24 ± 0.36	1.27 ± 0.38	.72
Clinical features and complications	Jaundicea	4 (10)	17 (41)	<.01
	Ascites	25 (61)	20 (49)	.27
	Hepatic encephalopathy	10 (24)	11 (27)	.80
	Variceal bleeda	12 (29)	4 (10)	.03
	Spontaneous bacterial peritonitis	1 (2)	2 (5)	.56

Values are mean ± SD or n (%), unless otherwise indicated.

CPH, cirrhotic portal hypertension; LSM, liver stiffness measurement; NCPH, noncirrhotic portal hypertension; SSM, splenic stiffness measurement.

^aSignificant result (P < .05)

^bVariables assessed with CPH as the positive determinant. Otherwise variables assessed with NCPH as positive.

Mean LSM was significantly higher in the CPH group than in the NCPH group (8.7 kPa vs 3.4 kPa; P < .01). With receiver-operating characteristic (ROC) analysis, an LSM cutoff value of ≤4.7 kPa had 95% sensitivity and 100% specificity for NCPH, with an area under the receiver-operating characteristic curve (AUROC) of 0.99. There were no significant differences between mean LSM of the 5 etiologies for NCPH. Mean SSM was not significantly different between the 2 groups (Table 1); however, there were significant differences in the SSM/LSM ratios. An SSM/LSM cutoff value ≥1.23 had 91% sensitivity and 85% specificity with AUROC of 0.93 for NCPH.

An LSM ≤3 kPa (standard cutoff to indicate the presence of fibrosis)⁶ would demonstrate a sensitivity of only 44% and a specificity of 100% for NCPH. In those subjects with LSM >3 kPa, an SSM/LSM cutoff >1.21 would yield 81% sensitivity, 85% specificity, and 0.86 accuracy for diagnosing NCPH.

Using a proposed algorithm of LSM ≤4.7 kPa and an SSM/LSM cutoff of >1.23 would yield 97.6% sensitivity, 100% specificity, and an AUROC of 0.99 with only 1 case of NCPH classified as CPH (Supplementary Figure 2). Comparison of ROC analysis for differentiating an LSM ≤4.7 kPa, an SSM/LSM cutoff of >1.23, and proposed combination algorithm (LSM ≤4.7 kPa or SSM/LSM cutoff >1.23 if LSM ≤4.7 kPa) showed no significant differences in the performance between an LSM ≤4.7 kPa and the proposed combined algorithm (P = .86). However, with an SSM/LSM cutoff of >1.23, performance was slightly but statistically significantly inferior to both LSM ≤4.7 kPa (P = .049) and the combination algorithm (P = “.045).

Among the MRI features, left lateral segmental hypertrophy, posterior hepatic notch sign, and esophageal varices were significantly higher in the CPH group but had poor sensitivity (Table 1).

Discussion

Our results demonstrate that the LSM is significantly lower and the SSM/LSM ratio is significantly higher in NCPH compared with CPH. An LSM ≤4.7 kPa would effectively rule out CPH. In patients with an LSM >4.7 kPa using an SSM/LSM cutoff of >1.23 resulted in only 1 false negative NCPH case and no false positive cases. These results are very promising but needs to be verified in a larger cohort comprising of as many different etiologies of NCPH as possible.

Our study has few limitations. Given its retrospective design there is inherent selection bias present with only histologically proven cases of NCPH included. There were no cases of postsinusoidal or posthepatic NCPH included in the cohort, and our proposed algorithm may not be applicable. The MRE sequence is optimized for LSM and evaluation of spleen stiffness may not be possible in all cases depending on the slice position.

In conclusion, the presence of increased LSM and increased SSM/LSM ratio are both significant features in differentiating CPH from NCPH. MRE-measured liver and spleen stiffness would be useful noninvasive parameters for differentiating NCPH from CPH.