A Prediction of Hepatocellular Carcinoma-Associated Cachexia: The Value of a Nomogram Based on Magnetic Resonance Imaging and Serum Characterisitcs

doi:10.21203/rs.3.rs-3785164/v1

Download PDF

Research Article

A Prediction of Hepatocellular Carcinoma-Associated Cachexia: The Value of a Nomogram Based on Magnetic Resonance Imaging and Serum Characterisitcs

https://doi.org/10.21203/rs.3.rs-3785164/v1

This work is licensed under a CC BY 4.0 License

Version 1

posted

You are reading this latest preprint version

Background and Objectives:

About 50.1% of hepatocellular carcinoma (HCC) patients are present with cachexia. Therefore, it remains urgent regarding an early detection of cachexia in HCC patients. This study reported the predictive value of a nomogram based on the clinical and magnetic resonance imaging (MRI) parameters in diagnosing cachexia in HCC patients.

Materials and Methods

A retrospective two-center study recruited the pretreatment clinical and MRI data of 489 HCC patients undergoing contrast-enhanced abdominal MRI. The data were divided into three cohorts for training and internal and external validation cohorts. Cancer-associated cachexia was initially diagnosed as weight loss > 5% over 6 months without diet control, or body mass index < 20 kg/m² and weight loss > 2%. Logistic regression analyses were performed to identify independent variables associated with cachexia in the training cohort used to build the nomogram. Area under receiver operating characteristic curve (AUC) and calibration curve were used to evaluate nomogram performance.

Results

The independent factors for HCC-associated cachexia included clinical characteristics (serum α-fetoprotein level > 100 ng/mL and serum carbohydrate antigen 19 − 9 level > 37 U/mL) and MRI findings (tumor size > 5 cm, intralesional fat, portal vein tumor thrombus, and intratumoral artery). The nomogram constructed with these variables showed good predictive performance with AUC of 0.83 (training cohort), 0.79 (internal validation cohort), and 0.85 (external validation cohort), and a satisfactory consistency between prediction and observation on the calibration curve in all cohorts.

Conclusion

The proposed multivariable nomogram suggested good performance in predicting the risk of cachexia in HCC patients.

Cachexia

Hepatocellular carcinoma

Magnetic resonance imaging

Nomogram

Imaging feature

About 50.1% of HCC patients had cachexia, which may reduce the chemotherapeutic response and indicate poor quality of life and a grave prognosis. The current study proposed a nomogram on the basis of MRI and serum characteristics for an early detection of the condition by retrospectively analyzing the MRI and clinical data of 489 HCC patients. The results showed that the proposed nomogram had a high accuracy in predicting HCC-associated cachexia, signifying the promising diagnostic value of the proposed nomogram.

Cancer-associated cachexia is defined as “a multifactorial syndrome characterized by an ongoing loss of skeletal muscle mass (with or without loss of fat mass) that can not entirely reversed by conventional nutritional support.” [1] It affects approximately 50–80% of cancer patients and is the direct cause of death in around 30% of those patients [2–4]. Cancer-associated cachexia is attributed to factors derived from the cancer cells and inflammatory and immune cells in the tumour microenvironment, which may induce an excessive catabolism of target tissues and alter the controls of energy intake and expenditure in central nervous system [2]. As cachexia is associated with reduced chemotherapeutic response, poorer quality of life, and grave prognosis [5] and, compared with patients with irreversable cachexia associated with advanced non-small cell lung cancer, those with reversable cachexia reported better therapeutic efficacy and longer survival [6], an early detection of cancer-associated cachexia promises enormous therapeutic benefits for the afflicted patients [7].

Of the various types of cancers, hepatocellular carcinoma (HCC) is one of the most frequently associated with cachexia [2]. Studies have documented cachexia in 50.1% of HCC patients [8] and more frequently in patients with advanced HCC [9]. Given that the diagnostic criterion suggests that individuals should be monitored for six months prior to a cachexia confirmation [2], it is crucial to explore a technique for an early cachexia detection in HCC patients before the diagnostic criterion.

As a noninvasive imaging modality, magnetic resonance imaging (MRI) plays a pivotal role in the diagnosis, staging, surveillance, and prognosis assessment of HCC [10]. Imaging features, including arterial phase peritumoral enhancement, absence of tumor capsule, intratumoral artery, hepatobiliary phase peritumoral hypointensity, and nonsmooth tumor margins, are potential predictors of poor clinical outcomes [11–13]. Several of these imaging characteristics can be utilized to discriminate the biological nature of HCC, including microvascular invasion, histopathologic subtype and differentiation, and molecular alterations [14–16]. Thus, it follows that these aggression-related imaging features may predict the HCC-associated cachexia.

Herein, we reported a nomograph constructed on the basis of pretreatment clinical and MRI parameters of 489 HCC patients to retrospectively detect the presence of cachexia in the patient cohort. The routine MRI scan features of HCC patients before treatment were analyzed to determine their relevance to the clinically diagnosed cancer-associated cachexia, in the hope of discovering potiental imaging indicators for cachexia prediction in HCC patients. The findings may provide some novel insights into an early cachexia detection in clinical practice.

Study Patients

This two-center retrospective study protocol was approved by the institutional review board (2018ZDSYLL162-P01) and the requirement of the informed consent was waived due to the retrospective nature of the study. The diagnosis of HCC was based on the criteria of the American Association for the Study of Liver Disease [17]. The workflow of the study design and patient selection was depicted in Fig. 1.

The inclusion criteria for the recruited data were as follows: (a) contrast-enhanced liver MRI performed within two weeks before treatment; (b) data of both clinical laboratory test and MRI; (c) images with an acceptable quality for quantitative analysis; (d) no previous treatment for HCC at the time of liver MRI scan; and (e) no comorbid diseases, such as other tumors, pancreatitis, acute cholecystitis, severe cirrhosis, severe fatty liver and massive ascites. Eligible cases were divided into three cohorts, i.e., the training cohort with patients scanned at one Institution A from May 2015 to March 2020 (n = 279); the internal validation cohort with patients scanned at the same institution from April 2020 to May 2021 (n = 120); and the external validation cohort with patients scanned in a different Institution B from January 2014 to June 2021(n = 90).

Data collection

Demographic data, including age, sex, height, and body weight at time of HCC diagnosis, were obtained from the database of clinical records. Patients were followed up for changes in body weight for six months, with body mass index (BMI) calculated for each patient. BMI was categorized according to the guidelines by the World Health Organization for Asian: underweight, below 18.5 kg/m²; normal weight, 18.5-<23 kg/m²; overweight, 23.0-<25 kg/m²; and obese, above 25 kg/m². We also collected the degree of liver dysfunction (Child-Pugh class), Barcelona Clinic Liver Cancer (BCLC) stage, and the causes of chronic liver disease. The results from laboratory tests included bilirubin, albumin, aspartate transaminase, alanine transaminase, platelet, prothrombin time, the international normalized ratio of prothrombin time, serum α-fetoprotein (AFP) level, and serum carbohydrate antigen 19 − 9 level.

The enrolled patients received different forms of treatment in their standard clinical care. For patients who underwent multiple HCC treatments, the treatments were ranked as surgical resection, local ablation (radiofrequency or microwave ablation), stereotactic body radiotherapy, transarterial chemoembolization or radioembolization, systemic therapy, best supportive care, with one preceding another in accordance with the stages of disease progression [18], which may facilitate the counting of each treatment type and exploration of the impact of treatments on this study.

According to the international consensus [1], cachexia was diagnosed when HCC patients reported a weight loss of more than 5% over six months without diet management, or a BMI of less than 20 kg/m² and a weight loss of more than 2%.

MRI Data Acquisition and Image Analysis

All patients underwent abdominal multiparametric contrast-enhanced MRI scan on a 3.0-T MR scanner (TrioTim, Siemens Medical Solutions, Erlangen, Germany; Discovery MR750, GE Healthcare, Milwaukee, USA; Ingenia, Philips Medical Systems, Best, The Netherlands; MEGNETOM Vida, Siemens Medical Solutions, Erlangen, Germany). When the contrast-enhanced MRI was perfomed, patients were intravenously injected at 1 mL/sec with either gadopentetate dimeglumine (Magnevist; Bayer) at a dose of 0.2 mL/kg (n = 249) or gadoxetic acid disodium (Primovist; Bayer) at a dose of 0.1 mL/kg (n = 240), followed by a saline wash (20 mL). MRI protocols routinely involved dual-echo spoiled gradient-echo T1-weighted in phase and out of phase imaging, T2-weighted fast spin echo imaging, diffusion-weighted imaging using b-values of 0 and 800 s/mm², and dynamic fat-suppressed 3D T1-weighted gradient-echo imaging. T1-weighted dynamic contrast-enhanced imaging was performed before and after the intravenous injection of the contrast agent. Briefly, after the contrast injection, consecutive contrast-enhanced dynamic images were acquired in the arterial phase (25–30 seconds), early portal venous phase (60–75 seconds), late portal venous phase (95–105 seconds), delayed/transitional phase (2–5 minutes), or hepatobiliary phase (15–20 minutes). The equipment, imaging sequences, and parameters used for liver MRI are summarized in Supplementary Table S1.

All liver images were read and interpreted by three experienced radiologists. The images collected from Institution A were reviewed by two independent radiologists (AA and AA) and those from Institution B by another two radiologists (AA and AA). In case of disagreement between the two radiologists, a joint review was conducted to obtain a consensus for the final report. All readers were blinded to the clinical diagnosis of cachexia and laboratory test results. The number of visible lesions, size of the lesion, and the presence or absence of the portal vein tumor thrombus, and extrahepatic metastasis were recorded for each patient.

In addition, readers assessed the presence or absence of other radiological findings of the tumors, including heterogeneity, necrosis, blood products in mass, intralesional fat, tumor capsule, hepatic capsular retraction, corona enhancement, nodule-in-nodule, mosaic architecture, intratumoral artery, nonperipheral washout, and restricted diffusion. Evaluattions were also made for the imaging findings of the signal intensity of tumors on the pre-contrast T1-weighted imaging (T1WI), T2-weighed imaging (T2WI), and those during the delayed or transitional phase, the presence or absence of ill-defined tumor margin, rim arterial phase hyperenhancement, and arterial phase peritumoral enhancement. In case of multiple lesions, only the largest one (main lesion) on the MRI scan was analyzed. The detailed descriptions of MRI findings identified and used in the study are summarized in the Supplementary information.

Statistical Analysis

All statistical analyses were performed using IBM SPSS Statistics (Version 26.0) and R software (version 4.2.2, http://www.r-project.org). The “rms” package was used to construct nomogram and calibration curve. The “pROC” and “rmda” packages were used to analyze the receiver-operator characteristic (ROC) and decision curve analysis (DCA) curves, respectively. A two-tailed P value of < 0.05 was considered statistically significant for the comparison of the area under the curve (AUC) of the ROC. Continuous variables were expressed as mean ± standard deviation or as median and interquartile range. Variables were expressed as odds ratio (OR) and 95% confidence intervals (CI).

The inter-observer agreement for each selected MRI feature or descriptor was evaluated by the κ statistic for categorical variables and intraclass correlation coefficient for continuous variables. The strength of the agreement is shown in Supplementary information.

For data from the training cohort, univariable and multivariable logistic regression analyses were performed to determine the MRI features and clinical laboratory indexes that significantly contributed to the correlation with clinically-confirmed cachexia. Variables with a P value of less than 0.10 in the univariable analysis were entered into the multivariable analysis, which produced a clinical data only model, an imaging feature model, and a combined model in a stepwise backward process. The predictive performances of the three models were assessed by the levels of discrimination and calibration. The nomogram was constructed with the inputs from the model with the best performance. The sensitivity and specificity of the models were also calculated. For the risk assessment of possible cachexia in each patient, the total points of each selected clinical and imaging feature were calculated with the logistic regression coefficients in the multivariable analysis by Eq. (1):

Points = β₀ + β₁X₁ + β₂X₂ + … + β_nX_n (1)

in which β₀ is the constant, β_n is the logistic regression coefficients, and X_n is the value of found for clinical or imaging features.

The level of discrimination by the nomogram was determined by the AUC of ROC. The quality of calibration was assessed by the calibration curve, which indicates the consistency between the predicted and observed risks of cachexia.

Stratification analysis was conducted for the type of treatment, type of contrast agents, and version of MRI system via AUC comparison by the Delong test. The DCA was used to evaluate the clinical efficacy of the obtained nomogram according to the net benefit at different threshold probabilities (0-100%), when the three models were compared with the strategy assumption of the diagnosis-all scheme (ALL) or the diagnosis-none scheme (NONE).

Patient Characteristics

The demographic information and MRI findings of HCC patients in the training and validation cohorts are summarized in Tables 1 and 2. Noticeably, among the eligible patients (410 men, 83.8% and 79 women,16.2%), a history of the Child-Pugh Class A liver diseases was found in most patients (409/489, 83.6%) and hepatitis B virus infection was the most common chronic liver disease (409/489, 83.6%). The portal vein tumor thrombus was present on MR images in 127 patients (26.0%). Extrahepatic metastasis to the lymph node or other organs was reported in 55 patients (11.2%). Restricted diffusion was evident in 370 patients (75.7%) and non-restricted diffusion in 119 patients (24.3%).

Table 1

Clinical information of the training, internal validation, external validation cohorts
Characteristics	Training cohort (n = 279)			Internal validation cohort (n = 120)			External validation cohort(n = 90)
Characteristics	(+) (n = 127)	(-) (n = 152)	P value	(+) (n = 56)	(-) (n = 64)	P value	(+) (n = 49)	(-) (n = 41)	P value
Age (years)	55.3 ± 11.0	56.1 ± 11.1	0.87	54.8 ± 13.5	59.3 ± 11.1	0.22	61.9 ± 11.2	59.4 ± 12.5	0.86
Sex (males: females)	107:20	121:31	0.32	47:9	59:5	0.25	40:9	36:5	0.56
BMI, kg/m²	24.1 ± 3.3	23.9 ± 3.1	0.88	23.5 ± 3.4	23.3 ± 3.1	0.50	24.7 ± 3.7	24.1 ± 4.6
BMI category, kg/m²			0.73			0.34			0.24
< 18.5	3 (2.4)	4 (2.6)		0	3 (4.7)		2 (4.1)	2 (4.9)
18.5-<23	45 (35.4)	62 (40.8)		28 (50.0)	27 (42.2)		14 (28.6)	20 (48.8)
23-<25	36 (28.3)	35 (23.0)		12 (21.4)	17 (26.6)		10 (20.4)	5 (12.2)
≥ 25	43 (33.9)	51 (33.6)		16 (28.6)	17 (26.6)		23 (46.9)	14 (34.1)
Etiology			0.56			0.44			0.51
Hepatitis B	113 (89.0)	131 (86.2)		47 (83.9)	58 (90.6)		42 (85.7)	38 (92.7)
Hepatitis C	2 (1.6)	7 (4.6)		2 (3.6)	0		4 (8.2)	1 (2.4)
Alcoholism	2 (1.6)	2 (1.3)		1 (1.8)	1 (1.6)		1 (2.0)	0
Unknown	10 (7.9)	12 (7.9)		6 (10.7)	5 (7.8)		2 (4.1)	2 (4.9)
Treatment			0.41			0.41			0.75
Surgical resection	57 (44.9)	60 (39.5)		25 (44.6)	23 (35.2)		14 (28.6)	11 (26.8)
TACE/TARE	49 (38.6)	68 (44.7)		21 (37.5)	23 (35.9)		29 (59.2)	27 (65.9)
Local ablation	11 (8.7)	18 (11.8)		5 (8.9)	12 (18.8)		0	0
Systemic therapy	8 (6.3)	4 (2.6)		4 (7.1)	6 (9.4)		5 (10.2)	3 (7.3)
NSC	2 (1.6)	2 (1.3)		1 (1.8)	0		1 (2.0)	0
Child-Pugh			0.68			0.08			0.68
A	116 (91.3)	134 (88.2)		43 (76.8)	57 (89.1)		31 (63.3)	28 (68.3)
B	10 (7.9)	16 (10.5)		13 (23.2)	6 (9.4)		15 (30.6)	12 (29.3)
C	1 (0.8)	2 (1.3)		0	1 (1.6)		3 (6.1)	1 (2.4)
BCLC stage			< 0.001			0.003			0.06
0 or A	43 (33.9)	101 (66.4)		17 (30.4)	37 (57.8)		20 (40.8)	25 (61.0)
B, C, or D	84 (66.1)	51 (33.6)		39 (69.6)	27 (42.2)		29 (59.2)	16 (39.0)
AST ( IU/L)	59.8 ± 48.4	52.9 ± 50.9	0.55	68.2 ± 51.0	68.2 ± 114.3	0.40	50.1 ± 45.5	73.9 ± 73.1	0.001
ALT (IU/L)	53.4 ± 49.8	47.6 ± 42.5	0.41	68.9 ± 121.9	72.2 ± 132.3	0.80	45.7 ± 46.0	65.9 ± 83.7	0.04
Albumin (g/dL)	39.2 ± 4.7	39.9 ± 5.8	0.40	38.9 ± 6.6	40.2 ± 5.1	0.04	37.0 ± 5.8	38.4 ± 5.8	0.5
Bilirubin ( mg/dL)	20.1 ± 17.0	22.8 ± 18.0	0.23	19.8 ± 9.6	20.3 ± 15.4	0.04	23.0 ± 15.1	21.6 ± 11.3	0.29
Platelet (1000/µL)	148.5 ± 94.5	131.4 ± 75.5	0.08	132.6 ± 89.3	142.4 ± 84.6	0.37	126.0 ± 57.3	138.4 ± 76.0	0.10
AFP (ng/mL)	472.2 ± 509.5	282.5 ± 451.7	0.001	407.6 ± 493.4	303.2 ± 456.5	0.15	323.3 ± 407.8	248.4 ± 423.8	0.88
CA 19 − 9 (U/mL)	55.6 ± 93.4	45.2 ± 86.7	0.85	65.1 ± 130.6	33.5 ± 31.7	0.18	65.7 ± 133.2	30.8 ± 27.8	0.18
PT (sec)	12.5 ± 2.7	12.7 ± 2.6	0.59	13.4 ± 2.5	12.7 ± 2.4	0.57	13.8 ± 5.8	13.2 ± 2.1	0.28
PT/INR	1.07 ± 0.2	1.07 ± 0.2	0.81	1.17 ± 0.2	1.11 ± 0.2	0.22	1.28 ± 0.5	1.22 ± 0.2	0.29
Note. Unless otherwise indicated, data are numbers of patients, with percentages in parentheses. AFP, α-fetoprotein; ALT, alanine transaminase; AST, aspartate transaminase; BMI, body mass index; NSC, nutritional supportive care; PT/INR, international normalized ratio of prothrombin time; PT, Prothrombin time; TACE, transarterial chemoembolization; TARE, transarterial radioembolization. (+), cachexia; (-), non-cachexia.

Table 2

MRI characteristics of training, internal validation and external validation cohorts
MRI features	Training cohort (n = 279)			Internal validation cohort (n = 120)			External validation cohort (n = 90)
MRI features	(+)	(-)	P value	(+)	(-)	P value	(+)	(-)	P value
Tumor size (cm)	7.5 ± 4.3	5.3 ± 3.4		7.6 ± 5.0	5.6 ± 4.0		6.3 ± 3.9	6.0 ± 4.2	0.94
Tumor size > 5 cm	92 (72.4)	49 (32.2)	< 0.001	36 (64.3)	25 (39.1)	0.006	29 (59.2)	16 (39.0)	0.06
Multifocality	57 (44.9)	54 (35.5)	0.11	35 (62.5)	25 (39.1)	0.01	22 (44.9)	16 (39.0)	0.57
Homogeneity	107 (84.3)	111 (73.0)	0.02	43 (76.8)	45 (70.3)	0.42	10 (20.4)	14 (34.1)	0.14
Substantial necrosis	63 (49.6)	61 (40.1)	0.11	21 (37.5)	20 (31.3)	0.47	19 (38.8)	15 (36.6)	0.83
Blood products	53 (41.7)	49 (32.2)	0.10	21 (37.5)	28 (43.8)	0.49	19 (38.8)	9 (22.0)	0.09
Intralesional fat	36 (28.3)	18 (11.8)	0.001	16 (28.6)	11 (17.2)	0.14	15 (30.6)	8 (19.5)	0.23
Pre-contrast T1WI			0.32			0.84			0.16
Iso- or hyperintense	35 (27.6)	34 (22.4)		8 (14.3)	10 (15.6)		16 (32.7)	8 (19.5)
Hypointense	92 (72.4)	118 (77.6)		48 (85.7)	54 (84.4)		33 (67.3)	33 (80.5)
T2WI			0.32			0.92			0.87
Iso- or hypointense	32 (25.2)	31 (20.4)		1 (1.8)	1 (1.6)		9 (18.4)	7 (17.1)
Mild-moderate hyperintensity	95 (74.8)	121 (79.6)		55 (98.2)	63 (98.4)		40 (81.6)	34 (82.9)
Tumor capsule			0.50			0.095			0.58
Complete	51 (40.2)	71 (46.7)		12 (21.4)	24 (37.5)		6 (12.2)	8 (19.5)
Imcomplete	53 (41.7)	54 (35.5)		36 (64.3)	29 (45.3)		27 (55.1)	19 (46.3)
None	23 (18.1)	27 (17.8)		8 (14.3)	11 (17.2)		16 (32.7)	14 (34.1)
Hepatic capsular retraction	67 (52.8)	58 (38.2)	0.02	28 (50.0)	21 (32.8)	0.06	25 (51.0)	19 (46.3)	0.66
Ill-defined tumor margin	88 (69.3)	75 (49.3)	0.001	35 (62.5)	33 (51.6)	0.23	29 (59.2)	24 (58.5)	0.95
Corona enhancement	37 (29.1)	49 (32.2)	0.58	10 (17.9)	14 (21.9)	0.58	12 (24.5)	8 (19.5)	0.57
Nodule-in-nodule	32 (25.2)	30 (19.7)	0.28	11 (19.6)	21 (32.8)	0.10	10 (20.4)	4 (9.8)	0.17
Mosaic architecture	78 (61.4)	73 (48.0)	0.03	27 (48.2)	29 (45.3)	0.75	19 (38.8)	21 (51.2)	0.24
Portal vein tumor thrombus	53 (41.7)	21 (13.8)	< 0.001	19 (33.9)	7 (10.9)	0.002	14 (28.6)	12 (29.3)	0.94
Intratumoral artery	93 (73.2)	47 (30.9)	< 0.001	37 (66.1)	25 (39.1)	0.003	30 (61.2)	15 (36.6)	0.02
Nonrim APHE	103 (81.1)	113 (74.3)	0.18	41 (73.2)	49 (76.6)	0.67	35 (71.4)	32 (78.0)	0.47
AP peritumoral enhancement	75 (59.1)	63 (41.4)	0.003	24 (42.9)	20 (31.3)	0.19	12 (24.5)	18 (43.9)	0.05
Nonperipherial washout	118 (92.9)	147 (96.7)	0.15	55 (98.2)	56 (87.5)	0.03	42 (85.7)	39 (95.1)	0.14
Delayed or transitional phase			0.41			0.03			0.79
Iso- and hyperintense	32 (25.2)	45 (29.6)		8 (14.3)	20 (31.3)		4 (8.2)	4 (9.8)
Hypointense	95 (74.8)	107 (70.4)		48 (85.7)	44 (68.8)		45 (91.8)	37 (90.2)
Extrahepatic metastasis	17 (13.4)	8 (5.3)	0.02	11 (19.6)	8 (12.5)	0.29	5 (10.2)	6 (14.6)	0.52
Restricted diffusion	90 (70.9)	114 (75.0)	0.45	47 (83.9)	49 (76.6)	0.31	38 (77.6)	32 (78.0)	0.96
Ascites	50 (39.4)	59 (38.8)	0.93	45 (80.4)	50 (78.1)	0.76	34 (69.4)	19 (46.3)	0.03
Note. Unless otherwise indicated, data are numbers of patients, with percentages in parentheses. APHE, arterial phase hyperenhancement; AP, arterial phase; T1WI, T1-weighted imaging; T2WI, T2-weighted imaging. (+), cachexia; (-), non-cachexia.

Interobserver Agreement

For the interobserver consistency for their findings in reviewing and interpreting MRI data of each case, the κ statistic value fell within the range of 0.49–0.94 from moderate to excellent in that order. An intraclass correlation coefficient within the range of 0.90–0.94 was ranked as excellent. Details of the interobserver agreements for each selected imaging finding are summarized in Supplementary Table S2.

MRI Features and Clinical Characteristics Associated with Cachexia: Univariable Analysis

In the univariable logistic regression analysis, variables with a P value of less than 0.1 were considered for an association with cachexia. The yielded clinical and MRI features that were statistically associated with cachexia included the bilirubin > 21 µmol/L (P = 0.094), the level of AFP > 100 ng/mL (P < 0.001), CA 19 − 9 > 37 U/mL (P < 0.001), tumor size > 5 cm (P < 0.001), homogeneity (P = 0.03), intralesional fat (P = 0.001), hepatic capsular retraction (P = 0.02), ill-defined tumor margin (P = 0.001), mosaic architecture (P = 0.03), portal vein tumor thrombus (P < 0.001), intratumoral artery (P < 0.001), arterial peritumoral enhancement (P = 0.004), and extrahepatic metastatic (P = 0.02) (Table S3).

Screening Variables Associated with Cachexia: Multivariable Analysis

Subsequently, the multivariable logistic regression analysis with a stepwise backward process was performed to identify the variables from the univariable analysis that might contribute to the discrimination of patients with cachexia. As a result, when only the clinical variables were evaluated for the clinical data only model, the clinical labortatory results of the AFP level > 100 ng/mL (OR: 3.69; 95% CI: 2.22–6.14; P < 0.001) and CA 19 − 9 > 37 U/mL (OR: 2.52; 95% CI: 1.51–4.19; P < 0.001) were statistically associated with cachexia (Table 3). For the imaging model, the imaging features associated with cachexia included tumor size > 5 cm (OR: 2.73; 95% CI: 1.44–5.20; P = 0.002), intralesional fat (OR: 2.37; 95% CI: 1.14–4.93; P = 0.02), mosaic architecture (OR: 0.50; 95% CI: 0.26–0.96; P = 0.04), portal vein tumor thrombus (OR: 2.58; 95% CI: 1.31–5.07; P = 0.006), and intratumoral artery (OR: 3.63; 95% CI: 1.90–6.95; P < 0.001). Meanwhile, for the combined model, both clinical variables and imaging features independently associated with cachexia were as follows: AFP > 100 ng/mL (OR: 2.76; 95% CI: 1.54–4.92; P = 0.001), CA 19 − 9 > 37 U/mL (OR: 2.58; 95% CI: 1.44–4.61; P = 0.001), tumor size > 5 cm (OR: 2.14; 95% CI: 1.10–4.12; P = 0.03), intralesional fat (OR: 2.24; 95% CI: 1.05–4.76; P = 0.04), portal vein tumor thrombus (OR: 2.13; 95% CI: 1.07–4.21; P = 0.03), and intratumoral artery (OR: 2.95; 95% CI: 1.55–5.64; P = 0.001) (Table 3).

Table 3

Combined model, clinical data only model, and imaging model associated with HCC-associated cachexia: multivariable analysis
Characteristics	Combined model		Clinical data only model		Imaging model
Characteristics	OR (95% CI)	P value	OR (95% CI)	P value	OR (95% CI)	P value
AFP > 100 ng/mL	2.76 (1.54, 4.92)	0.001	3.69 (2.23, 6.14)	< 0.001
CA 19 − 9 > 37 U/mL	2.58 (1.44, 4.61)	0.001	2.52 (1.51, 4.19)	< 0.001
Tumor size > 5 cm	2.14 (1.10, 4.15)	0.03			2.73 (1.44, 5.20)	0.002
Intralesional fat	2.24 (1.05, 4.76)	0.04			2.37 (1.14, 4.93)	0.02
Mosaic architecture	0.64 (0.32, 1.29)	0.21			0.50 (0.26, 0.96)	0.04
Portal vein tumor thrombus	2.13 (1.07, 4.21)	0.03			2.58 (1.32, 5.07)	0.006
Intratumoral artery	2.95 (1.55, 5.64)	0.001			3.63 (1.90, 6.95)	< 0.001
AFP, α-fetoprotein; OR, odds ratio; CI, confidence interval.

Development and Evaluation of the Predictive Nomogram

The details of the evaluation performance of the three models in training, internal validation, and external validation cohorts are summarized in Table S4. The ROC analysis revealed that compared with the clinical data only model and imaging model in the three cohorts, the combined model had a higher discrimination ability for cachexia prediction (Fig. 2a-c). Consequently, the nomogram was constructed on the basis of the combined model (Fig. 3). Based on the multivariable logistic regression analysis, all β_n coefficients were obtained as input parameters for Eq. (1) to calculate the points of the cachexia risk for patients, i.e, -2.435, 1.014, 0.947, 0.760, 0.805, 0.754, and 1.083 for β_0−6, respectively; for the binary value of X_1 − 6 (0 or 1), X₁ was set at 1 in case of AFP > 100 ng/mL, X₂ was at 1 in case of CA 19 − 9 > 37 U/mL, X₃ at 1 in case of a tumor size > 5 cm, X₄ at 1 for the presence of intralesional fat, X₅ at 1 for the presence of portal vein tumor thrombus, and X₆ at 1 for the presence of intratumoral artery. Additionally, the calibration curve showed a satisfactory consistency between the nomogram-predicted risks and the actual observed risks of cachexia in the three cohorts (Fig. 4a-c).

Based on the stratified analysis, the nomogram performance was not affected by the type of treatment, type of contrast agents, and version of the MRI system from the two institutions (P > 0.05 for all) (Fig. S1). The stratified analysis of treatments divided the patients into surgical resection subgroup, TACE/TARE, and other treatments (including local ablation, systemic therapy, and nutritional supportive care), with an AUC of 0.83, 0.79, and 0.84, respectively. The results showed no statistical difference when the three subgroups were compared with the all cohorts (Delong test P value of 0.60, 0.60, and 0.52, respectively). The type of the contrast agents, i.e., Magnevist and Primovist, yielded an AUC of 0.82 and 0.80, respectively, when compared with that of all cohorts (Delong test P value: 0.71 and 0.84, respectively). The version of the MRI system from the two institutions was divided into two subgroups: Trio Tim and the other version (including Discovery 750, Ingenia, and Vida), with an AUC of 0.82 and 0.76, respectively, when compared with that of all cohorts (Delong test P value of 0.75 and 0.39, respectively).

We also evaluated the benefit of the prediction nomogram using a decision curve. The decision curve analysis for the three models was presented in Fig. 5. The results revealed that compared with the clinical data only model, imaging model, the diagnosis-all scheme (ALL), and the diagnosis-none scheme (NONE), the combined model produced more net benefits for cachexia prediction in the training cohort (a threshold probability of 0–60%), the internal validation cohort (a threshold probability of 20–40%), and the external validation cohort (a threshold probability of 0–40% and 80–100%). Representative examples of HCC patients with cachexia are shown in Fig. 6 and S2.

In this retrospective study, we developed and validated a novel predictive nomogram with selected clinical and MRI parameters that were independently associated with cachexia in HCC patients. The predictive nomogram incorporated clinical and imaging predictors, including AFP > 100 ng/mL, CA 19 − 9 > 37 U/mL, tumor size > 5 cm, intralesional fat, portal vein tumor thrombus, and intratumoral artery. The nomogram displayed a promising improvement in discriminating and predicting cachexia in HCC patients with an AUC of 0.83, a sensitivity of 76% and a specificity of 79%.

The previous studies demonstrate that a tumor size of over 5 cm is associated with the existence of cachexia in HCC patients [19, 20]. These are consistent with our current study, saying tumor size over 5 cm is an predictive factor for cachexia in HCC patients. Besides, in the three cohorts of our study, cachexia was found in 67.7% of patients with a tumor size > 5 cm (157/232) and in 35.0% of patients with a tumor size < 5 cm (90/257). In addition, available evidence suggests that tumor size is related the tumor growth rate, in which a large tumor usually indicates more aggressive behavior in HCC [14].

Available literature demonstrates intralesional fat was commonly observed in the aggressive nature, i.e., poorly differentiated HCC [21], which is in agreement with our current study that there is an association between intralesional fat and an aggressive behavior, i.e., cachexia. However, other studies confirmed that intratumoral fat is associated with a lower risk for aggressive behaviors, such as microvascular invasion and metastases [22, 23]. Portal vein tumor thrombus is a form of vascular invasion with metastatic potential [24] and a poorer prognosis [25] in HCC. Besides, the prior study documented a close association between intratumoral artery and aggressive behavior, i.e., macrotrabecular-massive HCC, and poor prognostic outcome [16]. Taken together, in line with previous studies, our study confirmed that imaging features of the portal vein tumor thrombus and intratumoral artery were independent predictors of cachexia in HCC.

The incidence of cachexia is also linked to some serum biomarkers. Previous studies indicated that the elevated biomarkers of serum AFP and CA-19-9 are closely associated with aggressive biologic behaviors such as tumor recurrence and poor prognosis in cancer patients [26–28]. Similarly, in the present study, an AFP level greater than 100 ng/mL and a carbohydrate antigen 19 − 9 level greater than 37 U/mL were independently associated with cachexia. These results signify that the detection of these biomarkers is crucial for making possible a prediction for cachexia.

Several limitations remain in the current study. First, out of the initially included 2094 patients, patients with incomplete medical records (n = 1482) and HCC with co-found diseases (n = 106) were excluded from this retrospective study, thus introducing potential selection bias. Second, we included MRI parameters derived from tumors, without evaluation of the skeletal muscle and adipose tissue The quantitative assessment of these two types of tissues may increase the specificity of cachexia discrimination and prediction. For another, as the types of treatments for the patients were varied, especially for some patients receiving multiple or combined therapies, though no treatment-related effect on the findings was reported by the univariable analysis and stratified analysis in our current study, only one type of treatment option should be considered in our future study. Finally, although we included three independent cohorts from two institutions, an international multicenter study to confirm our current findings is preferable.

In conclusion, the proposed multivariable and discriminative nomogram can provide potent assistance in an early prediction and diagnosis of HCC-associated cachexia, in which the selected MRI scan features and serum biomarkers may provide important radiological and biochemical information of HCC patients. The nomogram may be of great guiding treatment decisions for an early intervention of cachexia in the management of HCC patients.

Funding statement This study has received funding by National Nature Science Foundation of China (81871412 and 82272064), Jiangsu Provincial Science and Technique Program (BK20221461), Postgraduate Research & Practice Innovation Program of Jiangsu Province (KYCX21_0159, KYCX22_0297 and KYCX23_0323).

Conflict of interest disclosure The authors declare that they have no conflict of interest.

Ethics approval and informed consent The study protocol was approved by the institutional review board (2018ZDSYLL162-P01). The requirement for the informed consent was waived based on the retrospective analysis of already collected clinical data.

Authors’ contributions Design of the study: XXL, XGP. Acquisition of data: XXL, BL, YJ. Statistical analysis: XXL, XGP. Drafting the article and revising it critically: XXL, HM, XGP. Final approval of the version to be published: all authors.

Fearon K, Strasser F, Anker SD, Bosaeus I, Bruera E, Fainsinger RL, et al. Definition and classification of cancer cachexia: an international consensus. Lancet Oncol 2011;12:489–495.
Baracos VE, Martin L, Korc M, Guttridge DC, Fearon KCH. Cancer-associated cachexia. Nat Rev Dis Primers 2018;4:17105.
Kumar NB, Kazi A, Smith T, Crocker T, Yu D, Reich RR, et al. Cancer cachexia: traditional therapies and novel molecular mechanism-based approaches to treatment. Curr Treat Options Oncol 2010;11:107–117.
Brown JL, Rosa-Caldwell ME, Lee DE, Blackwell TA, Brown LA, Perry RA, et al. Mitochondrial degeneration precedes the development of muscle atrophy in progression of cancer cachexia in tumour-bearing mice. J Cachexia Sarcopenia Muscle 2017;8:926–938.
Morimoto K, Uchino J, Yokoi T, Kijima T, Goto Y, Nakao A, et al. Impact of cancer cachexia on the therapeutic outcome of combined chemoimmunotherapy in patients with non-small cell lung cancer: a retrospective study. Oncoimmunology 2021;10:1950411.
Hakozaki T, Nolin-Lapalme A, Kogawa M, Okuma Y, Nakamura S, Moreau-Amaru D, et al. Cancer Cachexia among Patients with Advanced Non-Small-Cell Lung Cancer on Immunotherapy: An Observational Study with Exploratory Gut Microbiota Analysis. Cancers (Basel) 2022;14.
Laird B, Fallon M. Treating cancer cachexia: an evolving landscape. Ann Oncol 2017;28:2055–2056.
Anker MS, Holcomb R, Muscaritoli M, von Haehling S, Haverkamp W, Jatoi A, et al. Orphan disease status of cancer cachexia in the USA and in the European Union: a systematic review. J Cachexia Sarcopenia Muscle 2019;10:22–34.
Goh MJ, Kang W, Jeong WK, Sinn DH, Gwak GY, Paik YH, et al. Prognostic significance of cachexia index in patients with advanced hepatocellular carcinoma treated with systemic chemotherapy. Sci Rep 2022;12:7647.
Roberts LR, Sirlin CB, Zaiem F, Almasri J, Prokop LJ, Heimbach JK, et al. Imaging for the diagnosis of hepatocellular carcinoma: A systematic review and meta-analysis. Hepatology 2018;67:401–421.
Kang HJ, Kim H, Lee DH, Hur BY, Hwang YJ, Suh KS, et al. Gadoxetate-enhanced MRI Features of Proliferative Hepatocellular Carcinoma Are Prognostic after Surgery. Radiology 2021;300:572–582.
Rhee H, An C, Kim HY, Yoo JE, Park YN, Kim MJ. Hepatocellular Carcinoma with Irregular Rim-Like Arterial Phase Hyperenhancement: More Aggressive Pathologic Features. Liver cancer 2019;8:24–40.
Zhang L, Yu X, Wei W, Pan X, Lu L, Xia J, et al. Prediction of HCC microvascular invasion with gadobenate-enhanced MRI: correlation with pathology. European radiology 2020;30:5327–5336.
Yoneda N, Matsui O, Kobayashi S, Kitao A, Kozaka K, Inoue D, et al. Current status of imaging biomarkers predicting the biological nature of hepatocellular carcinoma. Jpn J Radiol 2019;37:191–208.
Lee S, Kim SH, Lee JE, Sinn DH, Park CK. Preoperative gadoxetic acid-enhanced MRI for predicting microvascular invasion in patients with single hepatocellular carcinoma. J Hepatol 2017;67:526–534.
Rhee H, Cho ES, Nahm JH, Jang M, Chung YE, Baek SE, et al. Gadoxetic acid-enhanced MRI of macrotrabecular-massive hepatocellular carcinoma and its prognostic implications. J Hepatol 2021;74:109–121.
Marrero JA, Kulik LM, Sirlin CB, Zhu AX, Finn RS, Abecassis MM, et al. Diagnosis, Staging, and Management of Hepatocellular Carcinoma: 2018 Practice Guidance by the American Association for the Study of Liver Diseases. Hepatology 2018;68:723–750.
Rich NE, Murphy CC, Yopp AC, Tiro J, Marrero JA, Singal AG. Sex disparities in presentation and prognosis of 1110 patients with hepatocellular carcinoma. Aliment Pharmacol Ther 2020;52:701–709.
Rich NE, Phen S, Desai N, Mittal S, Yopp AC, Yang JD, et al. Cachexia is Prevalent in Patients With Hepatocellular Carcinoma and Associated With Worse Prognosis. Clin Gastroenterol Hepatol 2022;20:e1157-e1169.
Imai K, Takai K, Miwa T, Taguchi D, Hanai T, Suetsugu A, et al. Rapid Depletion of Subcutaneous Adipose Tissue during Sorafenib Treatment Predicts Poor Survival in Patients with Hepatocellular Carcinoma. Cancers (Basel) 2020;12:1795.
Asayama Y, Nishie A, Ishigami K, Ushijima Y, Takayama Y, Okamoto D, et al. Fatty change in moderately and poorly differentiated hepatocellular carcinoma on MRI: a possible mechanism related to decreased arterial flow. Clinical radiology 2016;71:1277–1283.
Min JH, Kim YK, Lim S, Jeong WK, Choi D, Lee WJ. Prediction of microvascular invasion of hepatocellular carcinomas with gadoxetic acid-enhanced MR imaging: Impact of intra-tumoral fat detected on chemical-shift images. European journal of radiology 2015;84:1036–1043.
Siripongsakun S, Lee JK, Raman SS, Tong MJ, Sayre J, Lu DS. MRI detection of intratumoral fat in hepatocellular carcinoma: potential biomarker for a more favorable prognosis. AJR Am J Roentgenol 2012;199:1018–1025.
Fowler KJ, Burgoyne A, Fraum TJ, Hosseini M, Ichikawa S, Kim S, et al. Pathologic, Molecular, and Prognostic Radiologic Features of Hepatocellular Carcinoma. Radiographics: a review publication of the Radiological Society of North America, Inc 2021;41:1611–1631.
Zhang L, Zheng T, Wu Y, Wei H, Yang T, Zhu X, et al. Preoperative MRI-based multiparametric model for survival prediction in hepatocellular carcinoma patients with portal vein tumor thrombus following hepatectomy. European journal of radiology 2023;165:110895.
Lin K, Huang Q, Zeng J, Ding Z, Wang L, Chen Z, et al. Clinical Significance of Alpha-Fetoprotein in Alpha-Fetoprotein Negative Hepatocellular Carcinoma Underwent Curative Resection. Dig Dis Sci 2021;66:4545–4556.
Ding X, Sun W, Li W, Shen Y, Guo X, Teng Y, et al. Transarterial chemoembolization plus lenvatinib versus transarterial chemoembolization plus sorafenib as first-line treatment for hepatocellular carcinoma with portal vein tumor thrombus: A prospective randomized study. Cancer 2021;127:3782–3793.
Luo G, Jin K, Deng S, Cheng H, Fan Z, Gong Y, et al. Roles of CA19-9 in pancreatic cancer: Biomarker, predictor and promoter. Biochimica et biophysica acta Reviews on cancer 2021;1875:188409.

No competing interests reported.

Supplementarymaterial.docx

Download PDF

Version 1

posted

You are reading this latest preprint version

A Prediction of Hepatocellular Carcinoma-Associated Cachexia: The Value of a Nomogram Based on Magnetic Resonance Imaging and Serum Characterisitcs

Status:

Version 1

Abstract

Background and Objectives:

Materials and Methods

Results

Conclusion

Figures

Lay summary

INTRODUCTION

METHODS AND MATERIALS

Study Patients

Data collection

MRI Data Acquisition and Image Analysis

Statistical Analysis

RESULTS

Patient Characteristics

Interobserver Agreement

MRI Features and Clinical Characteristics Associated with Cachexia: Univariable Analysis

Screening Variables Associated with Cachexia: Multivariable Analysis

Development and Evaluation of the Predictive Nomogram

DISCUSSION

Declarations

References

Additional Declarations

Supplementary Files

Status:

Version 1