Serum biomarkers of liver fibrosis identify globus pallidus vulnerability

Highlights

• •Globus pallidus hyperintensities are often present in patients with liver cirrhosis.
• •Here, serum-derived, liver fibrosis markers were calculated in a large (n = 457) sample.
• •Serum fibrosis cutoff scores identified globus pallidus vulnerability.
• •Higher pallidal signal intensities correlated with sensory and motor impairment.
• •First time pallidal integrity is reported as affected in pre-cirrhotic liver disease.

Abstract

The CNS manifestation of chronic liver disease can include magnetic resonance (MR) signal hyperintensities in basal ganglia structures. Here, relations between liver (serum-derived fibrosis scores) and brain (regional T1-weighted signal intensities and volumes) integrity were evaluated in a sample of 457 individuals including those with alcohol use disorders (AUD), people living with human immunodeficiency virus (HIV), those comorbid for AUD and HIV, and healthy controls. Liver fibrosis was identified from cutoff scores as follows: aspartate aminotransferase to platelet ratio index (APRI) > 0.7 in 9.4% (n = 43) of the cohort; fibrosis score (FIB4) > 1.5 in 28.0% (n = 128) of the cohort; and non-alcoholic fatty liver disease fibrosis score (NFS) > -1.4 in 30.2% (n = 138) of the cohort. Presence of serum-derived liver fibrosis was associated with high signal intensities selective to basal ganglia (i.e., caudate, putamen, and pallidum) structures. High signal intensities in the pallidum, however, explained a significant portion of the variance in APRI (25.0%) and FIB4 (23.6%) cutoff scores. Further, among the regions evaluated, only the globus pallidus showed a correlation between greater signal intensity and smaller volume (r = -0.44, p <.0001). Finally, higher pallidal signal intensity correlated worse ataxia (eyes open ρ = -0.23, p =.0002; eyes closed ρ = -0.21, p =.0005). This study suggests that clinically relevant serum biomarkers of liver fibrosis such as the APRI may identify individuals vulnerable to globus pallidus pathology and contribute to problems with postural balance.

1. Introduction

Hepatic encephalopathy, the most widely recognized neurological complication of serious liver failure, presents acutely with a wide range of clinical manifestations including drowsiness, depression, confusion, tremor, or coma (Butterworth et al., 2009, Cheon and Song, 2021, Poordad, 2007) and upon magnetic resonance imaging (MRI) examination as diffuse edema generally considered to be due to CNS ammonia accumulation (Cudalbu and Taylor-Robinson, 2019, Grover et al., 2006, Prakash and Mullen, 2010). MRI in chronic liver disease, however, can also reveal bilateral, symmetric, high-intensity T1-weighted signals in basal ganglia structures, prominent in globus pallidus (Awada et al., 1995, Krieger et al., 1996, Taylor-Robinson et al., 1995) which likely represents a separate clinical entity than hepatic encephalopathy (Fukuzawa et al., 2006, Sparacia et al., 2022, Weissenborn et al., 2004, Weissenborn et al., 1995). Although often recognized on non-quantitative radiological readings, a “pallidal index”, the signal intensity ratio of the globus pallidus relative to another region – typically subcortical frontal white matter – has been used to describe signal intensity abnormalities associated with chronic liver disease (Kim et al., 1999, Krieger et al., 1995). The presumed pathophysiology of these basal ganglia hyperintensities in chronic liver disease is manganese deposition (Balachandran et al., 2020, Butterworth, 2003, Klos et al., 2006, Spahr et al., 1996; but see Maffeo et al., 2014), which may promote neuronal loss (Butterworth et al., 1995, Jayakumar et al., 2004). High pallidal signal intensities have been associated with postural tremor of the hands (Pujol et al., 1993); a high pallidal index has been associated with excessive postural body sway (Kim et al., 2007), slowed finger tapping (Kim et al., 2005, Shin et al., 2007), and impaired grooved pegboard (Chang et al., 2009, Chang et al., 2010) performance, but a consistent neurological consequence of T1-signal alterations in chronic liver disease has not been forthcoming.

Among clinical diagnoses that can increase the risk for liver disease are alcohol use disorders (AUD) and infection with the human immunodeficiency virus (HIV). Alcohol is rapidly overtaking viral hepatitis (e.g., hepatitis C virus, HCV) as a primary, modifiable, chronic liver disease risk factor (Åberg et al., 2022, Paik et al., 2020, Sepanlou et al., 2020, Vento and Cainelli, 2022). When considered as a primary cause and cofactor, alcohol accounts for upwards of 30% of cirrhosis-related deaths globally (Moon et al., 2020, Stein et al., 2016). The risk of developing cirrhosis appears to be a function of both quantity and duration of alcohol consumption (Bellentani and Tiribelli, 2001, Lelbach, 1974), although the prevalence of significant liver damage has recently increased in young adult drinkers (Kim et al., 2018, Tapper and Parikh, 2018, Yoon and Chen, 2018). Similarly, advanced liver disease remains a leading cause of death among people living with HIV (Sherman et al., 2017). Liver disease in HIV results from a complex pathology; HCV co-infections (Pokorska-Śpiewak et al., 2017, Weber et al., 2006) and AUD comorbidity (Justice et al., 2006, Muga et al., 2012), however, contribute significantly to liver disease incidence among those with HIV infections (Bhattacharya and Shuhart, 2003, Chaudhry et al., 2009, Lim et al., 2014).

Whether the CNS effects of liver disease can be identified in AUD and HIV independent of the primary diagnoses is unclear. Here, using retrospective analyses of an existing dataset, the question was asked whether serum evidence of liver disease would reveal a higher pallidal index in those with AUD, HIV, or their comorbidity. Several noninvasive serum biomarkers routinely used in clinical practice to diagnose and stage liver disease [i.e., aspartate aminotransferase to platelet ratio index (APRI); fibrosis score (FIB4); and NAFLD fibrosis score (NFS)] and a “pallidal index” were calculated. It was hypothesized that the presence of fibrosis would predict higher pallidal signal intensities independent of primary diagnoses (i.e., AUD, HIV) or presence of HCV. Further, given the extant literature, relations between the pallidal index and measures of motor function were evaluated to provide a functional consequence to the potential changes in pallidal integrity.

2. Methods

2.1. Study participants

This report used data from an existing cohort of study participants acquired between January 2006 and March 2022; portions of the data were published elsewhere (Pfefferbaum et al., 2018, Sullivan et al., 2018, Zahr et al., 2021). Data collection was conducted in accordance with protocols approved by the Institutional Review Boards of Stanford University and SRI International including written informed consent from all participants. A total of 457 individuals included healthy controls (n = 125, 52.4 ± 14.8 years), those diagnosed with an AUD (n = 186, 50.5 ± 10.7 years; currently sober as demonstrated by a negative Breathalyzer test given immediately after consent), those seropositive for HIV (n = 72, 53.5 ± 9.6 years), and those comorbid for AUD and HIV (n = 74, 51.8 ± 8.6 years).

Participants were screened using the Structured Clinical Interview for the Diagnostic Statistical Manual (DSMIV or DSM-5; SCID) (First et al., 1998), structured health questionnaires, and a semi-structured timeline follow-back interview to quantify lifetime alcohol consumption (Skinner and Sheu, 1982). Upon initial assessment, subjects were excluded if they had a significant history of medical (e.g., epilepsy, stroke, multiple sclerosis, uncontrolled diabetes, or loss of consciousness > 30 min), psychiatric (e.g., schizophrenia, bipolar disorder), or neurological (e.g., Parkinson’s disease) disorders other than an AUD (DSM-5). Other exclusionary criteria were substance dependence (other than alcohol for the 2 AUD groups) within the past 3 months or any other DSM disorder (for all 4 groups). The SCID interview also provided a Global Assessment of Functioning (GAF) score, which is a single rating scale of overall functioning ranging from 1 for sickest to 100 for healthiest individuals (Endicott et al., 1976, Williams and Rabkin, 1991). Socioeconomic status (SES) was derived from the Four-Factor Index of Social Status, which considers education and occupation level and wherein a lower score reflects higher status (Hollingshead, 1975).

The 4 groups were well-matched with respect to age, handedness, and body mass index (BMI, Table 1). As in other studies (Pfefferbaum et al., 2018, Sullivan et al., 2018, Zahr et al., 2021), the 3 diagnostic groups relative to the healthy control group were less educated, had lower SES, and lower GAF scores than controls (all p <.0001). Further, the 3 diagnostic groups relative to the healthy control group were more likely to include men (p <.0008), African American or Black individuals, and people who smoke (p <.0001). The 3 diagnostic groups were more likely to be seropositive for HCV than the control group (p <.0001); the comorbid relative to the single diagnosis groups had the highest proportion of individuals with HCV (AUD 23.1%, HIV 27.8%, AUD + HIV 47.3%; χ² = 15.0, p =.0006).

Table 1.

Demographic Characteristics by Diagnostic Group: mean ± SD or frequency count.

	Control (n = 125)	AUD (n = 186)	HIV (n = 72)	AUD + HIV (n = 74)	p-value*
N (women/men)	57/68	52/134	20/52	24/50	0.008
Age (years)	52.4 ± 14.8	50.5 ± 10.7	53.5 ± 9.6	51.8 ± 8.6	0.24
Self-Defined Race (White/AAa/otherb)	59/30/36	81/76/29	32/27/13	12/46/16	<0.0001
Handedness (Right/Left)	110/15	160/26	64/8	59/15	0.66
Body Mass Index (BMI)	26.3 ± 4.5	27.4 ± 5.0	25.9 ± 4.3	26.5 ± 4.2	0.06
Education (years)	15.7 ± 2.5	13.4 ± 2.4	13.6 ± 2.7	13.1 ± 2.0	<0.0001
Socioeconomic Statusc	27.5 ± 13.1	41.5 ± 15.1	38.6 ± 14.9	43.6 ± 12.2	<0.0001
GAFd	84.2 ± 7.0	68.7 ± 10.5	71.3 ± 10.6	65.5 ± 9.2	<0.0001
Smoker (never/past or current)	113/12	60/126	39/33	26/48	<0.0001
Age at first drink (years)	20.2 ± 6.9	16.5 ± 4.5	18.9 ± 5.6	16.9 ± 6.5	<0.0001
Lifetime Alcohol Consumption	45.6 ± 75.0	1236.8 ± 962.5	78.6 ± 93.1	957.4 ± 869.9	<0.0001
Days since last drink	–	200.0 ± 544.6	–	133.3 ± 345.8	0.25
AUDIT scorese	2.3 ± 2.3	21.5 ± 11.7	2.2 ± 2.7	12.7 ± 11.2	<0.0001
HIV duration (years)	–	–	17.9 ± 8.7	17.1 ± 7.8	0.56
CD4 cell count (100/mm3)	–	–	605.8 ± 252.6	569.6 ± 330.7	0.46
CD4 cell count nadir (100/mm3)	–	–	184.7 ± 145.5	205.7 ± 171.4	0.50
Viral Load (log copies/mL)	–	–	2.0 ± 1.1	2.3 ± 1.3	0.22
AIDS-defining eventf (yes/no)	–	–	11/61	4/70	0.22
ARTg (yes/no)	–	–	65/7	64/9	0.70
Karnofsky score	100	99.7 ± 2.1	99.3 ± 3.2	98.8 ± 3.2	0.02
Hepatitis C Virus (HCV, +/-)	3/122	43/143	20/52	35/39	<0.0001
Treatment for HCVh (yes/no)	0/122	9/177	3/69	6/68	0.03

*4-group (control, AUD = alcohol use disorder, HIV = human immunodeficiency virus, AUD + HIV) comparisons: ANOVA used on continuous variables (e.g., age), Pearson χ2 used on nominal variables (e.g., handedness), 2-group comparisons used t-tests; ^aAA = African American or Black; ^bother = Native American, Asian, Islander; ^clower score = higher status; ^dGAF = Global Assessment of Functioning; ^eAUDIT = Alcohol Use Disorders Identification Test; ^fAIDS-defining illness or CD4 prior nadir < 200cells/μl; ^gART = active retroviral therapy; ^hSelf report of HCV treatment; Bold = signficant at Bonferroni-corrected p-value = 0.006 (for 8 demographic variables).

2.2. Blood sample collection and liver fibrosis biomarkers

Serum samples were collected and analyzed by Quest Diagnostics for complete blood count (CBC) with differential, comprehensive metabolic panel, and HIV and HCV screening with RNA quantification for seropositive individuals. Laboratory results were used to calculate 3 validated, non-invasive indices of liver injury: the APRI, based on aspartate transferase (AST) and platelet levels (Fouad et al., 2012, Thandassery et al., 2016); the FIB4, based on age, AST, alanine aminotransferase (ALT), and platelet levels (Sterling et al., 2006); and the NFS, based on age, BMI, presence of diabetes, AST, ALT, platelet, and albumin levels (Lavender et al., 2020, McPherson et al., 2010).

$$\mathrm{A}\mathrm{P}\mathrm{R}\mathrm{I}=\frac{\left(\frac{\mathit{AST}(\frac{\mathit{IU}}{L})}{\mathit{AST}(upperlimitofnormal)(\frac{\mathit{IU}}{L})}\right)}{\mathit{platelet}count({10}^9/L)}\ast 100$$

$$\mathrm{F}\mathrm{I}\mathrm{B}4=\frac{age\left(years\right)\ast AST\left(\frac{\mathit{IU}}{L}\right)}{\mathit{platelet}count\left({10}^9/L\right)\ast \sqrt{\mathit{ALT}\left(\frac{\mathit{IU}}{L}\right)}}$$

$$\mathrm{N}\mathrm{F}\mathrm{S}=-1.675+0.037\left[age\left(years\right)\right]+0.094\left[BMI\left(\frac{\mathit{kg}}{m^2}\right)\right]+1.13\left[diabetes\left(1=yes,0=no\right)\right]+0.99[\left(\frac{\mathit{AST}}{\mathit{ALT}}\right)]-0.13[plateletcount\left({10}^9\right)]-0.66[albumin\left(\frac{g}{\mathit{dL}}\right)]$$

An APRI > 1.5 predicts cirrhosis; a score > 0.7 has a sensitivity of 77% and specificity of 72% for predicting significant hepatic fibrosis (Chou and Wasson, 2013, Lin et al., 2011, McPherson et al., 2010). A FIB-4 > 3.25 predicts advanced cirrhosis, while scores > 1.5 indicate fibrosis (Lavender et al., 2020, McPherson et al., 2010, Sterling et al., 2006). The NFS cutoff for cirrhosis is > 0.676; that for fibrosis is > -1.4 (Angulo et al., 2007, Lavender et al., 2020, McPherson et al., 2010, Treeprasertsuk et al., 2013).

2.3. Image acquisition and processing

Before scanning, each participant underwent and passed a metal screen designed to identify risks associated with the presence of a strong magnet field. In addition to verbal and pen-and-paper evaluation, participants passed a metal-detector screen immediately preceding the scan. MR data were collected and processed using an in-house pipeline as described (Pfefferbaum et al., 2018). Briefly, data were collected on a 3-Tesla GE whole-body MR system (General Electric Healthcare, Waukesha, WI) using an 8-channel phased-array head coil. The T1-weighted sequence was an axial Inversion-Recovery Prepared SPoiled Gradient Recalled (SPGR) (repetition time (TR) = 6.55, echo time (TE) = 1.56, inversion time (TI) = 300 ms, matrix = 256x256, thickness = 1.25 mm, skip = 0 mm, 124 slices, field of view (FOV) = 24 cm).

Preprocessing of T1-weighted SPGR data involved noise removal (Coupe et al., 2008) and brain mask segmentation using FSL BET (Smith, 2002), AFNI 3dSkullStrip (Cox, 1996), and Robust Brain Extraction (ROBEX) (Iglesias et al., 2011) generating 3 brain masks. In parallel, noise-corrected, T1-weighted images were corrected for field inhomogeneity via N4ITK (Avants et al., 2011), brain masks were segmented (Sadananthan et al., 2010), and the resulting segmented brain masks were reduced to one using majority voting (Rohlfing et al., 2004). Brain tissue segmentation (gray matter, white matter, and cerebrospinal fluid) of the skull-stripped T1-weighted images was generated via Atropos (Avants et al., 2011). Parcellated maps of tissue used the parc116 atlas to define cortical (gray matter) and subcortical (gray and white matter) volumes summed for bilateral hemispheres.

All regions of interest (ROIs) volumes were age and supratentorial volume (svol)-corrected based on a subset of 238 vetted healthy controls (121 men and 117 women; 158 white, 25 black, and 53 other race; aged 46.5 ± 17.1 years; 206 right-handed, 8 left-handed, and 9 ambidextrous; with an average BMI of 25.4 ± 4.1; average education of 16.3 ± 2.2 years; and average SES of 24.3 ± 11.2) as described in previously published studies (Pfefferbaum et al., 2018, Sullivan et al., 2018, Zahr et al., 2021). Volumes of 16 ROIs were evaluated: frontal, parietal, temporal, and occipital cortices; insular and cingulate cortices; hippocampus, parahippocampus, amygdala, caudate, putamen, pallidum, thalamus, pons; and subcortical white matter and corpus callosum.

For signal intensity calculations, ROI segmentations were overlaid on T1-weighted images. T1-weighted signal intensities per ROI as ratios relative to pontine T1-weighted signal intensity were calculated for cortical regions (frontal, cingulate, insular), basal ganglia structures (caudate, putamen, globus pallidus), and thalamus (Fig. 1). Three control participants (2 women, 1 man) had signal intensity, but not volume data.

Fig. 1.

Exemplary segmentation of basal ganglia (caudate = green, putamen = blue, pallidum = red) and thalamus (black) on coronal images from a 45.7-year-old man with AUD. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

2.4. Neuropathy and motor testing

As this was a retrospective analysis, measures that probe extrapyramidal symptoms (D'Souza and Hooten, 2022, Sanders and Gillig, 2012) such as the Extrapyramidal Symptom Rating Scale (Chouinard and Margolese, 2005) or the Columbia scale (Hely et al., 1993) were not available. Instead, scores were available for tests of neuropathy including 2-point discrimination (right and left palms of hands; right and left soles of feet; right and left combined) which uses a 3-point aesthesiometer to determine the minimal distance detected between 2 points (e.g., Corkin et al., 1970, Periyasamy et al., 2008); and perception of vibration (great toe, right and left combined) wherein impairment is perception of vibration < 10 s (Cherry et al., 2005).

Scores on alternate finger tapping (sum of all conditions) (Sullivan et al., 2002), fine finger movement (sum of all conditions) (Fama et al., 2007), and grooved pegboard (sum of left- and right-hand) (Trites, 1977) tasks were also available. Postural balance was evaluated using the walk-a-line ataxia battery, conducted first with eyes open and then with eyes closed under 4 conditions: stand heel-to-toe on a line with arms folded across the chest for a maximum of 60 s; walk a line heel to-toe for 10 steps; stand on the left foot for 30 s; stand on the right foot for 30 s. Each condition was conducted twice unless the subject achieved a perfect score on the first trial, in which case the second trial was also given a perfect score. Scores were summed for the eyes open or closed conditions (Fregly et al., 1972, Sullivan et al., 2000).

2.5. Statistical analysis

Statistics were performed using JMP® Pro 16.0.0 (SAS Institute Inc., Cary, NC, 1989–2021). Analysis of variance (ANOVA) were used for 4 group comparisons; two-group comparisons used χ² or t-tests as appropriate. Welch’s test was used for unequal variances. Correlations were evaluated using simple linear regressions or nonparametric Spearman’s ρ. Significance required Bonferroni-corrected p-values: for demographics, p≤0.006 (i.e., 0.05/8, the number of basic demographic variables in Tables 1 & 3); for signal intensities, p≤0.007 (i.e., 0.05/7 ROIs with signal intensities, Table 4); for volumes, p≤0.003 (i.e., 0.05/16 ROIs with volume data, Table 5); for neuropathies/motor performance, p≤0.006 (i.e., 0.05/8, the number of tests evaluated, Table 8). Nominal, multiple regression, or linear mixed (i.e., continuous and nominal) REML (restricted maximum likelihood) models were used to account for variance explained by relevant variables.

Table 3.

Demographic Characteristics by Liver Fibrosis Cutoff Scores and HCV: mean ± SD or frequency count *.

	normal n = 414	APRI > 0.7 n = 43	p-value	normal n = 329	FIB4 > 1.5 n = 128	p-value	normal n = 319	NFS > -1.4 n = 138	p-value	HCV neg. n = 356	HCV pos. n = 101	p-value
Diagnosis#	125/168/63/58	0/18/9/16	<0.0001	103/144/38/44	22/42/34/30	<0.0001	95/137/39/48	30/49/33/26	0.005	122/143/52/39	3/43/20/35	<0.0001
N (women/men)	144/270	9/34	0.07	117/212	36/92	0.13	104/215	49/89	0.55	117/239	36/65	0.60
Age (years)	51.8 ± 11.9	51.1 ± 7.8	0.64	49.4 ± 11.5	57.6 ± 9.5	<0.0001	48.9 ± 11.1	58.2 ± 9.8	<0.0001	51.3 ± 12.4	53.3 ± 7.7	0.05
Self-Defined Race (White/AAa/otherb)	170/161/83	14/18/11	0.51	135/122/72	49/57/22	0.29	130/112/77	54/67/17	0.004	158/118/80	26/61/14	<0.0001
Handedness (Right/Left)	360/54	33/10	0.04	284/45	109/19	0.29	275/44	118/20	0.07	309/47	84/17	0.54
Body Mass Index (BMI)	26.7 ± 4.7	26.9 ± 4.5	0.77	27.0 ± 4.7	26.2 ± 4.5	0.08	25.9 ± 4.3	28.8 ± 4.9	<0.0001	26.7 ± 4.7	26.8 ± 4.3	0.95
Education (years)	14.1 ± 2.6	13.0 ± 2.4	0.006	14.1 ± 2.6	13.8 ± 2.8	0.35	14.1 ± 2.5	13.8 ± 2.8	0.27	14.4 ± 2.6	12.6 ± 2.2	<0.0001
Socioeconomic Statusc	37.0 ± 15.5	42.8 ± 13.8	0.01	37.1 ± 14.8	38.8 ± 16.9	0.34	37.1 ± 15.0	38.8 ± 16.3	0.29	34.8 ± 15.0	47.3 ± 12.8	<0.0001
GAFd	73.5 ± 11.7	66.9 ± 11.7	0.001	73.2 ± 11.7	72.0 ± 12.3	0.32	73.1 ± 11.8	72.4 ± 12.0	0.56	74.5 ± 11.7	67.3 ± 10.8	<0.0001
Smoker (never/past or current)	223/191	15/28	0.05	168/161	60/68	0.34	166/153	72/66	0.94	203/153	35/66	0.0002
Age at first drink (years)	18.0 ± 5.9	17.3 ± 6.3	0.50	17.9 ± 5.8	18.1 ± 6.3	0.79	17.9 ± 6.1	18.2 ± 5.6	0.62	18.4 ± 5.7	16.5 ± 6.4	0.01
Lifetime Alcohol Consumption	651.5 ± 871.2	998.8 ± 1093.6	0.06	648.6 ± 858.0	774.1 ± 993.4	0.22	647.1 ± 819.2	768.2 ± 1059.9	0.24	575.6 ± 788.7	1065.9 ± 1133.9	<0.0001
Days since last drink	332.9 ± 1588.5	292.3 ± 1226.3	0.85	232.5 ± 1167.2	581.8 ± 2269.9	0.11	215.8 ± 1001.4	595.8 ± 2386.3	0.08	328.4 ± 1684.0	331.9 ± 999.2	0.98
AUDITe scores	12.0 ± 12.8	13.5 ± 11.4	0.44	12.8 ± 13.0	10.7 ± 11.7	0.12	12.3 ± 12.8	11.8 ± 12.4	0.67	11.4 ± 12.3	14.6 ± 13.4	0.05
HIV duration (years)	17.8 ± 8.5	16.1 ± 6.8	0.31	16.2 ± 8.6	19.1 ± 7.5	0.04	17.2 ± 8.4	18.0 ± 8.0	0.58	18.1 ± 8.8	16.6 ± 7.3	0.31
CD4 cell count (100/mm3)	617.2 ± 293.4	443.5 ± 258.3	0.005	643.7 ± 295.0	515.5 ± 279.5	0.008	649.8 ± 301.5	495.6 ± 259.5	0.001	581.0 ± 294.5\1	598.3 ± 297.0	0.73
CD4 cell count nadir (100/mm3)	196.7 ± 152.8	188.2 ± 183.9	0.85	200.4 ± 151.6	189.2 ± 167.1	0.72	202.5 ± 155.2	184.4 ± 164.3	0.57	171.9 ± 134.1	229.6 ± 185.4	0.08
Viral Load (log copies/mL)	2.0 ± 1.1	2.6 ± 1.3	0.07	2.0 ± 1.0	2.3 ± 1.3	0.15	2.0 ± 1.1	2.3 ± 1.3	0.25	2.1 ± 1.2	2.1 ± 1.1	0.75
AIDS-defining eventf (yes)	13	2	0.56	7	8	0.04	6	9	0.01	12	3	0.67
ARTg (yes)	102	23	0.56	72	53	0.29	78	47	0.34	76	49	0.90
Karnofsky score	99.6 ± 2.3	99.8 ± 3.3	0.20	99.6 ± 2.4	99.3 ± 2.6	0.28	99.5 ± 2.6	99.6 ± 2.0	0.57	99.6 ± 2.4	99.3 ± 2.7	0.32
Hepatitis C Virus (+/-)	71/343	13/30	<0.0001	41/288	60/68	<0.0001	53/266	48/90	<0.0001	0/356	101/0	<0.0001
Treatment for HCVh (yes/no)	15/399	3/40	0.28	9/320	9/119	0.03	9/310	9/129	0.06	0/356	17/84	<0.0001

*Welch's test (for unequal variance) used on continuous variables (e.g., age), Pearson χ2 used on nominal variables (e.g., handedness);^#control, AUD, HIV, AUD + HIV; ^aAA = African American or Black; ^bother = Native American, Asian, Islander; ^clower score = higher status; ^dGAF = Global Assesment of Functioning; ^eAUDIT = Alcohol Use Disorders Identification Test; ^fAIDS-defining illness or CD4 prior nadir < 200cells/μl; ^gART = active retroviral therapy; ^hSelf report of HCV treatment ever; Bold = signficant at Bonferroni-corrected p-value = 0.006  (for 8 demographic variables).

Table 4.

ROI Signal Intensities by Liver Fibrosis Cutoff Scores and HCV*.

Region of Interest (ROI)	APRI > 0.7		FIB4 > 1.50		NFS > -1.4		HCV status
	F ratio	p	F ratio	p	F ratio	p	F ratio	p
Cortical regions
Frontal	0.15	0.70	0.05	0.82	0.22	0.64	1.01	0.32
Cingulate	0.00	0.98	1.86	0.17	0.27	0.61	1.01	0.32
Insular	0.48	0.49	0.41	0.53	0.03	0.87	0.58	0.45
Subcortical regions
caudate	8.13	0.006	3.87	0.05	2.64	0.11	0.67	0.41
putamen	9.24	0.004	10.44	0.001	7.58	0.006	0.13	0.72
pallidum	11.50	0.002	17.62	<0.0001	5.98	0.02	1.28	0.26
thalamus	3.65	0.06	0.18	0.67	1.01	0.32	3.24	0.07

*nominal variables, Welch's test (for unequal variance); APRI n = 43 vs n = 414; FIB4 n = 128 vs n = 329; NFS n = 138 vs n = 319; HCV n = 101 vs 356; Bold = significant at Bonferroni-corrected p-value = 0.007 (for 7 ROIs).

Table 5.

ROI Volumes by Liver Fibrosis Cutoff Scores and HCV*.

		APRI > 0.7		FIB4 > 1.50		NFS > -1.4		HCV positive
Regions of Interest (ROI)		F ratio	p	F ratio	p	F ratio	p	F ratio	p
Cortical Regions
	Frontal	8.88	0.005	15.87	<0.0001	5.72	0.02	18.53	<0.0001
	Parietal	5.49	0.02	4.88	0.03	1.71	0.19	0.07	0.79
	Temporal	4.23	0.05	5.23	0.02	5.13	0.03	7.17	0.008
	Occipital	0.84	0.37	0.10	0.75	0.08	0.78	2.75	0.10
	Cingulate	2.51	0.12	2.99	0.09	0.03	0.87	3.72	0.06
	Insular	1.78	0.19	8.83	0.003	0.15	0.70	3.69	0.06
Subcortical Regions
	hippocampus	2.73	0.11	4.61	0.03	4.71	0.03	0.34	0.56
	parahippocampus	0.07	0.80	0.49	0.49	0.09	0.76	0.32	0.57
	amygdala	0.47	0.50	1.21	0.27	1.73	0.19	0.36	0.55
	caudate	6.60	0.01	0.04	0.85	0.03	0.87	0.52	0.47
	putamen	0.41	0.53	2.30	0.13	0.03	0.86	0.01	0.95
	pallidum	4.79	0.03	3.52	0.06	0.15	0.70	0.01	0.77
	thalamus	5.30	0.03	9.46	0.003	3.16	0.08	4.91	0.03
	pons	1.25	0.27	7.08	0.009	3.10	0.08	0.05	0.82
Other
	corpus callosum	6.47	0.01	1.15	0.29	0.99	0.32	0.03	0.87
	subcortical white matter	5.00	0.03	2.78	0.10	0.52	0.47	0.03	0.85

*nominal variables, Welch's test (for unequal variance); APRI n = 43 vs n = 414; FIB4 n = 128 vs n = 329; NFS n = 138 vs n = 319; HCV n = 101 vs 356; Bold = significant at Bonferroni-corrected p-value = 0.003 (for 16 ROIs).

Table 8.

ROI intensity by neuropathy scores and motor performance.

	Caudate		Putamen		Pallidum
	ρ	p	ρ	p	ρ	p
2-pt disc hands	0.08	0.22	0.12	0.05	0.11	0.07
2-pt disc feet	0.09	0.14	0.25	<0.0001	0.13	0.03
vibration sense	0.17	0.004	0.23	0.0001	0.18	0.003
AFT	−0.04	0.48	−0.06	0.30	−0.02	0.76
FFM	−0.11	0.06	−0.09	0.14	−0.08	0.19
grooved pegboard	0.18	0.003	0.17	0.005	0.14	0.02
ataxia open	−0.19	0.002	−0.20	0.0009	−0.23	0.0002
ataxia closed	−0.13	0.04	−0.16	0.009	−0.21	0.0005

*collapsed across the 3 patient groups (n = 332), nonparametric Spearman's ρ (rho) regression; AFT = alternate finger tapping; FFM = fine finger movement; Bold = significant at Bonferroni-corrected p-value = 0.006 (for 8 tests).

3. Results

3.1. Serum fibrosis Scores: Group differences

Few study participants met the threshold for cirrhosis: APRI > 1.5 (n = 6), FIB4 > 3.25 (n = 11), or NFS > 0.68 (n = 9) (Table 2). Cutoff scores for fibrosis had greater representation: APRI > 0.7 (n = 43, 9.4%); FIB4 > 1.5 (n = 128, 28.0%); NFS > -1.4 (n = 138, 30.2%) (Table 2). No healthy controls had APRI > 0.7; a similar portion of those with AUD or HIV had APRI > 0.7 (χ² = 0.44, p =.51); the AUD + HIV relative to the other 2 diagnostic groups had nominally more individuals with APRI > 0.7 (χ² = 6.7, p =.04). Control and AUD participants had a similar proportion of individuals with evidence for FIB4 liver fibrosis (χ² = 1.1, p =.29), but fewer than the HIV and AUD + HIV groups (χ² = 17.8, p =.0001), which had a similar fraction of FIB4-defined fibrosis (χ² = 0.66, p =.42). For the NFS, the HIV relative to the other 3 groups had the highest proportion of individuals above the cutoff (χ² = 12.8, p =.005; 3 group comparison without the HIV group, χ² = 3.1, p =.22).

Table 2.

Liver Biomarkers* by Diagnostic Group.

		Control (n = 125)	AUD (n = 186)	HIV (n = 72)	HIV + AUD (n = 74)	Total
APRI (mean ± SD)		0.22 ± 0.09	0.35 ± 0.46	0.45 ± 0.50	0.47 ± 0.44
	>1.5	0	3(1.6 %)	2(2.8 %)	1(1.4 %)	6(1.3 %)
	>0.7	0	18(9.7 %)	9(12.5 %)	16(21.6 %)	43(9.4 %)
FIB4 (mean ± SD)		1.04 ± 0.52	1.25 ± 0.76	1.61 ± 0.86	1.61 ± 0.96
	>3.25	0	4(2.2 %)	3(4.2 %)	4(5.4 %)	11(2.4 %)
	>1.5	22(17.6 %)	42(22.6 %)	34(47.2 %)	30(40.5 %)	128(28.0 %)
NFS (mean ± SD)		−2.34 ± 1.29	−2.03 ± 1.16	−1.53 ± 1.08	−1.76 ± 1.20
	>0.68	1(0.8 %)	2(1.1 %)	3(4.2 %)	3(4.1 %)	9(2.0 %)
	>-1.4	30(24.0 %)	49(26.3 %)	33(45.8 %)	26(35.1 %)	138(30.2 %)

*Frequency counts and ratios.

Demographic characteristics of the groups by fibrosis cutoff scores or HCV diagnosis are presented in Table 3. Of note, individuals with fibrosis scores above cutoffs were more likely than those below cutoff scores to have a lower current CD4 cell count and to be seropositive for HCV.

3.2. Serum fibrosis Scores: Relations with signal intensities

Regional signal intensities differentiated by fibrosis cutoff scores (Table 4) demonstrated that presence of fibrosis distinguished high signal intensities in the putamen (APRI >.07: p =.004; FIB4 > 1.5: p =.001; NFS < -1.4: p =.006); pallidum (APRI >.07: p =.002; FIB4 > 1.5: p <.0001); and caudate (APRI > 0.7: p =.006) (Fig. 2). A nominal logistic model including caudate, putamen, and pallidum singal intensities explained 25.6% of the variance in the APRI cutoff score and was driven by pallidal intensity (p =.0005), which alone explained 25.0% of the variance. Similarly, a nominal logistic model including putamen and pallidum explained 23.6% of the variance in the FIB4 cutoff score and was driven by pallidal intensity (p =.0004), which alone explained 23.6% of the variance. Conversely, 4 variables (APRI > 0.07, FIB4 > 1.5, NFS > -1.4, HCV status) explained 8.6% of the variance in pallidal signal intensity in a significant model (F_{3, 456} = 10.6, p <.0001) driven by APRI > 0.07 (p <.0001; FIB4 > 1.5: p =.02; NFS > -1.4: p =.87; HCV: p =.10).

Fig. 2.

Basal ganglia signal intensities by APRI (red, top) or FIB4 (blue, bottom) cutoff scores. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

Data evaluated using serum biomarkers as continuous variables (rather than cutoffs) and collapsed across the 3 patient groups (i.e., AUD, HIV, AUD + HIV, n = 332) demonstrated similar results: higher scores on all 3 fibrosis biomarkers correlated with higher signal intensities in the pallidum (APRI p =.007, FIB4 p <.0001, NFS p =.004) (Table S1). FIB4 (p =.0007) and NFS (p =.001) were additionally associated with higher signal intensity in putamen.

3.3. Serum fibrosis Scores: Relations with brain volumes

Of the 3 fibrosis biomarker cutoff scores, only the FIB4 discriminated regional volume differences (Table 5). FIB4 > 1.5 was associated with smaller volumes of the frontal (p <.0001) and insular (p =.003) cortices and thalamus (p =.003). Presence of HCV was also associated with smaller frontal volume (p <.0001). FIB4 > 1.5 and HCV together explained 6.4% of the variance in frontal volume in a significant model (F_{2, 456} = 10.6, p <.0001) with similar contributions from HCV (p =.002) and FIB4 (p =.002).

Evaluating ROI volumes using serum biomarkers as continuous variables and including only the 3 diagnostic groups (i.e., AUD, HIV, AUD + HIV, n = 332) demonstrated that higher FIB4 scores correlated with smaller volumes of frontal cortex (p =.001) and thalamus (p =.001, Table S2).

3.4. Prediction models

Multiple regression analyses to determine the relative contribution of the APRI to explaining the variance of intensity or volume of 7 ROIs included demographic variables different among those with scores above the APRI cutoff including nominal variables diagnosis and HCV status and continuous variables education, GAF, and CD4 count. The APRI explained a significant portion of the variance in signal intensities of putamen and pallidum even after considering contributions of the other evaluated variables (Table 6).

Table 6.

Variance explained in prediction models using APRI.

Predicted variable	Model p-value	Variance explained	Relevant variable	p-value
Intensity
caudate	0.20	6.7 %	APRI > 0.7	0.06
putamen	0.03	10.4 %	APRI > 0.7	0.002
pallidum	0.05	9.6 %	APRI > 0.7	0.005
thalamus	0.05	9.4 %	GAF	0.03
Frontal cortex	0.23	6.4 %	education	0.01
Cingulate cortex	0.06	9.1 %	education	0.004
Insular cortex	0.06	9.0 %	education	0.05
Volume
caudate	0.07	8.8 %	education	0.002
putamen	0.36	5.2 %	education	0.03
pallidum	0.46	4.6 %	education	0.03
thalamus	0.21	6.6 %	CD4 count	0.01
Frontal cortex	0.21	6.6 %	CD4 count	0.01
Cingulate cortex	0.57	4.0 %	CD4 count	0.08
Insular cortex	0.99	0.6 %	GAF	0.52

Model run across the 3 diagnostic groups (n = 332, excluding controls) included 3 nominal (diagnosis, APRI > 0.7, HCV status) and 3 continuous (education, GAF, CD4 count) variables.

A similar multiple regression to determine the relative contribution of the FIB4 to explaining the intensity or volume of 7 ROIs included demographic variables different among those with scores above the FIB4 cutoff included nominal variables diagnosis and HCV status and continuous variables age and CD4 count. The FIB4 explained a significant portion of the variance in intensity of putamen and pallidum even after considering contributions of the other evaluated variables (Table 7).

Table 7.

Variance explained in prediction models using FIB4.

Predicted variable		Model p-value	Variance explained	Relevant variable	p-value
Intensity
	caudate	0.06	8.1 %	age	0.03
	putamen	0.0007	14.8 %	FIB4 > 1.5	0.007
	pallidum	0.01	10.7 %	FIB4 > 1.5	0.008
	thalamus	0.12	6.8 %	age	0.007
	Frontal cortex	0.36	4.5 %	FIB4 > 1.5	0.08
	Cingulate cortex	0.05	8.3 %	age	0.004
	Insular cortex	0.14	6.6 %	CD4 count	0.06
Volume
	caudate	0.72	2.5 %	age	0.27
	putamen	0.54	3.4 %	age	0.07
	pallidum	0.75	2.4 %	FIB4 > 1.5	0.15
	thalamus	0.27	5.1 %	FIB4 > 1.5	0.05
	Frontal cortex	0.08	7.4 %	CD4 count	0.01
	Cingulate cortex	0.36	4.5 %	CD4 count	0.05
	Insular cortex	0.06	8.1 %	age	0.03

Model run across the 3 diagnostic groups (n = 332, excluding controls) included 3 nominal (diagnosis, FIB4 > 1.5, HCV status) and 2 continuous (age, CD4 count) variables.

3.5. Relations between signal intensities and brain volumes

ROI signal intensities relative to volume were evaluated for 7 regions (n = 457, Fig. 3). Only the pallidum showed a relation between higher signal intensity and smaller volume (r = -0.47, p <.0001). This relationship persisted when the 4 outliers were removed (r = -0.35, p <.0001, Fig. 3 inset). The cingulate showed an opposite relationship: lower signal intensity was associated with smaller volume (r = -0.16, p <.0001, Fig. 3). Thalamic signal intensity and volume were not related (r = -0.08, p =.07, figure not included).

Fig. 3.

Relations (simple linear regressions) between signal intensities and volumes evaluated across the 3 patient groups (n = 332). Only the globus pallidus demonstrated a significant relationship between higher signal intensity and smaller volume (inset, correlation without the 4 "outliers").

3.6. Relations between brain Measures, Neuropathy, and motor Functioning

ROIs that showed higher signal intensity (i.e., caudate, putamen, and pallidum) were evaluated for their relations with neuropathy measures, upper motor functions, and ataxia among individuals in the diagnostic groups (n = 332, Table 8, Fig. 4). Higher caudate signal intensity was associated with worse scores on vibration sense (higher score = worse performance), grooved pegboard performance (higher score = worse performance), and ataxia (eyes open). The putamen demonstrated the same correlations and was additionally associated with worse 2-point discrimination (feet). Higher pallidal signal intensity was associated with worse vibration sense and ataxia performance (both eyes open and closed). Together, caudate, putamen, and pallidal signal intensities explained 5.8% of the variance in vibration sense; putamen alone explained a significant portion of the variance (5.6%, p =.005). Caudate and putamen contributed equivalently to explaining 4.6% of the variance in grooved pegboard performance. Similarly, signal intensity in the 3 regions explained 6.0% of the variance in ataxia (eyes open) performance with equivalent contributions from each region. Putamen explained 4.2% of the variance in 2-point discrimination (feet); pallidum explained 2.9% of the variance in ataxia (eyes closed).

Fig. 4.

Correlations between higher signal intensities in basal ganglia structures and performance on sensory and motor tasks.

4. Discussion

To our knowledge, this is the first study to use serum biomarkers to determine relations between liver fibrosis and regional brain signal intensities and volumes. The key result is evidence for higher signal intensities unique to lentiform nucleus (i.e., putamen and pallidum) in those with APRI or FIB4 scores above cutoffs, even after considering confounding variables of original diagnosis (i.e., AUD or HIV), presence of HCV, CD4 cell count, age, education, and GAF. Further, high T1 signal intensities in these basal ganglia regions contribute to neuropathy (worse vibration sense), slower eye-hand coordination (grooved pegboard), and ataxia. Together, these findings suggest that in individuals with APRI or FIB4 scores above cutoffs, the lentiform nucleus is vulnerable to pathology that may contribute to impairment of sensory and motor functions.

The current cohort was not selected for clinically-evident liver disease, but the incidental presence of liver fibrosis occurred in study participants with AUD, HIV, and their comorbidity. The APRI cutoff indicated only 9.4% of the cohort as having liver fibrosis, had the greatest representation from the comorbid group, and was most selective for basal ganglia abnormalities (i.e., high signal intensities). The FIB4 identified 28.0% of the cohort as having fibrosis and had greater representation from the 2 HIV groups relative to the AUD and control groups. In addition to higher signal intensities in the basal ganglia, smaller volumes of frontal and insular cortices and thalamus were noted in those with FIB4 scores above cutoffs relative to those with scores below the cutoff. Although the NFS identified liver fibrosis in the largest fraction of the cohort (30.2%), it only identified abnormalities in putamen (i.e., high signal intensity). That the APRI and FIB4 might outperform the NFS in identifying brain changes affected by liver disease comports with a study using APRI, FIB4, and NFS indices to determine outcomes following intracerebral hemorrhagic stroke: APRI and FIB4, but not the NFS were associated with hematoma volume and expansion (Parikh et al., 2020).

Among the basal ganglia regions with high T1 signaling (i.e., caudate, putamen, pallidum), pallidal intensity explained a unique and significant proportion of the variance in APRI (25.0%) and FIB4 (23.6%) scores beyond that explained by signal intensities in caudate or putamen. A rich literature demonstrates a high “pallidal index” (ratio of signal intensity in pallidum relative to cortical white matter) in later stages of liver disease which appears to be independent of etiology [e.g., non-alcoholic cirrhosis (Awada et al., 1995); alcoholic cirrhosis (Pujol et al., 1993, Thuluvath et al., 1995); renal failure (da Silva et al., 2007); urea cycle disorders (Häberle et al., 2019); portosystemic shunt (Fukuzawa et al., 2006); Alagille syndrome (Nagai et al., 2022)]. Indeed, a hyperintense T1-weighted pallidal signal was recently shown to have fair diagnostic performance for identifying patients with cirrhosis (with 66% sensitivity and 78% specificity), but poorly discriminated hepatic encephalopathy (Sparacia et al., 2022). Further, among the ROIs evaluated herein, only the pallidum showed a relationship between higher signal intensity and smaller volume. Prevailing hypotheses regarding mechanisms of pallidal signal hyperintensities in liver disease involve manganese accumulation (e.g., Klos et al., 2006, Krieger et al., 1995, Layrargues et al., 1998, Layrargues et al., 1995, Rose et al., 1999). Whether manganese deposition in globus pallidus – associated with significant loss of dopamine D2 receptor binding sites (e.g., Mousseau et al., 1993, Watanabe et al., 2008) and astrocyte toxicity (Jayakumar et al., 2004) – is sufficient to promote volume loss remains unclear.

Relatively few studies have reported on regional brain volumes in liver disease absent hepatic encephalopathy. Contrasting with the current results, a small study including 30 individuals with hepatitis B virus (HBV)-induced cirrhosis reported a higher pallidal index associated with larger globus pallidus volume (Zheng et al., 2020). In HCV-related cirrhosis, all 34 participants showed pallidal signal hyperintensities, but of those, only the 18 patients with a history of overt hepatic encephalopathy demonstrated reduced pallidal volume relative to healthy controls (Marano et al., 2015). In other neuroimaging studies, pallidal volume has not been specifically identified. Steatosis in the general population (Evangelou et al., 2021) and non-alcoholic fatty liver disease (NAFLD) (Filipović et al., 2018) have been associated with non-specific CNS tissue volume deficits. In a study including 25 AUD participants, a fibrosis score comprising alpha 2-macroglobulin, platelets, prothrombin ratio, and hyaluronic acid was associated with smaller volumes of putamen and cerebellum (Lanquetin et al., 2021). Similarly, among basal ganglia regions with reduced mean kurtosis in cirrhosis patients, the putamen showed the best diagnostic performance in distinguishing those with impaired neuropsychological performance (Sato et al., 2019). In a large, multisite study, liver fibrosis assessed using the FIB4 was associated with smaller volume of the hippocampus (Parikh et al., 2022a) and increased risk for dementia (Parikh et al., 2022b).

The effects of liver fibrosis on lentiform nucleus signal intensity were independent of HCV infection as demonstrated by the multiple regressions presented in Tables 6&7. Instead, presence of HCV affected frontal cortex volume comporting with our previous work demonstrating the added burden of HCV to frontal cortex integrity in the context of AUD and HIV (Sullivan et al., 2018, Zahr et al., 2021). Structural neuroimaging studies of HCV (absent cirrhosis, AUD, or HIV) relative to healthy controls also report deficits specific to frontal cortex (Hjerrild et al., 2016); atrophy of insular cortex and thalamus has also been noted in HCV absent cirrhosis (Prell et al., 2019).

The functional relevance of high signal intensities in basal ganglia structures was demonstrated by their relations with neuropathy, motor, and ataxia measures. The pallidum explained some of the variance in performance of ataxia (eyes closed), comporting with a study showing a relationship between a high pallidal index and greater postural body sway (Kim et al., 2007). The putamen alone explained a significant portion of the variance in vibration sense and 2-point discrimination (feet). Although peripheral neuropathy can contribute to balance impairment (Ferris et al., 2020, Manor et al., 2012, Sullivan et al., 2021, Sullivan et al., 2022), evidence for direct relations between basal ganglia integrity and peripheral neuropathy is not prominent in the literature, and may be moderated by nutrition (cf., Comi et al., 2014). Here, caudate and putamen (but not pallidum) contributed similarly to explaining the variance in grooved pegboard performance, contrasting with results demonstrating an effect of a high pallidal index on grooved pegboard performance (cf., Chang et al., 2009, Chang et al., 2010) (also see, Bowler et al., 2018).

In summary, among three validated liver fibrosis indices, APRI and FIB4 cutoff scores identified high signal intensities in the lentiform nucleus. These results suggest that the APRI and FIB4, which can be derived from common clinical laboratory measures, may be useful in identifying patients with globus pallidus vulnerability and consequent sensory and motor impairment. Questions yet to be addressed include the generalizability of the current findings to liver disease from other etiologies (e.g., NAFLD); why the globus pallidus among other basal ganglia regions may be especially susceptible to liver disease; and mechanisms of globus pallidus pathology.

Ethics Statement: Data collection was conducted in accordance with protocols approved by the Institutional Review Boards of Stanford University and SRI International and included written informed consent from all participants.

Grant Support: This study was supported with grant funding from the National Institute of Alcohol Abuse and Alcoholism (NIAAA) including AA005965, AA017347, and AA010723 (NMZ) and K23AA029197 (AJK).

CRediT authorship contribution statement

Allison J. Kwong: Project administration, Funding acquisition, Validation, Writing – review & editing. Natalie M. Zahr: Conceptualization, Methodology, Funding acquisition, Formal analysis, Data curation, Resources, Project administration, Supervision, Validation, Writing – original draft, Writing – review & editing.

Declaration of Competing Interest

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Appendix A. Supplementary data

The following are the Supplementary data to this article:

Data availability

Data will be made available on request.