Granulocyte-monocyte/macrophage apheresis for steroid-nonresponsive or steroid-intolerant severe alcohol-associated hepatitis: A pilot study

Abstract

Background:

Patients with severe alcohol-associated hepatitis (SAH) have a high short-term mortality rate. Unmet needs exist in patients who are refractory to corticosteroids (CS) or are ineligible for early liver transplantation.

Methods:

This was a prospective, open-label, nonrandomized pilot study conducted at a liver transplant center in Tokyo, Japan, starting in October 2015. Lille model and Model for End-stage Liver Disease (MELD) score-defined CS nonresponsive or CS-intolerant patients with SAH who fulfilled the inclusion criteria (leukocytosis over 10,000/μL, etc.) were considered for enrollment. The median duration from admission to enrollment was 23 days (IQR, 14-31 days), after standard of care. Granulocyte-monocyte/macrophage apheresis (GMA) performed with Adacolumn twice per week, up to 10 times per treatment course, was evaluated.

Results:

13 GMA treatments were conducted through December 2021. Maddrey Discriminant Function was 53.2 $\mathrm{\pm }$ 17.7 at admission. The overall survival rate was 90.9% at 90 and 180 days. MELD scores significantly improved, from median (IQRs) of 23 (20–25) to 15 (13–21) after GMA (p<0.0001). Estimated mortality risks using the Lille model and MELD scores significantly improved from 20.9%±16.5% to 7.4%±7.3% at 2 months and from 30.4%±21.3% to 11.6%±10.8% at 6 months, respectively (both p<0.01), and were internally validated. The cumulative rate of alcohol relapse was 35.9% per year. No severe adverse events were observed. In exploratory analysis, granulocyte colony-stimulating factor levels were significantly correlated with prognostic systems such as MELD-Sodium scores after GMA (correlation coefficient= −0.9943, p<0.0001) but not before GMA (p=0.62).

Conclusions:

Compared to published studies, GMA is associated with a lower-than-expected 90- and 180-day mortality in patients with CS-nonresponsive or CS-intolerant SAH. GMA may meet the needs as a salvage anti-inflammatory therapy for SAH. (Trial registration: UMIN000019351 and jRCTs No.032180221) (274 words).

INTRODUCTION

Alcohol use is the leading risk factor for attributable disability-adjusted life-years for individuals aged 25 to 49 years worldwide.¹ Despite advances in basic and translational research activities, the short-term mortality of patients with severe forms of alcohol-associated liver diseases remains as high as ~40%–50%.²,³,⁴ Severe alcohol-associated hepatitis (severe AH; SAH), characterized by the recent onset of jaundice in patients with ongoing alcohol abuse, is often accompanied by subsequent liver decompensation and extrahepatic organ failure, a state known as acute-on-chronic liver failure (ACLF).⁴,⁵

In the short term, management for SAH aims to limit liver injury and inflammation; however, these aims are different in the long term.⁶ In addition to abstinence and nutritional support, the use of corticosteroids (CS), which target liver injury and inflammation, has demonstrated improved short-term (28-day) overall survival⁷,⁸ and is recommended as the standard of care (SOC) for SAH in the latest international guidelines and a recent expert review.⁹,¹⁰ Scoring systems, including the Lille model and the Model for End-stage Liver Diseases (MELD) system, are useful for clinical decision-making or risk estimation in the management of CS-treated patients with SAH.¹¹,¹²,¹³,¹⁴ Early liver transplantation within 6 months of abstinence is usually not considered due to regional allocation policies⁹,¹⁰,¹⁵; therefore, unmet needs exist in the management of patients with SAH who are refractory to medical treatment or ineligible for early liver transplantation.¹³,¹⁴ Moreover, because CS can only improve survival at 28 days,⁷ not only null responders of CS but also patients who present with partial responses (28-day overall survival 79.4±3.8% in a meta-analysis³) to CS are thought to be reasonable candidates for novel clinical studies targeting liver injury or inflammation.¹⁰,¹³,¹⁴

Leukocytosis, especially neutrophilia, was reported to be an indicator of poor prognosis for SAH in an observational study published in 1974.¹⁶ Recently, high-degree intrahepatic neutrophil infiltration, which characterizes a distinct histopathological phenotype of SAH, has been reported to be an important driver of liver injury and hepatocellular failure.¹⁷ Adacolumn (JIMRO, Ltd, Takasaki, Japan),¹⁸ a granulocyte (eg, neutrophils) and monocyte/macrophage apheresis (GMA) system that removes activated neutrophils and monocytes/macrophages from the peripheral circulation, has been approved for clinical application in inflammatory bowel diseases¹⁹,²⁰ and pustular psoriasis²¹ in Japan, with good safety profiles. Clinical observations demonstrated the possible efficacy of GMA in patients with SAH²,²²,²³,²⁴,²⁵; however, all of these were uncontrolled and did not consider of the Lille model-defined or MELD score–defined CS responsiveness.

In this prospective, open-label, nonrandomized pilot study, we aimed to demonstrate whether GMA improves 90- and 180-day overall survival in Lille model-defined and MELD score–defined CS-nonresponsive or CS-intolerant patients with SAH, with the efficacy of GMA as a salvage anti-inflammatory therapy.

METHODS

Study design and participants

Study conduct

This was a prospective, open-label, nonrandomized pilot study conducted at our tertiary medical center, a liver transplant center in metropolitan Tokyo, Japan, since October 2015. Ethical approval for this study (No. 20150156) was provided by the Institutional Review Board of Keio University School of Medicine, Tokyo, Japan, on October 15, 2015. All research was conducted in accordance with both the Declarations of Helsinki and Istanbul. Written consent was given in writing by all subjects. This study was registered at the University Hospital Medical Information Network (UMIN) Center as UMIN000019351 and Japan Registry of Clinical Trials as No. 032180221, http://links.lww.com/HC9/A767.

Screening for eligibility

In total, consecutively admitted 29 patients with AH were screened during the study period. The management flowchart of AH, along with the screening strategy for eligibility (Figure 1A) and the Consolidated Standards of Reporting Trials (CONSORT) diagram inclusive of detailed case numbers of eligibility assessment, enrollment, allocation, follow-up, and analysis (Figure 1B) are illustrated. Generally, the screening and management flow agreed with the latest international guidelines⁹ and expert opinions.¹³,¹⁴,²⁶ AH was diagnosed based on clinical features of chronic alcohol abuse and recent presentation of jaundice according to criteria raised by the Japanese Society for Biomedical Research on Alcohol, which was congruent to those by the National Institute on Alcohol Abuse and Alcoholism (NIAAA).²⁶ Viral hepatitis, such as chronic infection of HBV or HCV, acute exacerbation of HBV, and superimposed acute infection of HAV or HEV, was excluded serologically. Autoimmune hepatitis, DILI, and Wilson disease were excluded by history, serological assessments, and diagnostic criteria. Due to coagulopathy associated with liver failure, liver biopsy was only performed in patients in which AH could not be otherwise diagnosed. Immediately after admission, patients with AH underwent SOC (see below) and severity assessments. SAH was defined as AH with Maddrey Discriminant Function (mDF) ≥32 or MELD≥20.¹⁰ In addition to SOC, patients with SAH were considered for CS (prednisolone 40 mg/d as a standard dosage). The Lille model on day 7 or the MELD score was utilized for the assessment of responses to CS in accordance with previous reports or expert opinions.⁹,¹³,¹⁴ A Lille model score ≥0.56 (as null responders) or >0.16, <0.56 (as partial responders)³ or a MELD score consistently ≥20 was considered to taper CS quickly and to screen for inclusion. A patient who was tolerant to CS and had a Lille score ≤0.16 and a MELD score <20 continued CS for another 21 days. A patient with a worsening MELD score ≥16 during this period, indicative of a lower probability for recompensation,²⁷ was also screened. A patient who was intolerant to CS, such as suspicious of active infection at any time under SOC, was further screened for participation.

FIGURE 1.

Flowchart showing patient management, eligibility assessment and the complete Consolidated Standards of Reporting Trials (CONSORT) diagram of this study. (A) The management flowchart of patients with AH along with the screening strategy for eligibility of participation. (B) The complete CONSORT diagram inclusive of detailed case numbers of screening (for eligibility and for participation), enrollment, allocation, follow-up, and analysis. Abbreviations: ACLF, acute-on-chronic liver failure; AH, alcohol-associated hepatitis; CS, corticosteroid; GMA, granulocyte-monocyte/macrophage apheresis; mDF, Maddrey Discriminant Function; MELD, Model for End-stage Liver Disease; SAH, severe alcohol-associated hepatitis; SOC, standard of care; WBC, white blood cell count.

Participants: inclusion/exclusion criteria

After screening for eligibility, enrollment was assessed. The inclusion criteria were as follows: patients aged between 20 and 70 who met the diagnostic criteria of SAH with peripheral white blood cell (WBC)>10,000/μL. All patients were able to access vascular circulation to receive GMA procedure (see below). Patients with neutrophil count <2000 were excluded. Patients with complications such as severe extrahepatic organ failure, uncontrolled infection, severe heart diseases, systemic hypotension, or advanced malignancy were excluded from this study. Pregnant patients were also excluded due to unconfirmed safety profiles. Please refer to the CONSORT diagram illustrated in Figure 1B.

Standard of care

SOC was provided in accordance with the latest international guidelines⁹ and an expert review.¹⁰ The management of complications of decompensated cirrhosis and ACLF in accordance with the international guidelines²⁸ was also included as part of SOC. Details are described in the Supplemental Methods, http://links.lww.com/HC9/A766.

GMA device and procedural description

Adacolumn, the device for GMA, which was described in a previous report,¹⁸ was kindly provided by the manufacturer (JIMRO, Ltd, Takasaki, Japan). GMA was conducted twice weekly and was stopped as WBC<10,000/μL, for up to 10 times each treatment course. Circuit outlines and detailed treatment schedules are shown in Supplemental Figure S1, http://links.lww.com/HC9/A766. Refer to the Supplemental Methods for details, http://links.lww.com/HC9/A766.

Outcomes

The primary outcome was overall survival at 90 and 180 days. The secondary outcomes were (1) changes in clinical parameters, including prognostic scores before and after GMA administration, (2) cumulative rate of discharge from the Liver Unit and the duration of admission since the start of GMA, (3) cumulative rate of alcohol relapse, (4) overall survival during observation. Events associated with hepatic decompensation, including gastrointestinal bleeding, HE, ascites of new onset, and spontaneous bacterial peritonitis, were also recorded. Adverse events during GMA were also monitored. Please refer to the Supplemental Methods, http://links.lww.com/HC9/A766, for details of data collection.

Historical controls compiled from published literature

Overall survival at 90 and 180 days was compared to patients with SAH who were treated with CS and demonstrated Lille score–defined nonresponse to CS in previous clinical trials (historical controls) by a systemic search in published literature. Only patients who were not treated with any medication of interest other than CS were pooled. The literature search was last conducted in December 2022 by searching PubMed with the keywords “alcoholic hepatitis/alcohol-associated hepatitis” and “clinical trial.” Finally, 6 clinical trials³,⁷,²⁹,³⁰,³¹,³² were included (Supplemental Figure S2, http://links.lww.com/HC9/A766).

Exploratory analyses: immunoassays of serum samples

Serum samples were analyzed using immunoassays before and after GMA in 6 patients (Supplemental Table S1, http://links.lww.com/HC9/A766). MILLIPLEX Human Cytokine/Chemokine/ Growth Factor Panel A (Merck KGaA, Darmstadt, Germany) was used. Details are described in the Supplemental Methods, http://links.lww.com/HC9/A766.

Statistical analyses

Sample size estimation is described in the Supplemental Methods, http://links.lww.com/HC9/A766. Data were analyzed using JMP15 (SAS Institute Inc. Cary, NC). Graphs and linear correlations were constructed using Prism 9.0 (GraphPad Software Inc. San Diego, CA). For the primary outcome, Kaplan-Meier survival analyses with a log-rank test at a two-sided alpha level of 0.05 were applied. For combined survival analysis from published cohorts (Supplemental Table S2, http://links.lww.com/HC9/A766), individual patient survival data were retrieved from the total number and actuarial curve of survival according to the method described by Mathurin et al and others.³,³³ Briefly, the number of steps in the Kaplan-Meier survival curve, allowing for larger steps representing 2 or more nearly concurrent deaths, implies that any remaining patients not accounted for must have been censored. Careful measurement of survival probabilities at each step, and hence the size of successive steps, indicated the censoring times, which must have occurred. Because most previous reports apply 0.45 as a cutoff of the Lille model to determine CS nonresponse as reported in the original report,¹¹ a post hoc analysis that included patients with a Lille score over 0.45 was also conducted. For background comparison, Unpaired t tests with Welch correction were applied to evaluate the statistical differences of background MELD scores and mDF between our cohort treated with GMA and each historical control cohort individually.

Data were expressed as medians with interquartile ranges and/or averages ±SDs. Paired t tests were conducted to assess the differences in parameters before and after GMA. Differences between the 2 groups were analyzed using Mann-Whitney tests, or unpaired t test with Welch correction, as appropriate. A joint-effect model combining the Lille model and MELD score to calculate the 2- and 6-month overall mortality¹² was applied for risk estimation before and after GMA. Internal validation for risk estimation was performed by bootstrapping analysis using the R software (version 3.6.0). Spearman correlation was used for the correlation analysis. Results were considered significant at p<0.05.

RESULTS

Characteristics of the study subjects

From October 2015 to December 2021, 13 GMA treatment courses were conducted on 11 patients with CS-nonresponsive or CS-intolerant SAH. All of them were diagnosed as definite or probable AH, as classified by the NIAAA.²⁶ The baseline clinical characteristics are shown in Table 1. Detailed clinical demographics, including the reasons for inclusion and the number of times GMA was performed per treatment course, are shown in Supplemental Table S1, http://links.lww.com/HC9/A766. Two patients, Case No.1 and No.3, underwent a second treatment course of GMA due to alcohol relapse after recovery from the first episode of CS-nonresponsive SAH (with a 5-year and 2-year interval, respectively). Three patients had infectious complications at admission; 2 of them (Case No.2 and No.4) were included without pretreatment with CS owing to intolerance. The observation period ranged from 60 to 2220 days. Three patients had ACLF grade 1–3 upon admission. Their ACLF was managed appropriately that all patients were stabilized to the state of “no ACLF” before the start of CS or GMA. mDF at admission was 53.2±17.7. Two patients had Lille scores over 0.56 (null responders). Five patients had Lille scores between 0.56 and 0.45, while another 4 patients had scores between 0.16 and 0.45 (partial responders). On day 0 of GMA, WBC count was 24.3 (IQR, 21.1–25.7) ×10⁹/L. MELD score was 23 (IQR, 20–25).

TABLE 1.

Clinical characteristics at admission and just before GMA of 13 treatment courses of GMA in 11 patients^a

	At admission	Day 0 of GMA
N (by capita)/N (by course)	11/13	—
Female (%)	63.6% (7 of 11)
Parameters by course
Age (y)	46.6±9.7; 48 [38.5–55]
Alcohol consumption, g/d	112.0±44.4; 108 [72–138]
Days hospitalized before GMA	—	23±11.3; 23 [14–31]
Corticosteroid use before GMA	—	84.6% (11 of 13)
Infection before GMA, %	—	23.1% (3 of 13)
WBC count (×109/L)	23.3±7.9; 22.6 [15.6–30.7]	24.8±7.5; 24.3 [21.1–25.7]
Total bilirubin (mg/dL)	14.1±7.0; 13.4 [8.7–17.9]	13.0±7.4; 13.8 [5.7–19.1]
PT-INR	1.86±0.36; 1.75 [1.67–1.92]	1.72±0.25; 1.64 [1.49–1.93]
Serum creatinine (mg/dL)	0.68±0.38; 0.53 [0.46–0.74]	0.65±0.27; 0.54 [0.50–0.76]
Albumin (g/dL)	2.5±0.4; 2.6 [2.1–2.8]	2.7±0.5; 2.7 [2.3–3.1]
AST (U/L)	101.4±54.7; 76 [52–160]	85.8±58.1; 54 [41–129]
ALT (U/L)	31.8±24.4; 21 [15–52]	50.5±75.8; 25 [19–47.5]
γ-GTP (U/L)	263±228; 217 [77–454]	235±166; 217 [85–371]
CRP (mg/dL)	5.21±2.64; 6.1 [2.7–7.5]	2.77±1.65; 2.46 [1.66–4.03]
Prognostic scores by course
mDF	53.2±17.7; 51.3 [36.9–61.1]	—
MELD score	24±4; 23 [21–27]	22±3.3; 23 [20–25]
MELD-Na score	25.7±4.5; 25.1 [21.8–30.0]	24.0±4.0; 24.9 [20.7–27.5]
CLIF-C AD score	61.3±7.2; 60.1 [55.1–65.6]	60.4±5.3; 61.5 [55.8–65.0]
CLIF-C OF score	7.9±1.3; 8 [7–9]	7.5±0.9; 8 [7–8]
CLIF-C ACLF grade, no ACLF/Gr. 1/Gr. 2/ Gr. 3	10/1/2/0	13/0/0/0
CLIF-C ACLF score	44.2±7.0; 48.0 [39.6–48.6]	43.5 $\mathrm{\pm }$ 5.7; 45.0 [39.7–47.9]

Note: Statistics show both mean $\mathrm{\pm }$ SD and median with interquartile ranges in square brackets.

^aSecond treatment course of GMA was applied for case 1 and case 3, as demonstrated in Supplemental Table S1, http://links.lww.com/HC9/A766.

Abbreviations: ACLF, acute-on-chronic liver failure; AD, acute decompensation; ALT, alanine aminotransferase; AST, aspartate aminotransferase; CLIF-C, Chronic Liver Failure Consortium; CRP, C-reactive protein; GMA, granulocyte monocyte/macrophage apheresis; Gr, grade; γ-GTP, γ-glutamyl transpeptidase; mDF, Maddrey Discriminant Function; MELD, Model for end-stage liver disease; OF, organ failure; PT-INR, prothrombin time-international ratio; WBC, white blood cell.

Overall survival at 90 and 180 days

Two patients (case no. 5 and no. 9) lost follow-up at 72 and 60 days after GMA, respectively. They recovered from SAH and discharged from the Liver Unit at 15 and 24 days after GMA, respectively. Another patient, case no.7, recovered from SAH and was discharged from the Liver Unit to home 25 days after GMA, had an alcohol relapse 30 days after GMA, suffered from another episode of SAH-related ACLF, was readmitted and died 75 days later. The overall survival rate was 90.9% at 90 and 180 days (Figure 2A). Insignificant statistical differences were observed either in baseline MELD scores or in mDF between our cohort treated with GMA and the CS-nonresponsive patients of most historical cohorts (Supplemental Table S2, http://links.lww.com/HC9/A766). The overall survival rate of all 6 published cohorts (historical controls; Supplemental Table S2, http://links.lww.com/HC9/A766 for details, N=467) was 57.9% (95% CI: 53.1%–62.4%) at 90 days, and 51.7% (95% CI: 46.7%–56.6%) at 180 days. Compared to published cohorts, the GMA group demonstrated significantly better overall survival at 180 days (Log-rank test, HR, 8.0; 95% Cl: 3.9–16.4; p=0.0128). In addition, we included 2 post hoc analyses. In the first post hoc analysis, we included only published cohorts in which patients’ baseline MELD scores or mFD values that were calculable and not worse or better than our cohort. By this definition, 3 published cohorts⁷,³⁰,³² were included, and a significantly better overall survival was also observed (HR, 6.6; 95% Cl: 3.0–14.7; p=0.0283) (Figure 2B). In the second post hoc analysis inclusive of 7 patients (in 7 treatment courses) with Lille score over 0.45, this significantly better overall survival was maintained (HR, 5.8; 95% Cl: 2.5–13.4; p=0.0447) (Figure 2C).

FIGURE 2.

The 180-day overall survival of patients who underwent GMA compared to historical controls of corticosteroid nonresponders of severe alcohol-associated hepatitis. (A) Whole group of study subjects and all included published historical cohorts (see also Supplemental Figure S2, http://links.lww.com/HC9/A766). (B) A first post hoc analysis with published cohorts in which patients’ baseline MELD scores or mDF values that were statistically not worse or better than the GMA cohort (see also main text and Supplemental Table S2, http://links.lww.com/HC9/A766). (C) A second post hoc analysis of the study subjects with a Lille score over 0.45. Broken lines indicate the 95% CIs. *p<0.05. Abbreviations: Cont, historical control group; G, GMA group; GMA, granulocyte-monocyte/macrophage apheresis.

Changes in biochemical parameters and prognostic scores after GMA

Changes in biochemical parameters and prognostic scores that were collected before and after GMA were assessed with paired t tests. After GMA, the WBC count, predominantly driven by a decrease in neutrophil count, decreased significantly, while aspartate aminotransferase and alanine aminotransferase demonstrated statistically insignificant decreasing tendencies (Figure 3A). Parameters related to liver reserve function, such as total bilirubin levels, prothrombin time-international ratio, and albumin levels, significantly improved after GMA. Serum C-reactive protein, an inflammatory biomarker that was mainly produced by the liver under stimulation of IL 6, decreased significantly (Figure 3A). Accordingly, all composite prognostic scoring systems significantly improved; namely, MELD scores stabilized from 23 (median, IQR, 20–25) to 15 (median, IQR, 13–21) (Figure 3B). Compared to the Lille scores during the first 7 days of CS treatment with SOC (11 treatment courses) or SOC alone (2 treatment courses), the Lille scores calculated after GMA showed a significant improvement (Figure 3B). The estimation of overall mortality risk at 2 and 6 months, calculated using a joint-effect model combining both the Lille model and MELD score,¹² further demonstrated a significant improvement after GMA, from 20.9%±16.5% to 7.4%±7.3% at 2 months, and from 30.4%±21.3% to 11.6%±10.8% at 6 months, respectively (both p<0.01) (Figure 3C). These mortality estimates were also internally validated by bootstrapping, as shown in Supplemental Table S3, http://links.lww.com/HC9/A766.

FIGURE 3.

Comparisons of serum levels of biochemical analyses (A), prognostic scores, Lille scores (B) and mortality estimation at 2 and 6 months (C) before and after GMA. Assessed by paired t tests. Please refer to the main text and Supplemental Methods, http://links.lww.com/HC9/A766 for a detailed calculation of the prognostic scores, the Lille model, and the mortality estimation using a joint-effect model combining the Lille model and MELD scores. *p<0.05; **p<0.01; ***p<0.001; ****p<0.0001. Abbreviations: AD, acute decompensation; ACLF, acute-on-chronic liver failure; ALB, albumin; ALT, alanine aminotransferase; AST, aspartate aminotransferase; CLIF-C, Chronic Liver Failure Consortium; CRP, C-reactive protein; CS, corticosteroid; GMA, granulocyte-monocyte/macrophage apheresis; γ-GTP, γ-glutamyl transpeptidase; MELD, Model for End-stage Liver Disease; MELD-Na score, Model for End-stage Liver Disease-sodium; ns, not significant; OF, organ failure; PT-INR, prothrombin time-international ratio; SOC, standard of care; T-Bil, total bilirubin; WBC, white blood cell.

Other secondary outcomes

All patients recovered from SAH and were discharged from the Liver Unit at around a median duration of 25 days (range, 15–48 days) following GMA initiation (Supplemental Figure S3A, http://links.lww.com/HC9/A766). Excluding 2 patients who were lost to follow-up, 5 patients experienced first alcohol relapse (30, 45, 60, 100, and 380 days following GMA initiation, respectively). The cumulative rate of first alcohol relapse was 35.9%/year (Supplemental Figure S3B, http://links.lww.com/HC9/A766), which is comparable to a previous report.⁶ The long-term overall survival results are shown in Supplemental Figure S3C, http://links.lww.com/HC9/A766. Four patients died from liver failure owing to alcohol relapse. About events associated with hepatic decompensation, no gastrointestinal bleeding HE or spontaneous bacterial peritonitis was observed during GMA. Diuretics were initiated in 3 patients for new onset of ascites during GMA.

Safety

Patients tolerated GMA well without any major adverse events. No events of hemodynamic instability, bleeding, or allergic reactions due to extracorporeal circulation were observed during the procedures of GMA. One episode of new onset, uncomplicated urinary tract infection occurred in case no. 3. The patient completed antibiotic treatment and recovered well.

Exploratory analyses of cytokines/chemokines

Serum cytokines, chemokines, and growth factors levels were assessed at day 0 of GMA administration and after GMA administration in 6 cases (Supplemental Table S1, http://links.lww.com/HC9/A766). These levels are shown in Supplemental Figure S4, http://links.lww.com/HC9/A766. Their correlation with various markers of liver injury, prognostic biomarkers, and scoring systems has been extensively analyzed. Granulocyte colony-stimulating factor (G-CSF), a growth factor that has been the focus of many previous studies on patients with SAH,²⁹,³⁴ was particularly highlighted. Although the serum levels of G-CSF did not correlate well with prognostic scoring systems such as MELD-Na scores before GMA (Figure 4A, left), surprisingly, after GMA, G-CSF serum levels were highly and significantly inversely correlated with almost all prognostic scores, especially the MELD-Na scores (R=−0.99) (Figure 4A, right), suggesting a possible relative insufficiency of G-CSF in patients with high MELD-Na scores after GMA. Similar associations with mortality estimation calculated using the joint-effect model combining the Lille model and MELD scores at 6 months were also observed (Supplemental Figure S5A, B, http://links.lww.com/HC9/A766). Moreover, other growth factors, such as VEGF-A and PDGFs, and T_H2 cytokines such as IL4 and IL13, were significantly inversely correlated with the prognostic systems after GMA (Figure 4B). After GMA, serum levels of G-CSF, PDGF-AA, VEGF-A, interferon-γ, IL12p70, IL4, IL13, and IL22 were higher in patients with lower MELD-Na scores (cutoff at 18), while those of IL15 and IL18 were lower (Supplemental Figure S5C–E, http://links.lww.com/HC9/A766). These findings propose previously undescribed targets for the development of novel treatments for SAH.

FIGURE 4.

Correlations between clinical parameters and various cytokines/chemokines /growth factors before and after GMA in the exploratory analysis. Correlations between the serum levels of G-CSF and the MELD-Na scores before and after GMA (A) are shown. In (B), heat map of correlation coefficients between clinical parameters and various cytokines/chemokines /growth factors before and after GMA. R, correlation coefficient. The broken lines in A indicate 95% CIs. *p<0.05; **p<0.01; ***p<0.001. Abbreviations: G-CSF, granulocyte colony-stimulating factor; GMA, granulocyte-monocyte/macrophage apheresis; MELD-Na, Model for End-stage Liver Disease-sodium.

DISCUSSION

In this study, we demonstrated a lower-than-expected mortality in Lille model-defined or MELD score-defined CS-nonresponsive or CS-intolerant patients with SAH who were treated with GMA (Figure 2). Various biochemical parameters, prognostic systems, and mortality risk estimations improved significantly with the GMA (Figure 3). The inclusion criteria of this study had the following characteristics: (1) high-grade ACLF with extrahepatic organ failure is managed before CS; (2) GMA is considered a salvage anti-inflammatory/anti-injury treatment for CS in a setting where SOC with little heterogeneity is conducted; (3) cases of CS partial responders (Lille score 0.16–0.45), cases with high MELD but low Lille scores, and CS-intolerant cases are also screened. We believe that this design is congruent with recent expert opinions¹³,¹⁴,²⁶ and might help clarify whether and how GMA treatment can be best utilized.

The inclusion criteria of this study included the laboratory finding, peripheral WBC count>10,000/μL. In a recent insightful translational study, higher neutrophil infiltration and neutrophil cytosolic factor 1 were shown to characterize liver injury and disease progression in SAH.¹⁷ In the post hoc analysis of a randomized double-blind controlled trial utilizing GMA for active ulcerative colitis, GMA showed significant efficacy in a subgroup of patients with “severe acute inflammation” with erosion and ulceration.³⁵ Thus, the active inflammatory phase with neutrophils that characterized SAH satisfied the rational for utilizing GMA in these patients.

Previous studies have shown that successful short-term anti-inflammatory treatment may not ensure the medium-term or long-term survival of patients with SAH.⁶,²⁷ For medium-term survival, liver regeneration¹⁵ following the resolution of inflammation is needed in most patients with higher MELD scores.²⁷ Except for early liver transplantation within 6 months of abstinence, G-CSF has been focused on by various studies that targeted anti-inflammation and liver regeneration, but the outcomes were inconsistent.³⁴ Our exploratory analyses showed that during active phases before GMA, serum G-CSF levels correlated poorly with prognostic scores. However, the significantly high correlation with prognostic scores after GMA (Figure 4) suggests that after leukocyte apheresis, the real associations between low G-CSF levels and insufficient liver regeneration may become recognizable. It is also important to know that the beneficial roles of G-CSF are phase-specific or condition-specific, as demonstrated in a recent study using mouse models.³⁶ Thus, sequential utilization of G-CSF following GMA, especially in patients presented with higher MELD-Na scores, may be beneficial. Further investigations are needed.

The small sample size and the lack of randomization limit the generalizability of this study, even though most available data included in the historical controls did not demonstrate significant baseline statistical differences either in MELD scores or in mDF, as shown in Supplemental Table S2, http://links.lww.com/HC9/A766. Furthermore, the rarity and the high acute and medium-term mortality make two-arm randomization both difficult and unethical. In addition, the efficacy of GMA in patients with SAH with WBC <10,000/μL was not examined in this study.

In conclusion, this pilot study demonstrated a possible better survival benefit at 90 and 180 days among patients with CS-nonresponsive or CS-intolerant SAH who underwent GMA. These findings suggest a possibly beneficial role for GMA as a salvage therapy following CS. Further investigations with a larger number of patients are warranted.