Granulocyte colony-stimulating factor plus pentoxifylline increases short-term survival in patients with severe alcoholic hepatitis: a network meta-analysis

ABSTRACT Background: Optimal treatments for severe alcoholic hepatitis (SAH) remain controversial. Previous network meta-analysis showed that corticosteroid (CS) combined with N-acetylcysteine (NAC) was superior in reducing short-term mortality of patients with SAH. Recently, granulocyte colony-stimulating factor (G-CSF) treatments for SAH yielded promising results. Objectives: To determine how currently available treatments affect the survival and complications of patients with SAH. Methods: The study was conducted following the guidelines of PRISMA. The data from PubMed, Embase, MEDLINE, Cochrane Library, and clinicaltrials.gov to October 2022 were searched, and patients with SAH with pharmacotherapy were included in our study. The primary outcome was short-term survival, and the other outcomes were medium- (3/6 months) or long-term (12 months) survival and complications after treatment. R software was used to establish network meta-analysis models and the result was expressed by the odd ratio (OR) value and 95% credible interval (Crls). Results: A total of 31 randomized controlled trials, including 19 treatment regimens, were enrolled in our study. As the primary outcome, G-CSF+ pentoxifylline (PTX) ranked first in one-month survival and showed significant superiority when compared with the placebo (OR 8.60, 95% Crls 1.92–45.10) and CS (OR 4.95, 95% Crls 1.11–25.53). Also, G-CSF+PTX ranked first in improving three-month survival and reducing the occurrence of infection. PTX+MTD ranked first in six-month survival, and G-CSF ranked first in twelve-month survival. CS+MTD ranked first in the occurrence of gastrointestinal bleeding and hepatorenal syndrome. Conclusions: The combination of G-CSF and PTX showed a significant benefit in improving the short-term survival of SAH patients.


Introduction
As a severe form of alcohol-related liver disease (ARLD), alcoholic hepatitis is characterized by jaundice, fever, and elevated aspartate transaminase/alanine transaminase and may progress to liver failure in some patients.Severe alcoholic hepatitis (SAH) is defined by Maddrey's discriminant function (MDF) of ≥ 32.The short-term survival of patients with SAH without liver transplantation is not optimistic.The 28-day mortality rate was reported to be as high as 50% (1), which would be worse in steroid nonresponders (2).Multiple treatments have been tried for the management of SAH (3)(4)(5)(6)(7)(8)(9)(10)(11)(12)(13).However, the optimal treatment for SAH remains controversial.Liver transplantation, especially when performed early, is effective in saving the lives of patients with SAH (14,15).However, it is only available for a minority of patients due to the lack of liver donors and ethical problems.
With a high number of trials on SAH having been performed, a number of direct pairwise meta-analysis studies were implemented to compare the efficacy of different interventions for SAH (1,(16)(17)(18).However, conventional pairwise meta-analyses can only compare the efficacy of paired interventions.Nevertheless, more than 10 treatment regimens have been tried in patients with SAH.In this way, network meta-analysis is more useful, as it can compare multiple measures directly and indirectly and provide full-scale evidence for clinical decision-making.In 2015, Singh et al. (18) performed a network meta-analysis comparing five different interventions from 22 randomized controlled trials.The network meta-analysis showed the addition of N-acetylcysteine (NAC) to corticosteroids (CS) may be superior to CS alone for reducing short-term mortality.Since then, more studies on SAH presented

Materials and methods
The systematic review and meta-analysis were carried out, and the results were reported, following the guidelines of Preferred Reporting Items for Systematic Reviews and Meta-analysis (PRISMA) for Network Meta-analysis (Supplementary material 1).The study protocol was registered on the International Prospective Register of Systematic Reviews (PROSPERO) with a registration number CRD42022302784.

Search strategy
Four databases (PubMed, Embase, MELDLINE, Cochrane library) and one website (https://www.clinicaltrials.gov/)were searched from the establishment to October 2022 to identify the randomized controlled trials (RCT) research concerning on the effects of pharmacotherapy for patients with SAH compared with placebo or other therapy (for detail of literature searching, see Supplementary material 2).

Inclusion and exclusion criteria
Patients with SAH treated with pharmacotherapy were chosen to be included in our study, regardless of their country, race, and sex.SAH was defined as a MDF of ≥ 32.Studies meeting the following criteria were considered for inclusion in our study (1): patients >18 years old (2); randomized controlled trial (RCT) studies published in English (3); full-text research studies.Studies which met the following criteria were excluded (1): protocol, review, case report, or case series studies (2); cohort studies, cross sectional studies, etc (3).; duplicated research (including repeated publication and secondary analysis).

Data extraction and literature quality assessment
Two researchers (F.D. and C.L.) independently screened and extracted relevant data, including the country where the study was conducted, year of publication, study design, duration of therapy, dose of a related drug, follow-up time, and reported outcome of concern.The survival profile and occurrence of decompensated events were extracted separately.The Cochrane ROBINS-I tools were used to evaluate the risk of bias for the enrolled studies.Two independent reviewers (F.D. and C.L.) conducted the quality assessment and divergence were solved by discussion or consensus with a third reviewer (S.Y.).

Outcomes
The primary outcome was short-term (1-month survival) of patients with SAH.Other outcomes included medium-term (3/6 months), and long-term (12 months) survival and complications, including incidences of infection, gastrointestinal bleeding, and hepatorenal syndrome after the treatments.

Data analysis
The R software (4.1.2,R Foundation for Statistical Computing, Vienna Austria) was mainly used for statistical analysis and chart drawing, and the "gemtc" package of the R software was used to transfer "jags" software to realize network meta-analysis under a random effect Bayesian model.Odds ratio (OR) was selected as the effect quantity and the results were expressed by the OR value and 95% credible interval (Crls) (P < .05).The R software was used to construct the consistency model and compare the effect of different treatment regimens.
After the establishment of consistent models on the survival and complications of patients with SAH, the relationship between different treatment regimens was shown in the evidence network maps.Then, the rank probabilities of different treatment regimens concerning survival and complications were compared by the surface under the cumulative ranking curve (SUCRA) value and shown by ranking probability diagrams.Finally, the difference between direct and indirect evidence was compared using the node-splitting method, and related forest maps were drawn using the R software.I 2 was used to evaluate statistical heterogeneity.P > .05signified no inconsistency, whereas P ≤ .05signified the presence of inconsistency.

Study characteristics and literature quality assessment
A total of 4,159 studies were retrieved from PubMed, Embase, MEDLINE, Cochrane, and ClinicalTrial.gov.
After removing duplicates, irrelevant, and incomplete studies, 31 studies, including 19 treatment regimens related to the treatment of patients with SAH, were ultimately left.The specific screening process is shown in Figure 1.The characteristics of enrolled studies, including histological evidence of SAH, drug dose and study duration, follow-up time, concomitant therapies, and reported outcome of concern are listed in Table 1.Literature quality assessment was conducted according to the clauses of the Cochrane Handbook and shown by the summary and traffic plot (Figure 2a,b).Among the 31 enrolled studies, 9 studies showed low bias, 12 showed high bias, and 10 showed unknown bias.
For the rank of probabilities, G-CSF+PTX had the highest SUCRA score (0.94) of being the best short-term survival treatments.The two treatment regimens (G-CSF and PTX) alone got lower SURCA.However, there were only two studies mentioned the combined therapy and the results were limited.The SUCRA score and rankogram for the primary outcome appeared in Table 4 and Figure 4a.In the heterogeneity analysis based on pair-wise meta-analysis, only PTX vs placebo  showed high heterogeneity (I 2 = 78.0%).The forest plot compared the results of pair-wise meta-analysis and network meta-analysis appeared in Figure S1.

Sensitivity analysis and assessment of inconsistency.
The sensitivity analysis by eliminating the high-bias studies showed that the results were mostly consistent with the main meta-analysis (Table S8).The test for global inconsistency showed that transitivity was favorable concerning survival and occurrence of complications.For local inconsistency, no heterogeneity was observed between direct and indirect evidence for survival, occurrence of infection and gastrointestinal bleeding, indicating that the relevant models presented good robustness.However, there was heterogeneity between the direct and indirect evidence of CS+PTX vs. CS (p < .05)and PTX vs. CS+PTX (p < .05)for occurrence of hepatorenal syndrome, which limited the application of the results to some extent (Figure 5).

Discussion
Network meta-analysis is a powerful tool to compare multiple interventions from related studies to determine the best treatment for SAH.Previous network metaanalysis showed the superiority of CS+NAC for reducing the one-month mortality in patients with SAH.With more RCT trials conducted in patients with SAH, we performed this up-to-date network metaanalysis.We found the superiority of G-CSF+PTX for improving one-month/three-month survival and reducing infection.As for complications, we found that CS +MTD showed its advantage in improving mediumterm survival and reducing gastrointestinal bleeding and hepatorenal syndrome.
The efficacy and safety of GSF have already been verified in promoting hepatic regeneration and improving liver injury (9,(40)(41)(42).Recently, a total of three studies reported that G-CSF-related therapies in patients with SAH, including G-CSF alone (13), G-CSF plus PTX (9,12), and G-CSF plus PTX and NAC (12), all found the related therapies could significantly improve the 3-month survival were well tolerated.In our network meta-analysis, G-CSF +PTX showed significant superiority in improving one-month/three-month survival and reducing infection compared to placebo and CS.As SAH may progress to acute on chronic liver failure, which results in the death of patients, this finding was supported by a recent meta-analysis that showed G-CSF may decrease infection risk in patients with acute on chronic liver failure (43).However, further studies with larger sample sizes and long-term follow-up are needed to verify these preliminary results.
Higuera-De La Tijera et al. compared PTX, CS, PTX +MTD and CS+MTD for 3-month and 6-month survival and found the group treated with MTD owned a higher survival rate and abstinence rate compared with the group not treated (10).In our network metaanalysis, PTX+MTD ranked first in 3-month survival.Besides, CS+MTD ranked second in 3-month and 6-month survival and ranked first in reducing gastrointestinal bleeding and hepatorenal syndrome, although the role of MTD combination therapy in reducing infection is not prominent.As an antioxidant, MTD participates in the synthesis of the glutathione and the prevention of liver steatosis (44).Moreover, MTD also help patients remain abstinent (45), which makes it a promising treatment for patients with SAH.
survival of patients with SAH (6,18).However, the benefit of NACs in SAH remains controversial (46).Amjad et al. recently questioned the benefit of NAC+CS in SAH patients (47).In our network meta-analysis, we did not observe any difference in improving the 1-month survival when we compared CS and CS+NAC.The actual effect of NAC in SAH is still under investigation.In this study, only 61.2% (183/299) patients with SAH responded to standard CS treatment at day 7.For nonresponders, early liver transplantation is in urgent need.However, stigma to patients with alcohol-related liver diseases is one of the obstacles that hinder them from receiving transplant, when compared with liver disease due to other etiologies.Previously, 6-month abstinence is necessary for patients with ARLD to get access to liver transplantation.However multiple studies showed early liver transplantation might benefit these patients (48,49).Yet for patients whose abstinence is less than 6 months, multidisciplinary evaluation and management of abstinence was necessary, as patients with ARLD always comorbid with AUD.To overcome the critical barriers, lee et al. (50) proposed to integrate the psychotherapy and pharmacotherapy for patients in the transplant waiting list and got promising results (51,52).Recently, DiMartini
et al. (53) suggested that this integrated care model could expand to all alcohol-caused liver disease including severe alcoholic hepatitis.In the future, the dual management of liver disease and alcohol use disorders by hepatologists and addiction experts will optimize outcomes for these patients.
It is worth mentioning that our research bears certain limitations.First, there was non-negligible heterogeneity between studies, and even the specific scheme of the same drug may have been different.Second, the quality of some enrolled studies was low, which may have been related to the fact that blind methods were not implemented.Thirdly, including only literature published in English may lead to publication bias, although most of the studies might be interest to us were published in this language.Lastly, our results uncovered the advantage of G-CSF-related therapies and CS+MTD, but the application of our results in clinical treatment may limited due to the small sample size of relevant studies and the lack of long-term survival data.
In summary, we reviewed the published RCT studies on drug treatment of SAH patients, established network meta-analysis models of survival and complications, ranked the effects of different treatment regimens, and verified the established models.Finally, we concluded that G-CSF+PTX bore significant superiority in the 1-month/3-month survival and reducing infection and is promising for extending the patients' short-term and medium-term survival.Besides, we found that MTD-related therapy shows superiority in improving medium-term survival and reducing gastrointestinal bleeding and hepatorenal syndrome.However, more studies with larger sample sizes and longer follow-ups are needed to verify the effect of G-CSF+PTX therapy.

Figure 1 .
Figure 1.Flowchart of the screening of the present study.

Figure 4 .
Figure 4. Rankogram of different treatment regimens in survival and complications of patients with SAH.(a) 1-month survival.(b) 3-month survival.(c) 6-month survival.(d) 12-month survival.(e) Incidence of infection.(f) Occurrence of gastrointestinal bleeding.(g)Occurrence of hepatorenal syndrome.The column represents the possibility of different measures in different sequences, and the height of the column is positively correlated with the ranking.Note: A, placebo; B, corticosteroid; C, amlodipine; D, antioxidant cocktail; E, colchicine; F, corticosteroid combined with infliximab; G, corticosteroid combined with metadoxine; H, corticosteroid combined with N-acetylcysteine; I, corticosteroid combined pentoxifylline; J, corticosteroid combined with S-adenosyl-L-methionine; K, granulocyte colony-stimulating factor; L, insulin combined with glucagon; M, N-acetylcysteine; N, propylthiouracil; O, pentoxifylline; P, granulocyte colony-stimulating factor combined with pentoxifylline; Q, granulocyte colony-stimulating factor combined with pentoxifylline and N-acetylcysteine; R, pentoxifylline combined with metadoxine; S, anakinra combined with pentoxifylline and zinc.

Figure 5 .
Figure 5. Node separation forest map of different treatment regimens in survival and complications of patients with SAH.(a) 1-month survival.(b) 3-month survival.(c) 6-month survival.(d) 12-month survival.(e) Incidence of infection.(f) Occurrence of gastrointestinal bleeding.(g) Occurrence of hepatorenal syndrome.Compared with the results of direct and indirect comparisons between different measures, p < .05indicates a statistical difference.Note: A, placebo; B, corticosteroid; I, corticosteroid combined pentoxifylline; O, pentoxifylline.

Table 1 .
Baseline characteristics of enrolled studies.

Table 2 .
Network meta-analysis results of survival in SAH patients.

Table 3 .
Network meta-analysis results of complications in SAH patients.

Table 4 .
Ranking of surface under the cumulative ranking curve value of different treatment regimens in survival and complication of patients with severe alcoholic hepatitis.