Administration of Albumin Solution Increases Serum Levels of Albumin in Patients With Chronic Liver Failure in a Single-Arm Feasibility Trial

Background & Aims Infections are life-threatening to patients with acute decompensation and acute-on-chronic liver failure (AD/ACLF). Patients with AD/ACLF have prostaglandin E2–mediated immune suppression, which can be reversed by administration of albumin; infusion of 20% human albumin solution (HAS) might improve outcomes of infections. We performed a feasibility study to determine optimal trial design, assess safety, and validate laboratory assessments of immune function to inform design of a phase 3 trial. Methods We performed a prospective multicenter, single-arm, open-label trial of 79 patients with AD/ACLF and levels of albumin lower than 30 g/L, seen at 10 hospitals in the United Kingdom from May through December 2015. Patients were given daily infusions of 20% HAS, based on serum levels, for 14 days or until discharge from the hospital. Rates of infection, organ dysfunction, and in-hospital mortality were recorded. The primary end point was daily serum albumin level during the treatment period. Success would be demonstrated if 60% achieved and maintained serum albumin levels at or above 30 g/L on at least one third of days with recorded levels. Results The patients’ mean model for end-stage disease score was 20.9 ± 6.6. The primary end point (albumin ≥30 g/L on at least one third of days recorded) was achieved by 68 of the 79 patients; 75% of administrations were in accordance with suggested dosing regimen. Mean treatment duration was 10.3 days (104 ± 678 mL administered). There were 8 deaths and 13 serious adverse events, considered by the independent data-monitoring committee to be consistent with those expected. Twelve of 13 patients that developed either respiratory or cardiovascular dysfunction (based on ward-based clinical definitions) as their only organ dysfunction were alive at 30 days compared with 1 of 3 that developed renal dysfunction. Only 1 case of brain dysfunction was recorded. Conclusions In a feasibility trial, we found that administration of HAS increased serum levels of albumin in patients with AD/ACLF. The dosing regimen was acceptable at multiple sites and deemed safe by an independent data-monitoring committee. We also developed a robust system to record infections. The poor prognosis for patients with renal dysfunction was confirmed. However, patients with cardiovascular or respiratory dysfunction had good outcomes, which is counterintuitive. Severe encephalopathy appeared substantially under-reported, indicating that ward-based assessment of these parameters cannot be recorded with sufficient accuracy for use as a primary outcome in phase 3 trials. Trial registration no: EudraCT 2014-002300-24 and ISRCTN14174793.

See editorial on page 633, and related article on page 738. L iver disease represents 2% to 3% of deaths globally 1,2 and was estimated to be responsible for more than 1 million deaths in 2010. 1 Incidence rates are predicted to double over the next 20 years. 3 Cirrhosis patients with liver failure are termed acute decompensation (AD) or acute-on-chronic-liver failure 4 (ACLF). They are highly prone to bacterial infection 5,6 secondary to immune dysfunction, 7 with nosocomial infection rates of 35% compared with 5% in noncirrhotic patients. 8 Bacterial infection with organ dysfunction carries a mortality risk ranging from 60% to 95%. 9,10 Currently, there is no immune-restorative treatment for patients with AD/ACLF.
Intravenous albumin is used commonly in patients with AD/ACLF as a volume expander [11][12][13] and has been proven to prevent and improve renal failure. 14,15 Many hepatologists believe it has additional properties, 16 however, no multicenter evidence exists for these putative properties and critical-care studies have failed to establish a role. 17 In our previous work, we showed that increased circulating prostaglandin E 2 (PGE 2 ) was a potential key immune-suppressive mediator in AD. 18 We also developed an immune assay to assess the effects of patient plasma on healthy monocyte-derived macrophage tumor necrosis factor (TNF) production because monocyte deactivation is a key feature of AD/ACLF and reduced monocyte TNF production predicts poor outcomes in sepsis. 11 We recently showed, using a whole-blood stimulation assay, a reduction in TNF production by fresh circulating monocytes taken from ACLF patients. These data match our observations made when monocyte-derived macrophages are treated with AD/ ACLF plasma, validating our approach. 12 The only clinical characteristic that predicted immune suppression using this approach was a serum albumin value less than 30 g/ L. Albumin has been reported to bind and catalyze inactivation of PGE 2 , 13 and both circulating levels and binding capacity 19 decrease in AD/ACLF, making PGE 2 potentially more bioavailable. In our pilot study, ex vivo immune function improved after 20% human albumin solution (HAS) transfusions. 18 Therefore, we hypothesized that infusing albumin in hospitalized AD/ACLF patients with a serum albumin value less than 30 g/L with a target to increase this value to 35 g/L or higher would improve immune function, prevent new infections, and thus improve mortality. In view of the current discrepancies in ACLF definition 20 we believed that large-scale recruitment required simple inclusion criteria and wished to test whether basing these on a serum albumin level less than 30 g/L would select appropriate patients.
There were several other uncertainties regarding our protocol that required clarification before embarking on a large, randomized controlled trial (RCT) comparing albumin infusions with standard care. These included the use of a 72-hour window from admission to treatment, our end point definitions, and obtaining daily data and sample collection. Furthermore, only 1 previous UK multicenter interventional AD/ACLF trial has been performed, the Steroids or Pentoxifylline for Alcoholic Hepatitis trial. 21 Furthermore, a recent HAS infusion in sepsis study did not achieve the 30 g/L target 22 and a HAS in cirrhosis trial was stopped early because of potential adverse effects. Therefore, a feasibility study was essential to determine whether our suggested protocol was achievable, effective, and did not suggest safety concerns. We did not include a standard-care arm in this feasibility trial and the study was not powered to detect clinical outcomes.

Study Design
We conducted this prospective, multicenter, openlabel, feasibility trial of targeted 20% HAS infusions in UK AD/ACLF patients to inform the design of a phase 3 RCT. A protocol report was published 23 and the full protocol is available online. All authors had access to the study data, and reviewed and approved the manuscript.

Inclusion Criteria
All patients admitted to the hospital with AD or severe worsening of the complications of cirrhosis, who were older than age 18 years, with a serum albumin level less than 30 g/L, a predicted hospital admission longer than 5 days, and that were for full active management at admission were eligible. Patients were recruited within 72 hours of admission; the full inclusion criteria are listed in Supplementary Table 1

Intervention
In-patients received daily 20% HAS infusions intravenously for the treatment period, with a maximum of 14 days or until the patient was considered medically fit for discharge if fewer than 14 days. The volume of HAS prescribed was determined by the patients' serum albumin level that day (Supplementary Table 2), based on the Albumin Replacement in Patients with Severe Sepsis study 22 and clinical experience. All clinicians could deviate from the suggested regimen, but were requested to document their reasons for doing so.

Clinical Outcomes
The primary end point was daily serum albumin value for the treatment period. Success was defined as 60% of patients achieving and maintaining serum albumin levels at or greater than 30 g/L on at least one third of the days in the study with recorded levels.
The rates of infection, organ dysfunction (for definitions see Supplementary Tables 3 and 4), and in-hospital mortality, the component elements of the planned composite end point for the RCT, also were recorded. Definitions of organ failure were chosen from the European Foundation for the study of chronic liver failure Sequential Organ Failure Assessment score. 24 We used a recent definition of renal dysfunction in cirrhosis 25 reflecting deterioration in patient renal function rather than static values. Because studies have shown that up to 50% of liver patients considered to be infected may have negative microbiological culture, 26 we used antibiotic initiation as a surrogate for infection diagnosis. A subset of patients had clinical, biochemical, and microbiological data recorded for validation of infection diagnosis.
Data were summarized further within groups defined by baseline serum albumin levels (<20, 20-25, and 26-29 g/L) to investigate any apparent differences in outcome by group. This subgroup analysis was considered crucial to potentially identify whether any protocol amendments were necessary. The information was summarized for the total volume of albumin infused and of fluid administered, days spent in the intensive care unit, and the duration of hospital stay. Safety was assessed by our Independent Data Monitoring Committee (IDMC) who considered the number of serious adverse events (SAEs) reported during the trial.

Independent Data Monitoring Committee
The IDMC received monthly updates of the SAEs to ensure robust monitoring and met every 6 months to report to the Trial Steering Committee. The SAEs were recorded from entry into study until discharge or the end of treatment.

Statistical Methods
A sample size of 80 ensured 72 evaluable patients (this assumed a 10% loss to follow-up evaluation/withdrawal), with a probability of achieving 44 or more successes of 80%. Calculations assumed each patient had a 65% chance of albumin level being 30 g/L or greater on at least one third of the days when the level was recorded. If 44 successes or more were observed, then a single-sided 90% CI would suggest that the true rate is higher than 50%. Because this was a feasibility study, the emphasis was on producing relevant data summaries rather than on formal modeling or hypothesis testing. All analyses were undertaken by an intention-to-treat, including all consented and enrolled patients for whom outcomes were available, and according to a prespecified (approved before database lock) statistical analysis plan that is included in the Supplementary Materials and Methods section.

Trial Registration
The trial is registered with the European Medicines Agency (European Clinical Trials Database 2014-002300-24) and adopted by National Institute for Health Research (International Standard Randomised Controlled Trial Number 14174793).

Recruitment and Baseline Clinical Characteristics
Between May and December 2015, there were 517 AD/ACLF patients who were screened at 10 UK hospitals, with 124 eligible for inclusion and 80 enrolled. One patient was excluded from analysis ( Figure 1A) because their baseline albumin level was greater than 30 g/L. Their mean age was 53.4 years, with a male predominance (66%), and alcohol was the primary underlying cause in 96% ( Table 1). The mean Model for End-stage Liver Disease score was 20.9 (SD, 6.62), 17 of 79 patients had 1 or more extrahepatic organ dysfunction(s) at baseline according to our definitions, and 21 patients had ACLF grades 1 to 3 according to the European Association for the Study of the Liver-CLIF definition. 27 Sixty-seven percent of patients had baseline albumin levels less than 25 g/L. The mean time from admission to enrollment was 1.8 days. The most common reasons for ineligibility during screening were as follows: (1) albumin level of 30 g/L or greater; (2) admission more than 72 hours before screening; and (3) predicted hospital stay of fewer than 5 days. The most common reasons for nonparticipation of eligible patients were as follows: (1) patient declined; (2) informed consent was not possible; and (3) the site was unable to randomize within 72 hours.

Targeted 20% Human Albumin Solution Infusions Increase Serum Albumin Level to 30 g/L or More
A total of 68 of 79 patients achieved a serum albumin level of 30 g/L or greater on at least one third of days recorded, providing a success rate of 86% (95% CI, 76%-92%) ( Table 2, and Supplementary Table 5). All patients enrolled contributed to the analysis. The regimen was effective across all serum albumin subgroups, with the highest success in the 26 to 29 g/L group (96% success; 95% CI, 80%-100%) compared with less than 20 g/L (50% success; 95% CI, 19%-81%). The mean treatment length was 10.3 days and a mean of 1042 mL 20% HAS (SD, 678 mL) was administered; more than 50% of patients had their albumin level restored to 30 g/L or greater after 3 days, and more than 75% after 7 days. Deaths (n ¼ 8) or SAEs (n ¼ 12) during the trial period were not considered directly related to the HAS infusions by the IDMC (Table 3). Sixtyfour percent of administrations were in accordance with the suggested protocol, with 88% within AE100 mL of the suggested dose (Supplementary Table 6). No patient underwent a liver transplantation. On 161 of 657 occasions, 20% HAS was not administered despite a serum albumin level less than 35 g/L, suggesting an adherence rate of 75%.

Infection, Organ Dysfunction, Death, and Duration of Hospital Stay
Clinical end points were reported from days 3 to 15 after study entry (termed the treatment period) to ensure at least 24 hours since the patients first received 20% HAS treatment before measuring possible outcomes. The average time from admission to death or discharge was 16.9 days (SD, 15.7 d; range, 2-102 d). Forty-two of 817 patient days recorded were in the intensive care unit (approximately 5%). Supplementary Table 7 shows the data for days 2 to 15.  Table 4). The patients who had been prescribed antibiotics on admission had increased subsequent nosocomial infection rates compared with those who were not prescribed antibiotics on admission (24% vs 8%, respectively).
Infection case report forms were completed for 35 of the antibiotic prescriptions (either at baseline or after recruitment). A blinded microbiology review showed that 4 of 35 did not fulfill infection predefined criteria (Supplementary Table 4 (Table 5). Of these, the breakdown of components that triggered the composite end point first were as follows.  Table 9).
3. Twelve patients developed respiratory dysfunction as the first end point, with 5 of these patients triggering subsequent other end points (4 infections, 1 died). Of the 7 patients who developed solely respiratory dysfunction, the vast majority had a good outcome, with 6 alive at 30 days, 1 death during admission, and a median stay from respiratory dysfunction diagnosis of only 5 days,    with 3 patients discharged within 3 days of diagnosis with respiratory dysfunction (Supplementary Table 9).
4. Nine patients developed circulatory dysfunction as the first end point, with 3 patients triggering other end points, and 1 patient died during admission. All 6 patients who solely developed circulatory dysfunction were alive at 30 days, and the median stay after circulatory dysfunction diagnosis was 7 days, with 2 patients discharged within 2 days of diagnosis (Supplementary Table 9). 5. One patient died without previously triggering another end point.
Only 1 patient developed grade 3 hepatic encephalopathy after an infection, therefore no brain dysfunction diagnoses contributed to the end point.

Serious Adverse Events
The SAEs (n ¼ 13) during treatment were deemed to be unrelated to HAS by our IDMC (Supplementary  Table 10).

Discussion
Targeted albumin infusions effectively increased serum albumin levels to 30 g/L or greater in hospitalized AD/ACLF patients; a level below which we previously identified as predicting immune dysfunction. 18 Our infusion protocol was effective, acceptable at multiple sites, and deemed safe by the IDMC.
Our suggested protocol allowed dose reduction (eg, if safety concerns), as clinical practice remains variable. 28 We achieved the primary end point with 75% adherence, increasing and maintaining albumin even in patients with very low baseline values. Although a singlearm study, the rates of adverse events of particular concern were low, in particular there were no reports of pulmonary edema. Four variceal bleeds were reported, which potentially can be precipitated by increased portal pressure after albumin. Although not reported in previous albumin trials, this remains a concern. However, a 5% incidence during treatment is similar to expected rates. 21,24,29 The IDMC had no reason to alter our regimen; the absolute safety of our albumin regimen will be determined in our RCT vs standard care.
Our inclusion criteria were broad and straightforward and selected patients with almost exclusively alcohol-induced liver disease, a substantial spread of albumin values, and an ACLF score of 1 to 3 in 25% of cases. We reasoned that our intervention would be more successful in ward-based patients rather than in patients with multi-organ failure, and these criteria appeared to capture exactly this population. The majority were recruited within 2 days of admission, which is crucial because early intervention is widely believed to be imperative to improve outcomes in critically ill patients. As a trial aimed at the prevention of nosocomial infection, the primary composite end point will be reported from day 3 onward in line with established diagnostic criteria, which also ensures at least 24 hours of 20% HAS therapy. The observed in-hospital and 30-day mortality were similar to previous studies. 21,30,31 Survival at 3 and 6 months were not recorded but will be reported in the RCT. We included a medically fit for discharge category at the end of the trial period because many patients have social needs that prolong hospital stay beyond their medical needs.
Our RCT primary composite end point was planned as infection, organ dysfunction, and death because infection commonly triggers organ dysfunction and the combination substantially increases mortality. However, the feasibility of recording such data in a ward-based trial of AD/ACLF at multiple sites is not commonplace. Other than renal failure, 32 there is also no universally accepted definition for early (reversible) organ dysfunction/failures in patients with cirrhosis.
A robust diagnosis of infection in AD/ACLF is challenging because of the high rates of culture-negative sepsis. 26 The on-site clinician-reported infection rate on admission was 34%, which is in line with other studies; however, antibiotics were prescribed in substantially more patients (52%). This perhaps reflects a tendency to overprescribe, as reported elsewhere, 33,34 and therefore using a new/changed antibiotic prescription as a surrogate for infection diagnosis, as originally intended, appeared subject to potential bias and this cannot be standardized across multiple sites. In the RCT we therefore will define infection according to clinician diagnosis, triggering completion of an infection case report form, with a substantial proportion blindly scrutinized by a microbiology panel. This approach appeared feasible in a subset of 35 patients tested. Our data suggest that patients prescribed antibiotics at admission have a 3 times increased risk of subsequent nosocomial infection. This has been reported previously 6 and justifies inclusion of these patients in an RCT to prevent infection; furthermore, an antibiotic prescription will be included in stratification at randomization. Whether our reported nosocomial infection rate of 27% represents a beneficial effect of 20% HAS infusions will be determined in the RCT. Renal dysfunction uses an objective measurement, creatinine, and patients developing this had a poor prognosis, as expected. Renal dysfunction rates were lower than anticipated (9% vs up to z60% 35 ), perhaps related to additional fluid. 30 Cerebral, respiratory, and cardiac dysfunctions were recorded daily. Only 1 patient developed cerebral dysfunction reflecting severe hepatic encephalopathy (!grade 3), suggesting under-reporting, and objective assessment of encephalopathy is recognized as challenging. 34,36 The majority of patients who solely triggered respiratory and cardiovascular end points had a good outcome, with several discharged within a few days. This is counterintuitive because organ dysfunction is a key predictor of poor prognosis. Assessment may be subject to technical difficulties such as a standard-size blood pressure cuff for patients with sarcopenia and a oxygen saturation/fraction of inspired oxygen recording of respiratory dysfunction is influenced greatly by the amount of oxygen delivered by mask or nasal cannulae. We believe our data cast significant doubt over whether these dysfunctions can be recorded accurately in largely ward-based patients across multiple sites and therefore precludes use as part of our RCT primary composite end point; although these will be reported. Table 5 shows the incidence of our revised primary composite end point of infection, renal dysfunction, and mortality.
In summary, our multicenter feasibility study represents a highly pragmatic feasibility study in a challenging group of patients. The data presented confirm that our suggested protocol was deliverable across multiple sites and the dosing regimen increased serum albumin without significant safety concern. Albumin To PrevenT Infection In Chronic LiveR FailurE stage 2 is a multicenter, open-label, interventional RCT and began recruitment in April 2016. This is a phase 3 RCT to verify whether targeting a serum albumin level of 35 g/L or greater in patients admitted with acutely decompensated cirrhosis using repeated intravenous infusions of 20% HAS will reduce the incidence of infection, renal dysfunction, and mortality for the treatment period (maximum 14 days, or discharge if <14 days) compared with standard medical care. The trial will recruit 866 patients at more than 30 sites across the United Kingdom.
Large multicenter studies in patients with advanced liver disease rarely are performed and require significant resources. This feasibility study has enabled substantial improvement in design and end point definition/selection as well as providing important safety information and mechanistic insight. This approach could serve as a model for future trials in such patients.

Supplementary Material
Note: To access the supplementary material accompanying this article, visit the online version of Clinical Gastroenterology and Hepatology at www.cghjournal.org, and at https://doi.org/10.1016/j.cgh.2017.09.012. Clayton, Mr Brennan Kahan, Mr John Crookenden, and Ms Susan Tebbs. A list of trial sites and Principal Investigators can be obtained by contacting attire@ ucl.ac.uk.

Conflicts of interest
The authors disclose no conflicts.