Introduction

Liver-related deaths in England have increased by more than 250% since 1971, and now constitute the fourth commonest cause of premature death [1]. Ascites is the most common complication of cirrhosis (with approximately 40% of patients developing it within 5 years) and the most frequent cause for hospitalisation [2, 3]. Refractory ascites (RA), defined by intolerance or unresponsiveness to diuretics [4, 5], is a useful prognostic indicator as median transplant-free survival is about 6 months [6,7,8,9]. Although liver transplant can be curative, demand for organs continues to outstrip supply [10, 11]. The transplant process involves major surgery and life-long immunosuppression, meaning many patients with significant comorbidities or frailty are not potential candidates [12]. Similarly transjugular intrahepatic portosystemic shunts (TIPS) can be associated with improved survival but carries procedural risks and possible complications including worsening hepatic encephalopathy [13, 14•]. As a result, a significant proportion of patients with RA are reliant on symptomatic relief through repeated hospitalisation for large-volume paracentesis (LVP) [14•, 15]. Alternative management options include automated low-flow ascites pump (alfa pump) and palliative long-term abdominal drains (LTADs), with potential benefits to patient care [16, 17•].

Ascites (including RA) is associated with a high symptom burden and poor health-related quality of life (HRQoL) [18,19,20,21]. Although RA secondary to cirrhosis is being recognised as a potential trigger for palliative care (PC) involvement, there remains paucity of published data with researchers focused primarily on clinical outcomes and the impact of therapeutic interventions on survival [22]. For example, a recent systematic review included 77 studies assessing TIPS in patients with RA even though the majority of these patients will not receive one [14•]. There is a need to shift the focus of research in this group of patients living with this serious life-limiting illness. The aim of this manuscript therefore was to conduct a scoping review assessing outcomes in RA with emphasis on linkage to PC and provision of advanced care planning (ACP). We then conducted a retrospective cohort study in our own centre with a specific focus on PC provision.

Scoping Review Methodology

Eligible studies were defined as any non-interventional, cohort study or case series that reported on clinical outcomes of patients with RA secondary to cirrhosis. Studies had to have a clear definition of RA and report the outcomes of this cohort of patients separately. Authors of studies that included patients with RA but did not report on their outcomes were contacted for the raw data. Studies were identified by conducting a PubMed literature search using the following terms: (refractory ascit*) AND (cirrho*) AND (outcomes OR mortality OR survival).

Scoping Review Results

Six hundred seventy-four titles were reviewed of which 12 studies met our inclusion criteria. Of these 12 studies, one was excluded as the manuscript was in French and four as the authors did not provide the raw data for the cohort with RA. Seven studies were therefore included in the final analysis and data extracted are summarised in Table 1.

Table 1 Results of scoping review showing studies assessing outcomes in refractory ascites

Only four studies were published in the last 5 years. Six studies used the International Ascites Club (IAC) definition for RA in their inclusion criteria, but the methodology in establishing true resistance or intolerance was variable (as expected with the retrospective nature of the studies). One was published before the IAC definitions were established in 1993. The number of patients with RA varied between 24 and 241. One-year mortality was reported in six studies and ranged between 19.2 and 55%. Rates of transplant (1.7–23.2%) were reported in six studies and TIPS (1.0–4.3%) in two.

None of the studies looking at patients with RA mentioned or acknowledged the importance of PC.

Retrospective Cohort Study

Patients and Methods

Our retrospective cohort study was conducted in a secondary care 1100-bedded hospital in southeast of England. All patients coded as having a diagnosis of ascites from 1st of June 2012 to 30th June 2019 were identified by the clinical coding department and electronic hospital records. Initial records (discharge and clinic letters) were screened by MRE and in case of any uncertainty advice was sought from SV.

Inclusion Criteria

RA as defined by IAC criteria [5]. Patients had to have a documented diagnosis of RA made by a gastroenterologist, or three or more LVPs during the study period as a proxy for diuretic intractable ascites.

Exclusion Criteria

  • Non-cirrhosis cause of ascites and/or if aetiology of cirrhosis uncertain

  • Incomplete medical records

  • Ascites occurring after liver transplantation

Baseline clinical and demographic data were extracted from electronic discharge summaries and clinic letters at the time of RA onset and entered onto an anonymised database. These included the aetiology of liver disease, age, sex, use of diuretics, biochemical data, liver prognostic scores, ascitic fluid protein, past medical history including spontaneous bacterial peritonitis (SBP) and antibiotic prophylaxis. Comorbidities were categorised as follows: gastrointestinal disease (e.g. inflammatory bowel disease, pancreatitis), cardiovascular (e.g. hypertension, cerebral vascular accidents and heart failure), respiratory (e.g. chronic obstructive pulmonary disease, asthma or obstructive sleep apnoea), psychiatric (depression, anxiety, PTSD, psychosis), renal, neurological, non-HCC malignancy, diabetes, non-diabetic endocrinology (e.g. hypothyroidism, osteoporosis), infectious diseases and a category of ‘other’.

Study outcomes included referrals to PC and presence of documented end-of-life discussions. Clinical outcomes included number being referred to and receiving TIPS/transplant, number of LVPs and predictors of mortality. Time between RA diagnosis and referral to PC and the time between PC referral and death were also looked at.

Follow-up data was collected until 31st of March 2020 after which patients were censored. This date was chosen to avoid any potential impact that COVID-19 pandemic might have had on clinical outcomes.

Statistical Analysis

Data are summarised using counts, means ± standard deviations, medians (interquartile ranges [IQR]) or percentages. Potential predictors of mortality were defined a priori. These included age, gender, CPS, MELD-Na and UKELD scores; diuretic use, SBP incidence, comorbidity, TIPS/transplant or LTAD insertion. Variables associated with mortality on univariate analysis were included in a multivariable logistic regression analysis. Statistical analysis was performed using Stata 17.

Ethical Approval

As this study was an anonymised retrospective audit, formal ethical approval/informed consent was not deemed necessary by our Research and Development Department.

Results

Of the 1536 patients initially identified as having ascites, 869 were excluded due having a non-cirrhotic cause or incomplete medical records. Of those with ascites due to cirrhosis (n = 667, 43%), n = 88 (13%) meet our criteria for RA and were included in the final analysis. Baseline demographic data of the cohort is summarised in Table 2.

Table 2 Baseline demographic and clinical data in study cohort

This was a relatively young and predominantly male cohort with over two thirds being on diuretics and just under 40% having a prior history of SBP. The predominant aetiology of cirrhosis was alcohol-related liver disease (ALD) in just over 80%. The three most common comorbidities were cardiovascular, diabetes mellitus and psychiatric illnesses (Table 2).

At the time of RA diagnosis, median CPS was 9, while mean MELD-Na and UKELD scores were 17 ± 7 and 55.7 ± 5.2, respectively. About a third had hyponatraemia and/or renal impairment.

The median number of LVPs was 5 (IQR 3–8), suggesting high attendance rates to ambulatory care services. Six patients had over 20 LVPs each. Palliative LTADs were inserted in 11 (12.5%) patients.

Referral for Liver Transplant and TIPS

Overall, 31 patients (35%) were referred for liver transplant and six (7%) for TIPS. Eventually only 11 (13%) patients received a transplant, and one patient received a TIPS (1%). Outcomes are summarised in Table 3.

Table 3 Outcomes in study cohort

Palliative Care Input

Although 56% (n = 49) of patients had documented end of life care discussions, only 33% (n = 29) were referred to palliative care. The median time from RA diagnosis to PC referral was 15 weeks (IQR 4–32), time from PC referral to death being 9 weeks (IQR 4–16) (Table 4). Independent predictors of referral to PC were LTAD insertion (OR 11.0, CI 2.11–57.0, p < 0.01) and age (OR 1.57/10 years, CI 1.04–2.38, p = 0.03). There was a trend for PC referral to be a predictor of death outside hospital, though this failed to reach statistical significance (OR 2.83, CI 0.89–9.04–0.89, p = 0.078).

Table 4 Palliative care input

Mortality

Over a median follow-up of 298 days (IQR 115–125), the overall mortality was 61%, 1-year mortality being 51%. Compared to those receiving standard of care (LVP), receiving transplant/TIPS was associated with a significantly higher 1-year survival (43% vs. 83%, p = 0.013). Compared to those receiving standard of care (LVP), receiving LTAD was not associated with a 1-year survival benefit (43% vs. 36%, p = 0.985). Kaplan–Meier survival curves for each group are shown in Fig. 1.

Fig. 1
figure 1

Kaplan–Meier survival in patients with refractory ascites who underwent liver transplant/TIPS, long-term abdominal drains or standard of care (large-volume paracentesis)

Univariate and multivariate analysis was performed to assess predictors of mortality (Supplementary—Table 4) Independent predictors of mortality were protective effects of undergoing transplant/TIPS (OR 0.15, CI 0.04–0.64, p = 0.01) and having a psychiatric comorbidity (OR 0.32, CI 0.11–0.97, p = 0.044).

Discussion

Our retrospective cohort study and scoping review confirm the poor prognosis associated with refractory ascites secondary to cirrhosis. Our 1-year transplant free survival was 43%, with only a minority receiving curative options. Our mortality rates were higher than some of the studies identified in our scoping review [23, 24] but consistent with those that used more stringent criteria to diagnose RA (IAC) [5, 8, 25, 26].

Despite the high mortality associated with RA, none of the studies identified in the scoping review assessed PC input in this cohort. It is disappointing to note that even in our retrospective study, only a third were referred to PC services, most dying within 2 months of the referral. Our data is consistent with earlier studies that PC delivery in advanced cirrhosis is often too little, too late [27,28,29].

PC is specialised medical care for people living with a serious illness. It focuses on relieving symptoms and stress of the illness, thus improving HRQoL for patients and their families [30]. PC can mistakenly be considered synonymous with end-of life-care. Additional misconception amongst clinicians is that PC necessitates the end of active treatment. However, the modern PC model involves holistic care delivered to patients with a serious illness at any point in their clinical trajectory, including alongside curative treatments [31]. While the precise optimal timing of PC referrals remains controversial, early PC in advanced cancer does appear to improve quality of life and symptom burden [32, 33] and enables acceptable goals and advance care planning to be established. This might be beneficial in patients with RA who have a significantly reduced life expectancy [34].

Approaching challenging end-of-life care discussions in the context of a fluctuating disease trajectory in patients with ACLD can reduce physicians’ confidence in referring to PC [35, 36]. There may also be a perception that PC equates with excluding patients from receiving a liver transplant [37]. Additionally, physicians may have limited awareness of the poor prognosis of RA [29]. Indeed, as our retrospective study has demonstrated, traditional prognostic scores such as CPS, MELD-Na and UKELD were not predictors of mortality in patients with RA. Finally, potential barrier to PC provision in RA is the paucity of evidence-based palliative interventions in advanced chronic liver disease (ACLD).

Encouragingly, there is now increasing recognition of the need for evidence-based palliative interventions in ACLD, supported by national and international guidance [38, 39•]. This has led to randomised controlled trials (RCT) such as the PAL-LIVER study [40] and the advance care planning video study [41]. An earlier feasibility RCT showed preliminary evidence of safety and efficacy of palliative LTADs in RA due to cirrhosis [17•], leading to an ongoing national multicentre RCT (REDUCe 2 Study) assessing this intervention in RA [42]. Additionally, ‘gatekeeping’ needs to be addressed [43] as well as improving communication skills. There also needs to be increased emphasis on PC both at undergraduate and post graduate level. Finally, bringing together multidisciplinary colleagues (British Association for Study of Liver Disease End of Life Specialist Interest Group) [44] can only be a step in the right direction.

Our study does have limitations. Given its retrospective nature, no patient-reported outcomes are available to assess the impact of the PC intervention. There would have also been potential under reporting of both RA development as well as the PC intervention itself. Additionally, some of the results might not reflect true causation, for example, the presence of a psychiatric comorbidity was a significant protector of mortality in our cohort. But this could be the result of a lead-time bias, with improved recognition and documentation of these conditions in more recent years. It is also important to note that the study period overlapped with REDUCE trial [17•] at our centre—a feasibility study looking at the use of LTADs in RA. The study protocol mandated that patients get referred to PC which could have skewed our results and explains the strong association between LTAD insertion and PC referral. Additionally, a higher proportion may have received PC compared to other centres and our results might not be representative. Finally, as source of data was electronic letters rather than actual medical records, this may have resulted in some patients being inadvertently excluded. However, electronic records often do document information that clinicians find most important about patient care and is reflective of what community care teams have access to.

Conclusions

RA is associated with high mortality with only a minority being eligible for curative options. There needs to be a multifaceted approach to improve the end-of-life experience of patients and their caregivers with advanced cirrhosis including those with RA. Our findings suggest patients with RA are denied timely access to PC services. More research into the optimal strategy and timing of interventions is needed to improve the holistic care needs of this cohort of patients.

Supplementary Information.