Review article: towards a considered and ethical approach to organ support in critically-ill patients with cirrhosis
- This uncommissioned review article was subject to full peer review.
Dr P. A. Berry, Department of Gastroenterology, Hepatology and Nutrition, Frimley Park Hospital NHS Foundation Trust, Portsmouth Rd, Frimley, Surrey, GU16 7UJ, UK.
Increasing numbers of patients are being admitted to hospital with decompensated chronic liver disease in the UK. A significant proportion will develop complicating extra-hepatic organ dysfunction, but the selection of those who should be admitted to intensive care is complex and challenging. Alcohol-related liver disease also presents complex ethical dilemmas.
To review recent survival analyses and explore differences in secondary and tertiary care; to highlight strengths and weaknesses of prognostic models, therapeutic advances and shifts in prognostic expectation. We also aim to explore the ethical challenges presented by addiction and self-injury in an area of limited resource.
We searched PubMed for articles discussing ‘cirrhosis’, ‘prognosis’, ‘critical illness’, ‘organ failure’, ‘renal failure’, ‘alcohol’, ‘ethics’ and ‘addiction’. We also explored particular ethical dilemmas encountered by the authors and colleagues.
Prognosis has improved in many cirrhotic complications and historically poor outcomes in tertiary care may reflect a more complex patient cohort. Previously ‘untreatable’ complications are now being managed successfully. Estimates of survival are more accurate after a 48-h period of supportive care. Physicians are not best placed to make judgments with regard to deservingness, moral responsibility, rationing and access to organ support in cases of acute deterioration related to alcoholism, and the case for denying support must be made on purely medical grounds.
An early, aggressive approach to organ support is justified. Further discussions between hepatologists and critical care physicians are required to determine acceptable burden-to-benefit ratios for prolonged intensive care support in young alcoholic patients.
Background: Incidence of Decompensated Chronic Liver Disease and Inevitability of Multiple Organ Dysfunction
Liver disease accounts for an increasing number of hospital admissions in the UK. Between 1990 and 2003, admissions for chronic liver disease (CLD) rose by 71% in males and 43% in females. These changes were largely driven by alcohol-related liver disease (ARLD) with rates more than doubling in all age groups. There has been a five-fold increase in progression to cirrhosis in 35- to 55-year olds over the last 10 years. Between 1996 and 2005, the percentage of ICU admissions in England and Wales with ARLD rose from 0.65 to 1.35%, the estimated number of bed days rising from 3100 to over 10 000. A recent study from the Intensive Care National Audit and Research Centre (ICNARC UK) reported a rise in the proportion of ICU patients with a diagnosis of cirrhosis between the two eras of 2003–2005 and 2006–2008, from 2.8 to 5.4%. There is also a significant burden of hospital re-admissions in this patient group, further emphasising the need for a greater focus on early intervention and close follow-up. If these trends, driven by epidemics of NASH, alcohol misuse and hepatitis C continue, it can be anticipated that the number of patients who become critically ill with extra-hepatic organ failure will also rise, many of whom will be relatively young. Critical care facilities should therefore anticipate a significantly increased demand for organ support.
The liver's position at the apex of multiple synthetic, detoxifying, metabolic, immunological and hormonal processes predisposes to a number of complications. The most important of these are immunoparesis, renal failure and neurological dysfunction. Further liver injury often ensues in the context of an exaggerated systemic inflammatory response, due to alterations in visceral perfusion, impaired microvascular integrity and dysregulated hepatic cellular mediators.[6-8] Although support devices exist in other forms of essential organ failure, there is no reliable technology available, thus far, for liver failure. The incidence of multiple organ dysfunction (MOD) in decompensated liver disease is therefore high and, without transplantation resolution, can only occur if other acute precipitants (e.g. infection) are treated while vital extra-hepatic organs are supported.
Historical Prognosis and Therapeutic Pessimism
Historically, observational intensive care (ICU) studies have reported mortality rates for patients with CLD and critical illness of greater than 60%.[9-15] Most strikingly, a 1983 publication by Goldfarb et al. reported 89% ICU mortality in a group of 100 mechanically ventilated CLD patients. The group went on to report a statistical probability for ‘fatal evolution’ of this scenario of between 95% and 100%. A recent analysis of over 16 000 general ICU patients with cirrhosis admitted to 192 ICUs over a 13-year period found that overall ICU mortality was 65%, rising to around 90% with sepsis if more than 1 day of respiratory support and renal support were required. Secondary care data, which may reflect a less highly selected cohort of patients with CLD, has challenged this rather fatalistic message, demonstrating ICU and hospital mortality rates of only 38% and 47% respectively. This study also highlighted how important it is to differentiate MOF from a patient requiring single organ support (21% mortality for single organ failure vs. 63% for three or more failing organs). Although not reaching significance, mortality was also lower in those patients ventilated for airway protection rather than respiratory failure (33% vs. 67%).
Therefore, a general perception of poor outcome for patients with cirrhosis who require organ support remains prevalent in clinical practice. This cannot be shown to have translated to bias in the prioritisation of patients for admission to ICU as such data would be very difficult to source; however, most hepatologists will have confronted such prognostic pessimism when advocating for the admission of their patients, sometimes disproportionate to the actual degree of organ failure that has evolved. An over-simplistic and unchallenged approach to ‘organ failure’, if universally regarded as carrying a mortal prognosis, will not adequately discriminate between degrees of severity. There is a danger, in the authors' view, that such an approach will lead to potentially salvageable patients being denied life-saving therapy.
Renal failure – truly the end of the road?
Acute renal failure in cirrhosis, especially hepato-renal syndrome (HRS) in its rapid onset manifestation, has long been associated with a dismal prognosis. Historically, mortality has exceeded 50% at 1 month. However, not all renal failure in cirrhosis is HRS, and the exact aetiology of renal injury is of paramount importance with respect to prognosis. For instance, a large study demonstrated that sepsis and/or hypovolaemia accounted for 78% of renal failure cases and HRS for only 13%, 3-month mortality being twice as high in the former groups (31–46% vs. 15%). It is vital therefore that effort is made to determine the cause while resuscitative therapy is instituted. For those with definite HRS, the outlook has recently improved, a meta-analysis of four trials using Terlipressin and Albumin demonstrating consistent reversal in renal failure (relative risk: 3.66, CI: 2.15–6.23) with only an 8% risk of recurrence, and a trend towards improved survival at 90 days (relative risk: 1.86, CI: 1.0–3.4). A large systematic review of overall mortality in renal failure recently demonstrated that 58% die within 30 days, 12-month mortality rising modestly to 63%. If patients can be identified early and supported aggressively through the first month of illness, there is reason to expect improved survival. Indeed, the study showed a reduction in overall mortality from 74 to 63% since 2005, 30-day death rates differing further still (71% vs. 52%).
Transplantation for alcoholic hepatitis – salvaging the unsalvageable
Perhaps the most remarkable shift in expectation, and therapeutic paradigm, is in the case of urgent transplantation for severe alcoholic hepatitis. Historically, urgent transplantation has been reserved for patients with acute liver failure; that is patients who develop jaundice, coagulopathy and encephalopathy, on a background of entirely normal liver function. The definition is further subdivided into hyperacute (0–1 week), acute (1–4 weeks) and subacute (4–12 weeks) according to the time of encephalopathy onset.
Alcoholic hepatitis causes acute liver dysfunction, but by definition occurs after years of heavy drinking and is commonly associated with histological features of chronic liver injury. Patients are not therefore eligible for urgent LT. However, patients tend to be young and often have healthy hearts, lungs and kidneys, with good nutritional reserve. Prior to acute presentation, they might otherwise have been regarded as strong surgical candidates. A French team have begun to offer transplants in such cases, with, so far, acceptable results. Twenty-six patients were transplanted with 77% surviving 6 months. Survival benefit was maintained for the 2-year follow-up with only three cases of recidivism. If this trend continues, it is conceivable that patients for whom there was no hope previously will be selected for transplantation; however, it could be anticipated that only 1–2% of transplants would be for this indication if similar criteria were accepted generally. Although in early stages of development and yet to become standard of care, this change in approach demonstrates the gains that can be accomplished by treating such patients aggressively.
Bedside Prognostication in Acutely Decompensated Disease – an Inexact Science
Great efforts have been made to quantify and grade the severity of liver disease such that a more accurate prognosis can be arrived at in individual clinical scenarios. However, the variable performance of prognostication scores, especially in the early phase of presentation, is increasingly recognised.
Liver impairment is commonly quantified by scores such as Child-Pugh (CPS); Model for End Stage Liver Disease (MELD) or the United Kingdom Model for End Stage Liver Disease (UKELD). These scores use biochemical markers of hepatic synthetic function (bilirubin, albumin, prothrombin time), renal function (creatinine, sodium) and clinical sequelae of cirrhosis (ascites, encephalopathy) to evaluate disease severity. When used in the stable setting, they give accurate prognostic information with regard to 1-year survival and are used to guide organ allocation in the transplant arena. Recent EASL guidelines propose that organ support be offered to those with pre-morbid MELD <15, but questioned if MELD is >30 and there is ≥3 organ failure. However, the decline in liver function that accompanies episodes of acute illness can result in short-term elevation in these scores and it is therefore possible that they will give an inaccurate impression as to the patient's position in the natural history of disease. For example, in a cohort of 84 patients with cirrhosis admitted to a general intensive care, the MELD score increased by a median of six points and Child-Pugh score by 2.5 points compared with baseline outpatient data. This shifted 58% of grade ‘A’ patients to grade ‘B’ and 35% to grade ‘C’. A further 72% of grade ‘B’ upgraded to ‘C’ (Table 1). Although indicative of illness severity in the stable setting, these scores therefore remain unreliable in the scenario of acute critical illness.
Table 1. Child-Pugh score
Review of the literature confirms that CPS does not perform, as well as general critical illness scoring systems. From eight studies, the area under the receiver operator curve (AUROC) for CPS ranged between 0.7 and 0.75.[9-11, 13, 14, 28-30] MELD has been evaluated three times and performed marginally better (AUROC 0.78, 0.81 and 0.88) (Table 2).[11, 28, 29] In contrast, the Sequential Organ Failure Assessment (SOFA) score has produced AUROC values of 0.81–0.95 on the seven occasions it has been evaluated.[9, 11, 14, 28-31] Despite not including many CLD patients in the initial development datasets, current trends suggest that the general critical illness scores, specifically SOFA, are in fact more reliable for evaluating illness severity in this setting (Table 3).[32, 33]
Table 2. MELD calculation and prognosis
Table 3. SOFA score
|Mean blood pressure/inotrope requirement||>70||<70|| |
Dopamine <5 mc/Kg/min
or Dobutamine any dose
Dopamine >5 mc/kg/min
Adren/Noradren <0.1 mc/min/min
Dopamine >15 mc/kg/min
|Glasgow Coma Scale||15||13–14||10–12||6–9||<6|
|Creatinine or urine output||<110||>110||>171||>300/<500 mL/24h||>400/<200 mL/24h|
There is also evidence that patients who require ICU admission and recover prior to transplant (although not those transplanted out of ICU) have an equally good long-term prognosis. In a recent observational study arising from a tertiary French ICU, an 86% 6-month survival was reported in those patients successfully discharged from hospital. Therefore, it is the reversibility of the acute disease that is the most important factor and ‘end of the bed’ assessments of liver function made during inter-current illness may not be wholly representative of short-term prognosis.
More accurate identification of patients with improved prognosis at 48 h or greater
Recent data suggest that a more accurate prognosis can be achieved if assessment of liver and extra-hepatic organ function is made after a period of support.[28, 32] In 128 patients with cirrhosis admitted to ICU, mortality was 93% in those with SOFA score >10 at 48 h. AUROC was 0.88 compared with 0.81 at the point of admission. In the study by Das et al., liver disease severity did not predict mortality, and in those alive after 3 days only SOFA score performed usefully (Table 4). This phenomenon of improved utility over time has also been seen during studies into immunoparesis, thought to be a major determinant of outcome in cirrhosis-related multiple organ failure, where analysis of monocyte antigen presentation capacity and the immunosuppressive cytokine IL-10 after 48–72 h of support have improved prognostic accuracy.[6-8]
Table 4. SOFA score and prognosis in during first 48 h in medical ICU (from Ferreira et al.)
The data suggest therefore that making a clinical assessment of disease severity and outcome at a single point in time is potentially fraught with difficulties. Instead, a ‘stabilise and assess’ approach to complex patients to help define and identify treatment ‘responders’, might be advocated.
Alcohol: Professional Approach, Self-Harming Behaviour, Recidivism and Moral Responsibility
Refusing to admit a patient to ICU for organ support on the basis that ARLD is self-inflicted is ethically indefensible. Many diseases are treated aggressively despite strong links to voluntary behaviour. The General Medical Council ‘Good Medical Practice’ guidance is explicit on this:
‘The investigations or treatment you provide or arrange must be based on the assessment you and the patient make of their needs and priorities, and on your clinical judgement about the likely effectiveness of the treatment options. You must not refuse or delay treatment because you believe that a patient's actions have contributed to their condition…'.
The importance of abstinence, although self-evident in part, has been emphasised in studies that demonstrate life expectancy following diagnosis of cirrhosis depends not on the severity of liver injury, but on whether or not alcohol ingestion continues. In a study by Verrill et al., drinking status 1 month after diagnosis was the most important determinant of prognosis, with 7-year survival of 72% for the abstinent patients vs. 44% for recidivists. The patient who deteriorates during their ‘first presentation’ may well be given the benefit of the doubt. However, others who have been counselled, but who have continued to drink, or stopped and lapsed are often regarded less favourably.
Recidivism is due to overwhelming cravings for alcohol. The power of the individual to resist these cravings is equated, in the eyes of some authors, to their degree of ‘moral responsibility’. One argument is that there is ample time during an alcoholic's life to receive advice and comprehend the damage that it is doing; therefore, the decision to continue must involve a personal acceptance that the individual will become ill. There is also a perceived responsibility that individuals should actively seek help once it becomes clear that alcohol addiction has developed and those who do not must bear some responsibility for this failure. Those who do, but for whom attempts to break the addiction fail, might be regarded in a more positive light. It is not necessarily their ‘fault’ that the treatment of the addiction, a medical affliction in its own right, failed. Therefore, one might differentiate the patient who has never engaged in supervised detoxification or support programmes from those who have tried but failed.
An additional consideration is the quality of support that has been offered. The National Institute for Health and Clinical Excellence (NICE) has only recently published guidelines for identifying, treating and supporting alcohol-addicted patients. The interventions suggested (pharmacological, psychological and psychosocial) demand the provision of many suitably trained staff. It is likely that many patients presenting now will not have had the benefit of such structured therapy.
Complicating the attempt to attribute moral responsibility retrospectively even further are the numerous unasked for environmental and social factors that encourage alcohol addiction. Many alcoholics harbour tragic personal and family histories. If it is accepted that alcoholism is indeed a function of childhood and upbringing, it becomes very hard to deny treatment to a patient with liver disease secondary to that diagnosis while offering treatment to someone with cirrhosis due to other aetiologies beyond personal control, for example hepatitis B contracted at childbirth. More proximate still is the evidence that a genetic variation in the aldehyde dehydrogenase 2 gene influences the risk of developing alcohol dependence.[41, 42] Female gender also confers a greater risk of dependence  and recent laboratory work has started to identify other potential genetic associations (e.g. PNPLA3 polymorphism) with the development of alcohol-related liver disease. Alcoholism may therefore be inherited in part.
There is also the growing cohort of patients with ARLD for whom alcohol misuse has never been considered a detrimental feature in their lives. Many clinic patients harbour an alcohol history, which manifestly exceeds recommended guidelines, but have continued to live entirely ‘functional’ lives. From their perspective, the index diagnosis of cirrhosis is often unexpected and the implications of the term ‘alcoholic’ liver disease are frequently challenged. While ‘alcoholic’ patients are often lacking in advocacy and representation, this cohort can present with more pro-active next of kin whose expectations of outcome following acute presentation may be high.
Estimating the degree of moral (ir)responsibility, allowing this to influence the decision to offer treatment and explaining this to relatives are practices fraught with error and danger, and are arguably outside the skill set of the physician in an emergency situation. However, to disregard the history of alcoholism entirely before assigning priority to a patient appears naïve; a middle ground must be achieved whereby the physician avoids making ill-informed decisions as to moral responsibility and focuses instead on the practical question of whether organ support will in fact benefit the patient in the long-term.
Futility, Rationing and Public Opinion
We have seen that to deny admission on the basis of alcoholism alone is unjustifiable. A more nuanced objection to admitting such patients is that treatment may be ultimately futile. If the disease of alcoholism is manifestly ‘incurable’, as demonstrated by recidivism, a cogent argument can be formed that rescuing the patient from organ failure will not result in long-term benefit; liver injury will continue on the patient's return home, and death will inevitably ensue. In this scenario, a judgment is being made that discharge from hospital and a return to drinking is not an acceptable outcome. The patient, if asked this question and able to answer, could of course disagree (unless they had made an advanced directive). Futility in this context is therefore a very subjective, physician-centred quantity. This is an important point because the argument that ‘it's futile, they will just go back to drinking anyway’ has been used to deny ICU admission in the authors’ experience. A futility or, more positively a ‘utility’, calculation is required for all prospective ICU candidates. Estimates of short-term prognosis and the chances of success are made at the bedside according to various models and with clinical experience. However, medium-term futility calculations that invoke the patient's predicted likelihood of returning to self-harming behaviour is not applied in cases of, for example, smoking-related disease (e.g. obstructive airways, coronary artery). Projections of behaviour in ALD that are used to inform the immediate decision to admit or deny admission to ICU are therefore potentially unsafe and ethically unsound. The concept of futility as it applies to individual patients and the decision to ration services requires further investigation.
Analysing the relationship between futility and rationing, Jecker & Schneiderman wrote that whereas
‘medical futility signifies that a treatment offers no therapeutic benefit to a patient… rationing specifically acknowledges that a treatment does offer a benefit, and the issue becomes how to distribute beneficial but limited resources fairly. To clarify the distinction further: futility decisions are made at the bedside of a specific patient, whereas rationing decisions, involving categories of patients or treatments or circumstances, inevitably should be made at a policy level in order to assure just distribution of resources.’
Policy is informed by public opinion, and it must be accepted that when members of the public are asked to allocate healthcare resources in hypothetical clinical scenarios, patients with ‘self-inflicted’ disease are regarded more harshly. In Oregon USA, from a list of 714 paired diseases and treatments, liver transplantation for ARLD was ranked 19 from bottom (position 695), whereas transplant for a non-alcoholic aetiology was placed at position 364. A study that assessed views of primary care physicians and secondary care specialists e.g. gastroenterologists regarding LT specifically found that a hypothetical male patient with ARLD was ranked second from bottom within a list of 8 cases, only the man imprisoned for violence scoring lower. A third study confirmed these observations: patients with ARLD are generally regarded with lower priority than those with non-alcohol aetiologies.
Although studies looking at prioritisation for admission to ICU have not been performed, it is reasonable to anticipate that the public, and physicians, would maintain similar views. How can we expect physicians who have to juggle scarce facilities on a daily basis to promote the admission of increasing numbers of alcoholic patients in the face of such data? Rationing is not the job of individual physicians. Their single duty is to act in the best interests of those who come under their care. Rationing is an economic process, conducted by governments and budget administrators away from the frontline, where appropriate strategic decisions can be made. This may sound naive: how, for instance, should a doctor with one available bed on ICU decide between two patients, one alcoholic and one not. The conundrum is a false one. In general, such ‘his or her, this patient or that patient’ decisions rarely occur, because short temporary increases in ICU capacity can be made. In any case, it is not the ‘alcoholism’ that should be separated out from the equation and scrutinised, but other factors that are more pertinent to the potential benefits of organ support. In the case of ARLD, co-morbidities are especially prevalent and, once identified, can certainly support a futility based argument. This consideration would apply to any critically ill patient, but is worth some emphasis because relatives should be made aware that it is not alcohol alone that is the deciding factor. It is seen therefore that rationing should not enter the equation when an individual patient is being considered, for this cannot be in the patient's best interest. A decision based on futility, whereby that patient is spared a burdensome admission, is a wholly patient-centred process.
Lack of Strategy Regarding Treatment Withdrawal
In promoting a more aggressive stance towards organ support in these patients, it is necessary to present a strategy for treatment withdrawal in those who do not respond. A patient may continue to deteriorate, with escalating vasopressor requirements, ventilator demands or metabolic derangements such as acidosis, in which case, as with other types of patients, a decision is usually made with the family to limit further increases in therapy or to withdraw. Alternatively, improvements are seen within 48 h by way of reduced levels of support (reflected in falling SOFA or APACHE scores) and it is clear that treatment should be continued. The more difficult challenge is how to manage the patient who achieves a degree of stability with multiple organ support, but who shows no signs of sustained improvement. This common scenario is one that must be recognised and addressed by the advocating hepatologist, so that death, if manifestly inevitable, is not delayed unnecessarily. However, thresholds in terms of underlying liver function or degree of ongoing organ failure have not been identified.
Recent reviews of outcome in cirrhosis and society guidelines have addressed this issue. A recent EASL guideline states: ‘…the persistence of 3 or more [organ] failures after 3 days spent in the ICU may lead to consider a limitation in life-sustaining treatments as a fatal outcome is almost constant.’ The recent review by O'Brien et al. suggests that ‘given the extremely high mortality in patients with multi-organ failure, support should be limited/withdrawn in such patients.’ Details as to what parameters indicate failure to improve are lacking, however. It would appear perverse to withdraw therapy in a patient who remains dependent on mechanical ventilation, norepinephrine and haemofiltration at day 3, but whose Fi02 and vasopressor doses had reduced by 75% in the last 24 h. Further work focussing on the precise trends in support parameters (e.g. vasopressor requirements, ventilatory support settings/FiO2) that correlate with ultimate survival rather than ‘organ failure – yes/no’ is much needed in this context.
Towards a strategy for accurate and fair identification of patients who may benefit from ICU admission
In light of the available evidence and discussion above, we suggest that the following propositions require further debate in the hepatology and critical care communities:
- An awareness that recent therapeutic developments have resulted in improvements in prognosis, and that previously ‘untreatable’ conditions are now being considered for aggressive therapy. We should strive to ensure equity of access to early specialist input for all patients presenting acutely to secondary care to temporise against further physiological decline and organ dysfunction, and hence, critical care requirements.
- An acceptance that CLD and ARLD patients’ relative youth may justify a recalibration of the benefit–burden ratio with respect to intensive care treatment. The gains to be made in ensuring the survival of a younger patient may outweigh the burden, and cost, of prolonged organ support, if a clinical response is seen.
- An understanding that judgments concerning moral responsibility, although valid in an academic sense, cannot be made safely in the acute setting. Especially unsatisfactory are rationing decisions, with respect to critical care beds, in which the perception of ‘addiction’ is allowed to influence prioritisation.
- An acceptance that prognosis made after a period of organ support is more accurate than that made in the first 24 h of admission. Therefore, serious consideration should be given to offering patients with the most deranged physiology and evidence of poor liver function a ‘trial’ of therapy.
- Agreement as to what constitutes a clinical response to therapy, and what are the features of a non-improving patient. ‘Stability’ at a very low level of hepatic or extra-hepatic function may not constitute survivability. It may be constructive to agree, at the outset, with family and involved clinicians, at what point this judgement will be made and what will be done (e.g. treatment withdrawal).
Guarantor of the article: Dr Berry.
Author contributions: Dr Berry conceived the article. Drs Berry and Thomson wrote the article and additional refinements were made by Drs Rahman and Ala. All authors approved the final version of the manuscript.
Declaration of personal interests: None.
Declaration of funding interests: Dr Ala is supported by the NIHR/CLRN Surrey and Sussex, UK.