AT THE INTERNATIONAL LIVER CONGRESS 2016
BARCELONA (FRONTLINE MEDICAL NEWS) – An algorithm that has been developed to help clinicians decide whether or not to refer a patient with end-stage liver disease (ESLD) with organ failure for intensive care treatment predicted mortality in the vast majority of cases, but came under fire during discussion as being potentially “quite dangerous” at the International Liver Congress.
When retrospectively applied to data on 465 patients with end-stage liver disease (ESLD), the algorithm correctly predicted 30-day mortality in 87% of cases.
Determining which patients with severe cirrhosis and acute-on-chronic liver failure (ACLF) should be transferred for intensive care is challenging, explained presenting study author Dr. Katrine Lindvig of Odense University Hospital in Denmark.
“More than 30 studies have looked at mortality in cirrhotic patients with organ failure, and we know that the mortality ranges from 35% to 89% between 15 centers,” Dr. Lindvig said at the meeting sponsored by the European Association for the Study of the Liver. But, she asked, what is an acceptable mortality rate? With limited health care resources was there a point of no return at which referral for intensive care became futile or perhaps unethical?
Dr. Lindvig and colleagues developed an algorithm to help decide if a patient with ESLD is a candidate or not for intensive care based on three scoring systems: the premorbid Child-Pugh Score, the Model of End stage Liver Disease (MELD) score, and the ACLF grade. Data were obtained from four centers in Belgium, Austria, and Denmark, which included two hepatology wards and two ICUs.
A green light for ICU referral was given to patients’ with premorbid Child-Pugh A or B, or a premorbid MELD score of less than 20. Patients with premorbid Child-Pugh C or higher MELD score could also be referred if they had an ACLF grade of 2 or less.
Conversely, patients with Child-Pugh C or a premorbid MELD score of 20 or higher and who had an ACLF grade of 3 or more were given a red light as they would be unlikely to benefit from ICU therapy. Special cases could be discussed, of course, such as those listed for liver transplantation.
Results showed that 394 would get the green light to be referred to ICU and 70 would get a red light, indicating that there would probably be no benefit. Of these 155 (39%) and 53 (75%) in each group, respectively, died within 30 days. Of the 17 patients who survived in the red light group, 12 were still alive at 6 months, and nine were still alive at 1 year.
Comparing the results with the ACLF, there were 35 (30%) of 114 patients with a score of three or higher that were still alive at 30 days and had been coded red by the algorithm. Dr. Lindvig noted, however, that the ACLF was purely descriptive and not developed as a predictive tool.
“However, when we combine the ACLF with the premorbid liver function the algorithm is significantly superior to the ACLF grade alone,” Dr. Lindvig said.
“We have no intention to replace good clinical judgment,” she added. “We simply want to assist the clinical decision making to make the decision more objective.” A computer-based program and app is now being developed in order to optimize and test the algorithm further.
“If the first duty of a physician is to do no harm, then we must continually review our decision-making tools and favor those that have the highest predictive value of treatment success and – importantly – treatment failure,” EASL spokesman Dr. Tom Hemming Karlsen of Oslo University Hospital Rikhospitalet said in a press release. “This study adds to our knowledge of existing, well-recognized scoring systems, and provides an interesting approach for review and wider discussion by the liver community.”
However, the dichotomous nature of the algorithm’s output was criticized during the discussion session by several concerned delegates who felt that it was oversimplified and potentially “quite dangerous” to be giving a green or red light for ICU referral. “There are lots of different variables to consider,” maintained one. Another argued that futility was “clearly not proven” in almost one-third of cases when compared with using the ACLF. Another said that the output should “not be a verdict.”
Nevertheless, others thought that while it was “provocative,” there were perhaps signs of “getting close” to a very good prediction tool.
In an interview, Dr. Lindvig defended the algorithm by reemphasizing that this was envisaged purely as a decision aid. “It is dichotomous, and we are just trying to aid the decision-making process,” she said, noting that this was particularly relevant perhaps for more junior doctors who may have to make a life-saving decision in the middle of the night.
Dr. Lindvig had nothing to disclose.