TORONTO — The risk for decompensation and death from cirrhosis can be accurately predicted when a CT-derived liver surface nodularity score is used in combination with the Model for End-Stage Liver Disease (MELD) score, new research has shown.
"A lot of these patients are already getting CT scans, so once somebody has cirrhosis, it's pretty commonplace to screen them for hepatocellular carcinoma," said Andre Smith, MD, from the University of Mississippi Medical Center in Jackson. However, there is "no good noninvasive method to predict how they are going to do and when they are going to decompensate."
"Having a tool like this that is widely applicable and noninvasive that tells us what they are going to do will be very useful," he told Medscape Medical News.
Dr Smith presented the study results here at the American Roentgen Ray Society 2015 Annual Meeting.
For their retrospective study, the investigators identified a baseline cohort of 858 patients with compensated and decompensated cirrhosis for whom liver CT scans and MELD scores were available. Scores had to have been calculated within 6 months of imaging.
They used custom software to generate a liver surface nodularity score for each patient, which was classified as low (<3) or high (≥3). MELD scores were classified as low (<10), intermediate (≥10 to 20), or high (≥20).
In the baseline cohort, the mean liver surface nodularity score was higher in patients with decompensated cirrhosis than in patients with compensated cirrhosis (3.83 vs 2.82; P < .001).
"As fibrosis progresses to cirrhosis, the nodularity score increases," explained study investigator Cody Branch, a medical student at the University of Mississippi Medical Center.
Liver surface nodularity and MELD scores were combined in a risk model to predict overall survival.
For patients with decompensated liver disease at baseline, 19.2% had a low nodularity score and a low MELD score, and 51.0% had a low nodularity score but an intermediate or high MELD score.
Of patients with a low MELD score but a high nodularity score, 57% had decompensated cirrhosis at baseline. For patients with an intermediate or high MELD score and a high nodularity score, 85% had decompensated cirrhosis at baseline.
For those classified by the risk model as being at low risk, median time to decompensation was 6.6 years; for those classified as moderate risk, median time was 3.8 years; and for those classified as severe risk, median time was 1.6 years.
Median survival in the low-risk group was 4.69 years, in the moderate-risk group was 2.76 years, in the severe-risk group was 1.41 years, and in the critical-risk group was 0.08 years.
In a follow-up cohort of 346 patients with compensated cirrhosis, there were 137 decompensation events.
Again, both decompensation and death were independently associated with nodularity and MELD scores.
Table. Association Between Scores and Outcomes in the Follow-up Cohort
| Outcome | Hazard Ratio | P Value |
| Decompensation | ||
| Nodularity score (per unit increase) | 1.26 | .041 |
| MELD score (per unit increase) | 1.05 | .006 |
| Death | ||
| Nodularity score (per unit increase) | 1.30 | .014 |
| MELD score (per unit increase) | 1.05 | .001 |
Median time to decompensation for patients with a nodularity score below 2.5 was 6.14 years, with a score of 2.5 to 3.25 was 4.35 years, and with a score of 3.25 or higher was 2.26 years.
In a random sample of 48 patients, there was high to very high intraobserver and interobserver agreement on the nodularity score among three readers.
"If somebody is likely to decompensate sooner rather than later, you'd watch them a lot more closely. While this information might not change your treatment plan, it could definitely change the surveillance regimen," said Dr Smith.
"The nice thing about the software is that it's vendor-neutral. It doesn't matter what equipment you did the CT scan with, it works with all machines. I'm pretty excited by it," he said. "In the future, we could watch nodularity scores over time, and that could give us more information than simply assessing the liver at a single time point."
The MELD score by itself can predict death in patients with cirrhosis, so the computer software is not the only tool available, said Erick Remer, MD, from the Cleveland Clinic.
"I think the advantage here is to help better achieve prediction of decompensation," he told Medscape Medical News.
Cirrhosis is still a large clinical problem in the United States as the prevalence of hepatitis B and C increase. "As the prevalence of cirrhosis grows, the need for liver transplants will continue to grow as well," Dr Remer said.
The new antivirals for the treatment of hepatitis C, although highly effective, can cost upward of $85,000 for a single course. Dr Remer explained that it is his understanding that this is the reason the new antivirals are being used "very sparingly" at this point in time.
Dr Smith has a patent pending on the liver surface nodularity software. Dr Remer has disclosed no relevant financial relationships.
American Roentgen Ray Society (ARRS) 2015 Annual Meeting. Abstract 5008.Presented April 22, 2015.