Novel Model Performs Well in Identifying Liver Fibrosis Risk in Patients with Diabetes

Credit: Pexels/ SHVETS production

The fibrosis investigating navigator in diabetes (FIND) score identified patients at risk of liver fibrosis and predicted all-cause and liver-related mortality better than established models.¹

“Our model, trained and validated with [National Health and Nutrition Examination Surveys (NHANES)] data and further tested with liver biopsy data from a Chinese dataset, demonstrated robust predictive accuracy for [liver stiffness measurement] thresholds ≥8 kPa and ≥12 kPa,” wrote investigators, led by Mingkai Li, MD, from Shenzhen Hospital in People's Republic of China.

Type 2 diabetes increases the risk of cirrhosis and liver cancer, making it crucial for this patient population to receive early assessment of liver fibrosis.^2,3 Investigators sought to develop a score to help assessments identify liver fibrosis among patients with diabetes.¹

The team conducted a 3-part study with new model training, validation, and prognostication. The research included 25,740 participants with diabetes after excluding those who did not meet diagnostic criteria, had viral hepatitis, or missed variables for score construction.

Fibrosis was defined as a liver stiffness measurement of ≥ 8.0 kPa, and advanced fibrosis was defined as ≥ 12 kPa. The primary endpoint was the model’s ability to predict a liver stiffness measurement of ≥ 8 kPa, and the secondary endpoint expanded that to ≥ 12 kPa.

In the first part of the study, investigators developed and validated the FIND score using the NHANES dataset from 2017 – 2020, ultimately including 1200 participants.

The second part of the study included 110 patients who underwent liver biopsy at the Third Affiliated Hospital of the Sun Yat-sen University in China between January 2016 and December 2020. Investigators tested the new model against this liver biopsy data, and the Chinese population served as a reference to externally validate the model.

Part 3 investigated the predictive effects of the FIND index by examining 24,430 participants in the UK Biobank cohort from 2006 – 2010. This data was used to assess the model’s ability to predict all-cause and liver-related outcomes in the UK diabetic population.

Participants in the training and validation sets had comparable demographics regarding median age (62 vs 61 years), height (165 vs 166 cm), BMI (31 vs 32 kg/m2, weight circumference (108 vs 108 cm), AST (19 vs 19 IU/L), ALT (19 vs 19 IU/L), GGT (25 vs 26 IU/L).

During the study’s development and validation stage, 21.6% had a liver stiffness measurement of ≥ 8 kPa, and 7.7% had a measurement of ≥ 12 kPa. Among the biopsy validation sample, 62.7% had significant fibrosis and 49.1% had advanced fibrosis. In the UK Biobank cohort, which had a median follow-up of 155 months, 17.9% developed all-cause mortality and 1.2% developed liver-related mortality.

The study showed a moderate linear correlation between FIND and transient elastography results (P < .001), with the FIND index significantly increasing across the 3 transient elastography strata (P < .01). The FIND score performed better at predicting liver fibrosis mortality than other available models, yielding an AUROC of 0.781 (P < .05).

The study also revealed that FIND performed better than other existing models with the greatest AUROC in staging a liver stiffness measurement (P < .01). In the NHANES dataset, the FIND score achieved an AUC of 0.77 (95% CI, 0.744–0.810) for liver stiffness ≥ 8 kPa and 0.821 (95% CI, 0.776–0.866) for ≥12 kPa (P < .05).

When examining the UK Biobank cohort, FIND was linked to an increased risk of all-cause mortality (hazard ratio [HR], 1.75; 95% confidence interval [CI], 1.62 – 1.89) and liver-related mortality (HR, 23.59; 95% CI, 13.67 – 40.69) in patients with diabetes.

A subpopulation analysis revealed that FIND’s diagnostic accuracy was consistent and unaffected by age, sex, alcohol abuse, hypertension, BMI, hepatic steatosis, and serum transaminase levels. FIND also performed strongly in ruling out individuals with liver stiffness measurement ≥ 8 kPa in a lower threshold (0.16) due to the high negative predictive value (91.9%). Compared with the currently recommended FIB-4, FIND had a greater negative predictive value and similar accuracy (P = .011).

“The clinical implications of our study are profound, particularly in the context of managing patients with diabetes who are at risk for liver fibrosis,” investigators concluded. “This model could be instrumental in identifying patients at an early stage of liver fibrosis, thereby facilitating timely intervention and potentially improving patient outcomes.”

References

1. ^{Li M, Yu H, Wan S, Hu F, Luo Q, Gong W. The fibrosis investigating navigator in diabetes (FIND): A tool to predict liver fibrosis risk in subjects with diabetes. Diabetes Obes Metab. 2024 Dec 5. doi: 10.1111/dom.16109. Epub ahead of print. PMID: 39638772.}
2. ^{Garcia-Compean D, Jaquez-Quintana JO, Gonzalez-Gonzalez JA, Maldonado-Garza H. Liver cirrhosis and diabetes: risk factors, pathophysiology, clinical implications and management. World J Gastroenterol. 2009 Jan 21;15(3):280-8. doi: 10.3748/wjg.15.280. PMID: 19140227; PMCID: PMC2653324.}
3. ^{Conger, K. Diabetes and liver cancer — Stanford Medicine study suggests new screening guidelines. Stanford Medicine. February 1, 2024. https://med.stanford.edu/news/all-news/2024/02/diabetes-liver-cancer.html#:~:text=The%20distinction%20matters%20because%20people,in%20the%20absence%20of%20cirrhosis. Accessed December 11, 2024.}