Nonalcoholic fatty liver disease (NAFLD) is today the most common cause of chronic liver disease and second only to viral hepatitis as a cause of liver transplantation in the U.S. (1,2). It encompasses conditions from simple steatosis (NAFL), believed to be associated with slow disease progression, to the more severe and progressive form known as nonalcoholic steatohepatitis (NASH). NASH is characterized by hepatocellular injury in the form of hepatocyte ballooning (necrosis) and predominantly lobular inflammation. The severity of hepatic fibrosis is defined in stages. They range from stage F0, or no fibrosis, to mild (stage F1), moderate (stage F2, with zone 3 sinusoidal fibrosis plus periportal fibrosis), or advanced fibrosis, with bridging fibrosis (stage F3) or cirrhosis (stage F4). NASH may lead to cirrhosis and to the development of hepatocellular carcinoma, but even moderate-to-severe fibrosis (F2-F3) is associated with higher mortality (1,2). Advanced liver fibrosis and cirrhosis occur more often in obesity but, in particular, in patients with type 2 diabetes (T2D) (3). Endocrinologists should be aware that patients with NAFLD are also at a two- to threefold increased risk of both progression from prediabetes to diabetes and development of cardiovascular disease (4,5). Taken together, there is a consensus that patients with T2D and NASH are at a much higher risk of hepatic and extrahepatic morbidity and premature death than in the absence of liver disease.
Within this context, Younossi et al. (6) report in the current issue of Diabetes Care an important study on the clinical and economic burden of NASH in patients with T2D in the U.S. This is so far the most comprehensive effort to systematically outline the magnitude of the problem in patients with diabetes. The authors used 2017 annual direct medical costs attributed to diagnosed diabetes reported by the American Diabetes Association (7) and applied prevalence rates and well-validated statistical models from prior work in populations with NASH (3,8–10). The results were staggering for anyone involved in the care of patients with diabetes. The overall prevalence of NAFLD was >70% (47% with NAFL plus 26% with NASH), for a total of >18 million patients with T2D having NAFLD (not including patients in the U.S. with undiagnosed T2D). The economic burden was, as expected, driven by diabetes care, as about two-thirds had simple steatosis (NAFL), which infrequently will develop into advanced liver disease. However, health care expenses were substantially higher in those with NASH. The total liver-related cost of NASH versus NAFL was about 24 times higher at $2,275 versus $95 per person-year, respectively. The economic burden for the group with prevalent NASH and T2D was $642 billion, with $482.2 billion (75%) attributable to diabetes care and $160.3 billion (25%) to NASH-related liver care. Over the next 20 years, NASH and T2D would be potentially responsible for 64,900 liver transplants (29% of the total estimated liver transplants performed), 812,000 liver-related deaths, 1.37 million cardiovascular deaths, 1.27 million cases of decompensated cirrhosis person-years, and 479,000 hepatocellular carcinoma person-years. Weaknesses of the study include those intrinsic to the assumptions of any disease model. The true prevalence of NASH will not be exactly known in the foreseeable future until a fully reliable noninvasive test is available. Systematic liver biopsies for all potential patients with NASH (i.e., anyone with NAFL) would expose many patients to an unnecessary risk and would be clearly unethical. Still, for this study the authors used data from the best sources available on both T2D and NASH. The model was conservative as it did not include potentially higher diabetes care costs from worse diabetes micro- and macrovascular complications from having NAFLD/NASH (5,11,12). A second limitation is that the natural history of steatohepatitis and liver fibrosis is still unclear, in large part because of the same diagnostic/monitoring limitations as exist for estimating the prevalence of NASH (i.e., doing repeated liver biopsies over time). However, recent prospective long-term follow-up studies from the multicenter NASH Clinical Research Network (CRN) consortium (13,14) appear to confirm the assumptions included in the model by Younossi et al. (6).
The above results call on endocrinologists to view NASH as a frequent and serious complication of T2D and to be proactive in the early identification of patients at risk for liver fibrosis. However, the challenge remains how to separate patients with diabetes with the more benign form of the disease (NAFL) from those who have steatohepatitis (NASH), as well as those with or without moderate-to-severe fibrosis (≥F2) that would benefit from aggressive lifestyle intervention and therapies such as pioglitazone or glucagon-like peptide 1 receptor agonists (GLP-1RAs). There are several diagnostic algorithms with the goal to identify patients with NASH fibrosis, as reviewed elsewhere, some more focused on patients from hepatology clinics (15,16) or for primary care physicians and endocrinologists (17). They are based on a combination of blood testing and imaging (usually elastography either at point-of-care [Fibroscan] or by magnetic resonance). Blood diagnostic panels combine clinical demographics (BMI, T2D) with routine chemistries, while there are also specific commercially available plasma biomarkers to diagnose NASH or fibrosis. To summarize a large body of data, noninvasive plasma tests have had limited accuracy to diagnose NASH or early stages of fibrosis but have been useful for the diagnosis of advanced disease (≥F3). In other words, they are used best to rule out severe disease (good specificity and negative predictive value) than to establish an early diagnosis (they usually have modest sensitivity and positive predictive value) or monitor the disease (1,2,18). Another caveat is that they usually perform better in hepatology clinics, where there are more patients with advanced liver fibrosis and cirrhosis, than in nonhepatology settings, where cirrhosis is less common (19–22). Finally, few studies have focused only on patients with T2D.
Acknowledging the above gaps, the study by Bril et al. (23) in this issue of Diabetes Care did a head-to-head comparison of the most commonly used plasma biomarkers and diagnostic panels to establish their true value in 213 patients with T2D not being followed in a hepatology setting. Their key findings were that for the diagnosis of NASH none of the currently available panels or biomarkers (cytokeratin 18 [CK-18], NashTest 2, HAIR, BARD, or OWLiver) was able to outperform plasma ALT (area under the curve [AUC] 0.78 [0.71–0.84]), while for advanced fibrosis (≥F3) none of the plasma tests (fragments of propeptide of type III procollagen [PRO-C3], APRI, FIB-4, Fibrotest, or NAFLD fibrosis score) was significantly better than plasma AST (AUC 0.85 [0.80–0.91]). These results suggest that plasma ALT and AST as stand-alone tests can be helpful, although they often fall short of clearly guiding management on an individual basis. On the more positive side, plasma PRO-C3 showed a trend to be better than plasma AST (AUC 0.90 [0.85–0.95] vs. 0.85 [0.80–0.91]) and held hope that sequential use of plasma AST (≥26 IU/L) followed by PRO-C3 or future biomarkers may help limit the number of liver biopsies. Still, a high number of patients (about one-third) would have needed a liver biopsy. Of note, noncommercial diagnostic panels such as FIB-4, NAFLD fibrosis score, and APRI also performed relatively well compared with PRO-C3. The limitations of the study were the relatively small sample size and its cross-sectional nature that did not allow conclusions to be drawn regarding the value of screening relative to disease progression or clinical outcomes. However, it contributes to the field about the value of AST/ALT in a population with diabetes and by calling for more reliable tests for the diagnosis of NASH and advanced fibrosis in T2D. The overall poor results speak to the complex biology of liver fibrosis, where both fibrogenesis and fibrinolysis determine fibrosis progression over time with fluctuations depending on profibrotic stimulus (24,25). Significant work is being done in the field to validate novel and more sophisticated fibrosis biomarkers (26). Future studies will help us enter a new era of precision medicine where biomarkers will identify and target therapy to those with more active disease at risk for cirrhosis.
The above sets the stage for discussing what treatment options we currently have for patients with T2D and NASH. Vitamin E is effective in patients with biopsy-proven NASH without diabetes (1,4) but has had more mixed results in a recent proof-of-concept randomized controlled trial (RCT) in patients with T2D (27). A larger, long-term study may be needed to clarify its role in this setting and dispel lingering safety concerns. Given that diabetes and NASH often overlap, an approach that treats both would be logical and cost-effective. Among pharmacological agents approved by the U.S. Food and Drug Administration (FDA) for the treatment of T2D, the insulin sensitizer metformin has not shown clinical efficacy (1,4). In contrast, there are now five RCTs where pioglitazone has consistently improved steatohepatitis, with a treatment difference of about 30 to 40 percentage points compared with placebo (27–31). A modest effect on fibrosis has been reported in some of these trials (28,30). The thiazolidinedione has been incorporated as an option into current liver (1,2) and diabetes (32) treatment guidelines. In RCTs, dipeptidyl peptidase 4 inhibitors have been largely negative for the treatment of NASH (33,34; reviewed in 4,35,36). However, in several small proof-of-concept RCTs, GLP-1RAs have been reported to normalize plasma aminotransaminases and decrease hepatic steatosis (35,36) and even improve liver histology (37,38). Most studies have used liraglutide and the benefit has usually been proportional to the degree of weight loss, although other mechanisms may be at play. Results from the semaglutide NASH trial in the second half of 2020 are awaited with significant expectation (ClinicalTrials.gov reg. no. NCT02970942).
Studies in animal models of NAFLD (39–41) and uncontrolled clinical trials (42–47) support the notion that sodium–glucose cotransporter 2 inhibitors (SGLT2i) could play a valuable role in NAFLD. However, only recently more carefully designed RCTs have assessed their safety and efficacy in patients with T2D (Table 1) (48–52). In this issue of Diabetes Care, Kahl et al. (52) report on the first RCT with empagliflozin, where 84 well-controlled patients with T2D (baseline HbA1c 6.6% ± 0.5%) were randomized to empagliflozin or placebo for 24 weeks. The study included state-of-the-art liver fat (1H-MRS) and metabolic measurements. The main findings were a 22% reduction in liver fat (P = 0.009 vs. placebo) associated with a 2.5-kg placebo-corrected weight loss (∼2.4%). Treatment did not improve hepatic, muscle, or adipose tissue insulin sensitivity, although there was modest increase in plasma high-molecular-weight plasma adiponectin, suggesting an improvement in adipose tissue biology and function. Still, adiponectin remained significantly low and less than ∼50% of normal. There were no significant placebo-corrected changes in several adipokines (interleukin [IL]-1Ra, tumor necrosis factor α, IL-6, and fibroblast growth factor 21) and in biomarkers of liver fibrosis (CK18-M30 and -M65), in contrast to an earlier study with dapagliflozin (49). The relative liver fat reduction was in the range observed in earlier RCTs with canagliflozin (50) but somewhat greater than with dapagliflozin (48,49,51) in patients with T2D and NAFLD. An improvement in hepatic insulin sensitivity and insulin secretion was observed in an earlier study with canagliflozin, along with a 38% reduction in liver fat compared with a 20% decrease with placebo (placebo-corrected difference of 18%; P = 0.09). Of note, in this RCT all patients received dietary advice, which likely accounted for the significant change in liver fat with placebo and highlighted the need for placebo-controlled studies in the field. Nevertheless, more patients on canagliflozin versus placebo lost ≥5% of body weight and had a ≥30% reduction in liver fat (38% vs. 7%, P = 0.009).
Effect of SGLT2i in NAFLD
| Author, year . | Agent . | n . | Duration (weeks) . | Comparator . | Main study results . | ||
|---|---|---|---|---|---|---|---|
| Body weight* . | ALT . | Liver fat* . | |||||
| Prospective open-label studies | |||||||
| Ito et al., 2017 (42) | Ipragliflozin | 66 | 24 | Pioglitazone | ↓ 3.7% | ↓ | ↓¶ |
| Ohta et al., 2017 (43) | Ipragliflozin | 20 | 24 | Standard care | ↓ 2.5% | ↓ | ↓ 39% |
| Shibuya et al., 2018 (44) | Luseogliflozin | 32 | 24 | Standard care | ↓ 3.2% | Unchanged | ↓¶ |
| Kuchay et al., 2018 (45) | Empagliflozin | 50 | 20 | Standard care | ↓ 1.1% | ↓ | ↓ 26% |
| Shimizu et al., 2019 (46) | Dapagliflozin | 57 | 24 | Standard care | ↓ 3.1% | ↓ | ↓† |
| Inoue et al., 2019 (47) | Canagliflozin | 20 | 52 | Standard care | ↓ 3.4% | ↓ | ↓ 31% |
| Randomized controlled trials | |||||||
| Bolinder et al., 2012 (48) | Dapagliflozin | 67 | 24 | Placebo | ↓ 2.2% | — | Unchanged |
| Eriksson et al., 2018 (49) | Dapagliflozin | 84 | 12 | Placebo | ↓ 2.2% | ↓ | ↓ 10%§ |
| Cusi et al., 2019 (50) | Canagliflozin | 56 | 24 | Placebo | ↓ 3.4% | Unchanged | ↓ 18%§ |
| Latva-Rasku et al., 2019 (51) | Dapagliflozin | 32 | 8 | Placebo | ↓ 2.1% | Unchanged | ↓ 13% |
| Kahl et al., 2019 (52) | Empagliflozin | 84 | 24 | Placebo | ↓ 2.4% | Unchanged | ↓ 22% |
| Author, year . | Agent . | n . | Duration (weeks) . | Comparator . | Main study results . | ||
|---|---|---|---|---|---|---|---|
| Body weight* . | ALT . | Liver fat* . | |||||
| Prospective open-label studies | |||||||
| Ito et al., 2017 (42) | Ipragliflozin | 66 | 24 | Pioglitazone | ↓ 3.7% | ↓ | ↓¶ |
| Ohta et al., 2017 (43) | Ipragliflozin | 20 | 24 | Standard care | ↓ 2.5% | ↓ | ↓ 39% |
| Shibuya et al., 2018 (44) | Luseogliflozin | 32 | 24 | Standard care | ↓ 3.2% | Unchanged | ↓¶ |
| Kuchay et al., 2018 (45) | Empagliflozin | 50 | 20 | Standard care | ↓ 1.1% | ↓ | ↓ 26% |
| Shimizu et al., 2019 (46) | Dapagliflozin | 57 | 24 | Standard care | ↓ 3.1% | ↓ | ↓† |
| Inoue et al., 2019 (47) | Canagliflozin | 20 | 52 | Standard care | ↓ 3.4% | ↓ | ↓ 31% |
| Randomized controlled trials | |||||||
| Bolinder et al., 2012 (48) | Dapagliflozin | 67 | 24 | Placebo | ↓ 2.2% | — | Unchanged |
| Eriksson et al., 2018 (49) | Dapagliflozin | 84 | 12 | Placebo | ↓ 2.2% | ↓ | ↓ 10%§ |
| Cusi et al., 2019 (50) | Canagliflozin | 56 | 24 | Placebo | ↓ 3.4% | Unchanged | ↓ 18%§ |
| Latva-Rasku et al., 2019 (51) | Dapagliflozin | 32 | 8 | Placebo | ↓ 2.1% | Unchanged | ↓ 13% |
| Kahl et al., 2019 (52) | Empagliflozin | 84 | 24 | Placebo | ↓ 2.4% | Unchanged | ↓ 22% |
Arrows indicate statistically significant changes vs. comparator.
Comparison-corrected (open-label) or placebo-corrected relative treatment difference in weight and liver fat measured with MRI-based imaging techniques.
Liver fat measured as liver-to-spleen attenuation ratio on computed tomography. In Ito et al. (42) the decrease in liver fat was similar to pioglitazone (comparator).
Significant improvement in liver fat by controlled attenuation parameter (CAP; Fibroscan).
Not significant compared with placebo.
Where do we go from here with SGLT2i in NASH? In the RCTs summarized in Table 1, patients’ HbA1c was rather well controlled (6.5–7.6%), so one may speculate a greater effect in NAFLD in a “real-world” cohort of patients with uncontrolled T2D. We also do not have information from controlled trials on how changes in ALT and steatosis with SGLT2i treatment translate to liver histology, but benefit has been reported in small uncontrolled clinical studies (53–55). A recent meta-analysis of 11 studies in 6,745 patients with T2D treated with canagliflozin reported a significant reduction in AST (56), ALT, and γ-glutamyl transferase, although results from smaller RCTs have been less consistent and depended on the baseline AST/ALT. Given the clinical cardiovascular and renal benefits of SGLT2i by mechanisms not initially anticipated, some on inflammation and profibrotic pathways (39–41), this class deserves further study in patients with steatohepatitis. In many studies (Table 1), SGLT2i reduced hepatic steatosis more than expected for the rather modest weight loss, suggesting additional weight-independent mechanisms. Furthermore, a reduction of liver fat may not necessarily be proportional to the improvement in necroinflammation or fibrosis, as recently suggested with pioglitazone (57).
In summary, it is time to include NASH in the management plan of patients with T2D in the same way as today it includes diabetic retinopathy or nephropathy. The American Diabetes Association in the 2019 Standards of Care guidelines recommends that “patients with type 2 diabetes and elevated liver enzymes (alanine aminotransferase) or fatty liver on ultrasound should be evaluated for the presence of nonalcoholic steatohepatitis and liver fibrosis” (recommendation 4.14) (58). Given the potential cardiometabolic and liver-specific complications associated with NASH, endocrinologists and the diabetes team must be at the forefront of disease prevention. Future studies should aim to better understand the natural history of liver disease in patients with diabetes, the biology of liver fibrosis to find novel plasma biomarkers that will identify “rapid disease progressors,” and the impact of NAFLD on micro- (11) and macrovascular (5,12) diabetes complications. This knowledge will be essential to develop cost-effective screening and long-term monitoring algorithms. We are also still at the dawn of treatment for NASH. While weight loss and exercise remain the cornerstone of NAFLD management, only a few short-term (≤12 months) controlled studies have been performed. Large, long-term multicenter lifestyle intervention studies are needed. Within this context, the role of SGLT2i will need to be better established, in particular, their efficacy as add-on therapy to new or already available FDA-approved medications for T2D with proven efficacy in NASH (i.e., add-on to pioglitazone, GLP-1RAs?). Finally, many novel pharmacological agents are being tested and will likely soon expand our treatment options. All health care providers taking care of patients with diabetes need to embrace today the evolving clinical challenge posed by NASH, educate their patients, and be proactive in the diagnosis and monitoring of patients with this “new complication” of T2D.
This article is part of a special article collection available at https://care.diabetesjournals.org/collection/nafld-in-diabetes.
Article Information
Duality of Interest. K.C. has received research support as principal investigator from the National Institutes of Health, Cirius, Echosens, Inventiva, Novartis, Novo Nordisk, Poxel, and Zydus. K.C. is a consultant for Allergan, AstraZeneca, Bristol-Myers Squibb, Boehringer Ingelheim, Coherus, Deuterex, Eli Lilly, Genentech, Gilead, Janssen, Madrigal, Pfizer, Poxel, Prosciento, Novo Nordisk, and Sanofi. No other potential conflicts of interest relevant to this article were reported.
