Lean patients with fatty liver but no recorded cardiometabolic risk factors showed more evidence of liver injury and a higher adjusted risk of liver-related death than lean patients with neither finding, according to a multinational analysis published in Gut.
The study examined cryptogenic steatotic liver disease (SLD), which the investigators operationally defined as lean SLD in people with no recorded cardiometabolic risk factors and no other identifiable major cause of liver fat. Drawing on 693,936 individuals across five analytic cohorts, the analysis found that cryptogenic SLD accounted for 13.7% of lean SLD in the study's primary cohort and carried a 2.5-fold higher adjusted risk of liver-related death in a Korean national cohort (95% CI, 1.4-4.3). That association was not present before adjustment for sex and age.
The investigators built five analytic cohorts from four data sources: the UK Biobank (UKB), from which they derived separate cohorts based on magnetic resonance proton density fat fraction (MR-PDFF) and on the hepatic steatosis index (HSI); the National Health and Nutrition Examination Survey (NHANES) 2017-2018; the Korean National Health Insurance Service (KNHIS) health check-up cohort from 2011-2012; and a Korean health check-up cohort with magnetic resonance elastography (MRE). Depending on the cohort, steatosis was defined by MR-PDFF of at least 5%, HSI of at least 36, controlled attenuation parameter of at least 248 dB/m, or B-mode ultrasonography.
Across all cohorts, the investigators excluded people younger than 18, along with those who had baseline malignancy, hepatitis B or C, missing cardiometabolic data, or significant alcohol intake, defined as at least 420 g per week for men and 350 g per week for women. Lean was defined as a body mass index (BMI) below 25 kg/m² in non-Asian participants and below 23 kg/m² in Asian participants. The comparison group throughout was lean participants with neither steatosis nor any cardiometabolic risk factor.
Long-term outcomes were assessed in the UKB HSI and KNHIS cohorts only. The MR-PDFF cohort was not used for prognostic analysis because follow-up time in the analytic dataset was not indexed to the imaging visit.
The share of lean SLD classified as cryptogenic varied widely by cohort: 13.7% in the UKB MR-PDFF cohort, 9.8% in the UKB HSI cohort, 14.7% in NHANES, 27.2% in the MRE cohort, and 44.4% in KNHIS. The investigators attributed part of that spread to healthy participant bias, noting that people who volunteer for screening tend to be relatively well, which can lead to underestimation of cardiometabolic risk factors.
In the MR-PDFF cohort, used for phenotyping because liver fat was directly quantified, the cryptogenic group had a mean alanine aminotransferase (ALT) of 19.6 IU/L compared with 16.8 IU/L in the comparison group, along with higher gamma-glutamyl transferase and triglycerides and lower HDL cholesterol. Despite carrying no recorded cardiometabolic risk factors, this group also had a higher mean BMI, at 23.1 versus 22.2 kg/m², and reported higher weekly alcohol intake, at 115 g versus 88 g. About half were men, compared with roughly a quarter of the comparison group.
Mean iron-corrected T1 (cT1), an MRI marker of hepatic inflammation, was 711.3 ms in the cryptogenic group versus 674.0 ms in the comparison group.
That laboratory pattern did not hold in KNHIS, where the cryptogenic group had a lower mean ALT than comparators, at 11.1 versus 25.7 IU/L, along with lower gamma-glutamyl transferase, lower triglycerides, and higher HDL. The investigators linked the reversal to sex distribution: the KNHIS cryptogenic group was about 87% women, compared with roughly a third of the comparison group. They cautioned that cohort-specific hazard ratios should not be directly compared or over-interpreted.
Fibrosis findings were mixed and threshold-dependent. In the MRE cohort, the cryptogenic group had a higher fibrosis burden at cut-offs of 3.2 kPa and above — 3.7% versus 1.0% at that threshold — but not at 3.0 kPa. In NHANES, where fibrosis was assessed by transient elastography, the two groups were similar at every cut-off examined. The investigators noted that mean age in NHANES was in the mid-30s versus the mid-40s in the MRE cohort, and suggested fibrosis burden in cryptogenic SLD may differ by age or cohort setting.
In the MR-PDFF cohort, PNPLA3 risk variants were present in 57.3% of the cryptogenic group versus 37.7% of comparators, and TM6SF2 risk variants in 31.1% versus 13.1%. Neither enrichment was reproduced in the UKB HSI cohort, and HSD17B13 variants did not differ. PNPLA3 frequencies in the lean MASLD group were similar to those in cryptogenic SLD, at 54.0%, which the investigators read as shared genetic susceptibility across lean phenotypes rather than a feature distinguishing cryptogenic disease.
In KNHIS, cryptogenic SLD was not associated with liver-related death in unadjusted analysis, and Kaplan-Meier analysis showed no significant difference from the comparison group. The association emerged only with adjustment, reaching a hazard ratio of 2.2 after accounting for sex and age and 2.5 in the fully adjusted model, which added smoking, comorbidity, socioeconomic status, and frailty. Lean metabolic dysfunction-associated steatotic liver disease (MASLD) carried a hazard ratio of 7.8 in the same model.
In the UKB HSI cohort, liver-related death occurred in one of 96 participants with cryptogenic SLD, or 1.0%, versus 17 of 22,917 comparators, or 0.1%. The fully adjusted hazard ratio was 13.2, but its confidence interval ran from 1.9 to 92.4, and the investigators said the estimate should be interpreted cautiously given that it rested on a single event.
Cryptogenic SLD was not associated with all-cause mortality in either longitudinal cohort, nor with major adverse cardiac events after adjustment — unlike lean MASLD, which showed an adverse cardiovascular profile in both. The investigators said this pattern suggests the phenotype may carry liver-specific rather than broad cardiometabolic risk.
In KNHIS, cryptogenic SLD was associated with a modestly higher risk of extrahepatic cancer, with a hazard ratio of 1.2, a finding not reproduced in the UKB HSI cohort. Hepatocellular carcinoma risk was nominally lower, at a hazard ratio of 0.9, which the investigators cautioned should not be read as protective given the small number of events and differences in age, sex, baseline fibrosis, and follow-up duration.
Alcohol intake in the range used to define MASLD with increased alcohol intake was more common in the cryptogenic group than comparators in the MR-PDFF cohort, at 22.6% versus 16.1%, but not in either HSI-based cohort. After stricter alcohol exclusion, the KNHIS association with liver-related death attenuated to a hazard ratio of 2.1 but remained significant, while the UKB HSI estimate became inestimable because no liver-related deaths remained in the cryptogenic group. Elevated cT1 and the PNPLA3 and TM6SF2 enrichment persisted in the MR-PDFF cohort. The investigators concluded that subthreshold alcohol exposure may contribute to the phenotype in some people but is unlikely to be the sole explanation.
The investigators said the findings suggest lean SLD without apparent cardiometabolic risk factors may still warrant clinical attention, and called for further work to characterize the phenotype using more objective measures of steatosis for classification and risk stratification.
The investigators acknowledged that neither longitudinal cohort used a direct measure of liver fat. Because the HSI incorporates BMI and diabetes status, both of which overlap with the cardiometabolic criteria used to define MASLD, the index may systematically under-ascertain cryptogenic SLD. Within the UK Biobank, cryptogenic SLD accounted for 13.7% of lean SLD by MR-PDFF but only 9.8% by HSI. Exploratory analyses at lower HSI thresholds increased the cryptogenic group from 96 to 330 and then 1,326 participants but yielded only three and four liver-related deaths, respectively, with fully adjusted estimates attenuated and imprecise.
In the MR-PDFF cohort, imaging and laboratory measurements were not obtained at the same time, so the reported phenotype may reflect an earlier metabolic profile. Alcohol intake was self-reported and may have been under-reported. The investigators also lacked data on medication exposure, nutritional status, and rare liver diseases, and said subclinical cardiometabolic dysfunction may have gone undetected — meaning some cases may reflect unrecognized metabolic disease or other uncaptured causes.
The Korea National Institute of Health, the National Research Foundation of Korea, and a Regional Innovation System and Education grant through the Seoul RISE Centre, funded by the Ministry of Education and the Seoul Metropolitan Government, supported the study. The authors declared no competing interests.
Expert Insight
GI & Hepatology News invited Binu V. John, MD, chief of gastroenterology and hepatology at the Miami VA Health System, to comment on the study. His responses have been lightly edited for length and clarity.
How might the findings influence clinical practice?
Dr. John: Clinicians should consider patients with steatosis but without cardiometabolic risk factors or harmful alcohol use as a distinct entity that may still carry liver-related risk, potentially through genetic factors. These patients should undergo further risk assessment like we currently do with patients with other SLDs.
Is there anything else you'd like physicians to know about this study?
Dr. John: A few limitations to highlight.
The finding that cryptogenic SLD was associated with higher major adverse liver outcomes was driven by finding of a single outcome among 96 patients in the UK Biobank hepatic steatosis index (HSI) cohort. These findings were not seen across the other cohorts.
Two of the cohorts that had longitudinal data both use the HSI as a surrogate of hepatic steatosis. We and others have shown that the HSI is a poor surrogate for imaging confirmed steatosis. In fact, we observed that an elevated HSI was nearly as common among veterans with and without steatosis on cross-sectional imaging, indicating poor specificity and accuracy to estimate steatosis. While the cohorts that used MRI-PDFF and controlled attenuation parameter scores estimated steatosis reliably, the operational definition of steatosis in the other cohorts is suboptimal. This is a major limitation.
Additionally, it is possible that some patients who had cardiometabolic risk factors did not have them documented in the chart and therefore, labelled cryptogenic because of suboptimal coding. Undiagnosed prediabetes, elevated blood pressure below the threshold for a diagnosis of hypertension, or under-coded comorbidities could be misdiagnosed as cryptogenic SLD.
The sensitivity analyses showing that cryptogenic SLD retained higher alcohol intake in the range used to define MASLD with increased alcohol intake in the UK Biobank MR-PDFF cohort raises the possibility that some cryptogenic SLD may represent under-reported alcohol-related liver disease. The study, like many large database studies, lacked biomarker-based alcohol assessment like phosphatidylethanol, which is a limitation.
Overall, this is an interesting study that is well-executed and is hypothesis-generating. This calls for more rigorous, prospective studies examining this subgroup of patients, and long-term follow-up is critical to examine if this is truly associated with worse liver-related outcomes.
Dr. John received institutional research support from Exact Sciences, Genentech, and Takeda, unrelated to this topic, and has consulted for Madrigal, Gilead, and Ipsen.