Causality Between Non-alcoholic Fatty Liver Disease and Risk of Cardiovascular Disease and Type 2 Diabetes

A Meta-analysis With Bias Analysis

Amy E. Morrison; Francesco Zaccardi; Kamlesh Khunti; Melanie J. Davies


Liver International. 2019;39(3):557-567. 

In This Article

Abstract and Introduction


Background & Aims: A causal association of non-alcoholic fatty liver disease (NAFLD) with cardiovascular disease (CVD) and type 2 diabetes (T2DM) remains unproved. We aimed to quantify the likelihood of causality examining the sensitivity of observational associations to possible confounding.

Methods: Studies investigating longitudinal associations of NAFLD with CVD or T2DM were searched on 5 June 2018. Study-specific relative risks (RRs) were combined in random-effects meta-analyses and pooled estimates used in bias analyses.

Results: Associations of NAFLD with CVD and T2DM were reported in 13 (258 743/18 383 participants/events) and 20 (240 251/12 891) studies respectively. Comparing patients with NAFLD to those without, the pooled RR was 1.48 (95% CI: 0.96, 2.29) for CVD and 2.17 (1.77, 2.65) for T2DM. In bias analyses, for an unmeasured confounder associated to both NAFLD and CVD with a RR of 1.25, the proportion of studies with a true (causal) effect of NAFLD on CVD surpassing a RR of 1.10 (ie, 10% increased risk of CVD in participants with NAFLD) was 0.67 (95% CI: 0.42, 0.92) while for 75% increase, it was 0.36 (0.11, 0.62). Corresponding figures for T2DM were 0.97 (0.91, 1.00) for a 10% increased risk of T2DM in participants with NAFLD to 0.70 (0.49, 0.92) for a 75% increase.

Conclusions: The results of this study are strongly suggestive for a causal relationship between NAFLD and T2DM, while the evidence for a causal link between NAFLD and CVD is less robust. Therapeutic strategies targeting NAFLD are likely to reduce the risk of developing T2DM.


Non-alcoholic fatty liver disease (NAFLD) is defined as hepatic fat accumulation (steatosis) in more than 5% of hepatocytes on histological examination, following exclusion of other causes of chronic liver disease including hepatitis B and C, autoimmune hepatitis, hepatotoxic drugs and excessive alcohol consumption.[1] The prevalence of NAFLD is rising worldwide, with it rapidly becoming the most common cause of chronic liver disease, and currently affecting 25% of population worldwide[2–4] Up to one fifth of patients affected with NAFLD will develop non-alcoholic steatohepatitis (NASH), with evidence of hepatocyte injury (ballooning degeneration); NASH could further progress to liver fibrosis, cirrhosis and its complications including hepatocellular carcinoma.[5]

In prospective longitudinal studies, the presence of NAFLD has been associated with an increased risk of cardiovascular disease (CVD) and type 2 diabetes (T2DM);[6,7] however, it remains unclear whether NAFLD is causally related to these conditions or, rather, it is just an epiphenomenon of other cardiometabolic factors causally related to CVD and T2DM.[8–10] In fact, the rising prevalence of NAFLD concurs with that of overweight and obesity (particularly visceral fat).[11,12] Moreover, as excess body weight is also a risk factor for CVD and T2DM,[13,14] it has been argued that the associations of NAFLD with CVD and T2DM could be spurious (ie, non-causal) and reflect the link between overweight/obesity with CVD and T2DM.[8,9] To account for such possible confounding, previous epidemiological studies have adjusted associations for several potential confounders, including body mass index;[6,7] however, residual confounding resulting from unmeasured factors is possible.

Recently, bias analysis (also known as sensitivity analysis) for single observational studies and meta-analysis has been proposed as a methodology to infer causal associations from non-randomised, observational data.[15–17] By quantifying the sensitivity of causal conclusions with respect to unmeasured confounding, this methodology allows assessing the strength of a causal evidence between the exposure and the outcome of interest using pooled estimates obtained from meta-analysis of observational studies.[17]

In this view, we conducted a systematic review, meta-analysis and bias analysis to help elucidate the nature of the association of NAFLD with CVD and T2DM by assessing the likelihood of a causal relationship.