Coffee, caffeine and non-alcoholic fatty liver disease?

We read with interest the meta-analysis [Shen et al. 2016] in Therapeutic Advances in Gastroenterology regarding the association between caffeine and non-alcoholic fatty liver disease (NAFLD). The authors analysed separately the associations of (1) caffeine from ‘regular coffee’, which has been described elsewhere as ‘filtrated coffee’ [Anty et al. 2012], and (2) ‘total caffeine’, which included all coffee types, including espresso, and non-coffee caffeine (e.g. tea and soda). The authors found that ‘regular coffee caffeine’ but not ‘total caffeine’ was inversely associated with NAFLD fibrosis severity, and that neither was associated with NAFLD prevalence.

Coffee is a complex mixture of biologically active compounds, including caffeine, chlorogenic acid, cafestol and kahweol. These compounds possess anti-oxidative and anti-fibrotic properties and may protect against a range of liver diseases, including fibrosis, cirrhosis and hepatocellular carcinoma [Saab et al. 2014]. Caffeine is thought to prevent fibrosis by antagonism of adenosine receptor A2a [Feld et al. 2015], which inhibits hepatic stellate cells: the primary mediators of hepatic fibrosis [Wang et al. 2014]. Therefore, it is surprising that the meta-analysis found an inverse association between NAFLD hepatic fibrosis severity and ‘regular coffee caffeine’ but not ‘total caffeine’. One possible explanation for this is that a compound in regular coffee other than caffeine is responsible. An equally plausible explanation is that no association between ‘total caffeine’ and NAFLD fibrosis severity was found because the data in the meta-analysis was unadjusted for confounders, such as type 2 diabetes mellitus (T2DM) and body mass index (BMI). This was apparent in an additional sensitivity analysis we performed on these data, which involved rerunning the meta-analysis while excluding the individual studies one at a time to examine their influence on the pooled association. When we excluded the study by Bambha and colleagues [Bambha et al. 2014], we found a statistically significant association between ‘total caffeine’ and NAFLD hepatic fibrosis severity. However, when we included the study by Bambha and colleagues, the association was null, as reported in the meta-analysis. This is paradoxical since Bambha and colleagues reported that the sole measured source of caffeine (i.e. ‘coffee’ described as ‘coffee {unspecified}’ in the meta-analysis) was actually inversely associated with fibrosis, but only after adjustment for confounders. Unfortunately, the adjustment was not incorporated into the meta-analysis because ‘total caffeine’ was calculated from raw unadjusted coffee consumption.

Another notable finding from the meta-analysis is the lack of an association between either ‘regular coffee caffeine’ or ‘total caffeine’ and NAFLD prevalence. The risk of NAFLD is increased in T2DM. Observational studies report inverse associations between coffee and T2DM [Ding et al. 2014]. The mechanism of action is thought to be largely mediated by chlorogenic acid, which improves glucose metabolism by inhibiting gut absorption and gluconeogenesis [Ding et al. 2014]. Chlorogenic acid is present in regular coffee, so a protective effect against diabetes and, thus, reduced NAFLD prevalence might be expected. The lack of an association may reflect the study designs (e.g. cross-sectional and case control) which provide low levels of evidence (e.g. due to selection/information bias and lack of temporality). However, above all it underscores the need for robust randomized trials to investigate the effects of coffee on liver disease and other health outcomes.