This bidirectional Mendelian randomization study leveraged genetic data from a large number of published GWASs to explore a bidirectional hypothetical causal association between NAFLD and neuropsychiatric disorders. From a genetic perspective, our results revealed that MDD was the potential risk factor for NAFLD, and NAFLD also increased significantly the risk of anxiety disorder.
Anxiety disorders are common in Western populations [46]. A study comparing 19,871 NAFLD patients with 19,871 matched control individuals found that new-onset anxiety was linked to NAFLD even after multiple complications were controlled [46]. In a cross-sectional study, anxiety and depression scale scores were used to assess depression and anxiety symptoms in 567 patients with NAFLD confirmed by biopsy in the United States [47]; 45% and 25% of the patients had subclinical and clinical anxiety symptoms, respectively, indicating that anxiety is one of the common symptoms of NAFLD and that it is related to advanced NAFLD. On the other hand, to some extent, NAFLD may progress to nonalcoholic steatohepatitis (NASH) [48]. A smaller study involving 36 NASH patients and 36 control individuals without liver disease after cholecystectomy revealed that NASH increased the risk of anxiety disorders than the control group. It was also found that anxiety was related to more serious changes in liver histological characteristics [49]. This is consistent with the evidence that NAFLD increased significantly the risk of anxiety disorder observed in our MR study.
MDD is a common recurrent mental illness that leads to decreased quality of life and increased mortality in the relevant population [50]. Meantime, many studies have found a high prevalence of MDD in patients with NAFLD [51, 52]. A study found that NAFLD patients with MDD had more severe histological liver steatosis and higher NAFLD activity scores compared with NAFLD patients without MDD [53]. Pathologically, MDD is associated with a severe hepatocellular balloon [54], which is one of the characteristics of NAFLD [55]. In another aspect, the rising levels of inflammatory markers in depressed patients, such as intercytotin-6, C-reactive protein and tumor necrosis factor α, provide another reasonable explanation for the association with NAFLD [56].
Our MR results found no causal relationship between NAFLD and AD, which goes against the previous studies. However, these studies were basically cross-sectional studies, and the longitudinal associations between them should still be supported by further research data. At the same time, there are also many studies showing that there is no significant direct correlation between NAFLD and AD [57, 58]. As mentioned above, differences in the mean ages of the study population and the cognitive level tests used are also responsible. And the severity of liver fibrosis with disease progression (rather than NAFLD itself) may be associated with cognitive dysfunction [59, 60]. There are also many literature highlighting the role of the degree of fibrosis in the development of diseases (such as cardiovascular function [61] and stroke [62]).
A cross-sectional study suggested that there was no significant relationship between NAFLD and migraine [63], which is consistent with our results. Although a study has found that there may be a causal relationship [64], we believe that the sample size used is too small and the appropriate control cases are not used, so the causal relationship between them still needs further studies to verify.
The results of this study contradict previous observational studies that reported a link between schizophrenia and NAFLD. Similarly, it has been reported that there may be a causal link between bipolar disorder and NAFLD [65], but this is not consistent with our results. The results may be affected by the confusion or selection bias inherent in traditional observational studies, such as differences in diet structure, lifestyle and genetic susceptibility of different populations. The connection between NAFLD and these factors may be caused by lots of factors rather than simply the disease itself. For example, a research explained that the use of atypical antipsychotic drugs for the treatment of schizophrenia and other diseases will produce many metabolic and endocrine-related side effects, like insulin resistance, dyslipidemia and systemic inflammation, which will increased the risk of NAFLD [66]. In young male patients, different drug doses and combinations are also risk factors for NAFLD [67]. Although glucose homeostasis is altered before the onset of schizophrenia [68], insulin sensitivity in migraine patients is reduced [69], and abnormal glucose metabolism is a risk factor for NAFLD [70], these only suggest that both have a similar pathogenesis. The possible potential causal relationships between them still need to be further explored.
The main strength of our study was that the MR design avoids the confusion or selection bias common in traditional observational studies and supports a robust causal relationship between exposure and results. Besides, our study used aggregate data from large genetic databases, which made our research assumptions more accurate. The use of multiple sensitivity analyses combined with "leave one out" analysis enabled the testing of the stability of key results and provided additional reliable evidence. Third, this was a bidirectional causal-association study that expands the evidence on the relationship between neuropsychiatric disorders and NAFLD. Finally, this analysis had guiding recommendations for future studies on MASLD.
However, our research also had some limitations. First, the population studied was exclusively of European descent; therefore, further studies should be encouraged to verify these results in other populations. Second, although the sensitivity analysis did not reveal any evidence of multiple effects that may affect the results, the potential for horizontal pleiotropic effects couldn’t be completely excluded. Third, we chose some disease-related instruments with a relatively loose threshold of 5 × 10− 6, which was considered to be a reasonable threshold [71, 72], but it will affect the statistical efficacy of causal estimation to some extent. Finally, in studies, exposure and outcome were not only likely to have potentially duplicate participants, but it was also difficult to predict the extent to which the sample overlap. However, in this study, the use of sufficiently powerful tool variables (F statistics > 10) could minimize the potential deviation caused by sample overlap [32].
In summary, our study implied causal relationships of NAFLD with MDD and anxiety disorders. For other related diseases, more genetic instruments and larger GWAS data are needed to verify the results. On the other hand, our study further verified the bidirectional causality between NAFLD and neuropsychiatric disorders.