In this study, we found that there is a relationship between RNFL thickness and schizophrenia and bipolar disorder, which has also been observed in previous observational studies [29, 30], and this study further indicates that this relationship is a positive causal relationship, whereas RNFL thickness is not related to depressive disorder. There is no reverse causal effects of psychiatric disorders on RNFL in this study. Accordingly, we can regard the RNFL as a promising biomarker to aid in the diagnosis of schizophrenia and bipolar disorder disorders, and although we cannot discriminate between the two disorders by relying on the RNFL alone, a combination of imaging and assessment of retinal function (e.g., ERG measurements) reveals that reduced b-wave amplitude can be observed in schizophrenia only. This difference may be important in distinguishing between the two diseases [8, 9].

Although the MR analysis led us to the above results, it is important to recognize that the mechanisms linking the state of RNFL to the onset of psychiatric disorders are not yet clear. Therefore, caution should be exercised in drawing conclusions from our findings. Previous studies have pointed out that schizophrenia is a refractory severe mental disorder, and its clinical symptoms such as visual hallucinations is closely related to RNFL [6, 7, 31], and this study also supports this relationship, that is, there is a positive causal relationship between retinal fiber layer thickness and schizophrenia. However, previous studies have shown that there is no significant difference in retinal nerve fiber layer thickness between schizophrenia and healthy controls [32]. The two different results in observational studies on the same issue. We analyzed that this may have potential influence factors, for example, the different duration of disease in the recruited subjects of the disease group when they were enrolled for observation, whether they had taken antipsychotic drugs, and the type of drugs they had taken may be potential factors affecting the final results. This question has been considered and studied in observational studies, but the number of studies is limited. Bipolar disorder is a chronic disabling mental disorder. In this study, a positive causal relationship between RNFL and bipolar disorder was found. Similar to previous observational studies on schizophrenia and RNFL, previous observational studies showed two different results on the relationship between RNFL and bipolar disorder [33]. We analyzed that this may be due to the inconsistent and small number of subjects in the clinical observational studies, and the inconsistent duration of onset may be the potential factors causing the completely opposite results of the clinical observational studies. In addition, the GWAS data on bipolar disorder included in this paper do not clearly reflect the clinical subtypes of bipolar disorder, so we were unable to clarify the correlation of retinal changes with bipolar disorder I and bipolar disorder II, respectively, as well as the differentiation of these two subtypes by retinal changes. The absence of a causal relationship between depression and RNFL in the present study, but the association between depression and RNFL in previous studies, led us to speculate that other factors may be responsible for the association between depression and RNFL in the clinical observational studies.

Our study also has strengths, firstly, reviewing the previous literature, we can find that most of the studies on the relationship between retinal nerve fiber layer thickness and mental disorders are observational studies, and exploring the causal relationship between the two is novel in this study. Second, our research methodology also offers advantages. Although the development of psychiatric disorder series is the result of multifactorial interactions, and these disorders are not only inherited but also heavily influenced by early life adversities such as perinatal infections, maltreatment or bullying, and substance-related factors such as drugs and alcohol [34,35,36], one of the major strengths of the Mendelian randomization approach, as opposed to observational studies where it is difficult to accurately control for all confounders, lies in the fact that it utilizes genetic variations as an instrumental variable that can be used to study potential health risks reflecting susceptibility by avoiding confounding factors [11, 12]. Therefore, we can explain that patients with lesions in the RNFL have a higher genetic predisposition to develop schizophrenia and bipolar disorder compared to normal. Indicance threshold of each instrumental variable to P < 5 × 10− 8 according to the principle of Bonferroni Correction, which minimizes the possibility of Type I error due to multiple test comparisons, and used various methods to reduce the bias of instrumental variables, such as MR Presso, to further confirm the robustness of our results.

Finally, this study has several limitations. First, the source of exposure and outcome data in our study was selected from a single database, which may have caused some selection bias. Second, although we tested and adjusted for sensitivity in the data we analyzed, we could not fully assess horizontal pleiotropy. Third, limited by the lack of sample introduction in the database, we could not accurately grasp the information of the samples in the data, and could not exclude the confounding factors affecting the analysis results. Fourth, the population samples in our database were mostly from Europe, so the results of this study cannot be replicated among Asians.