Utilizing the most extensive publicly accessible GWAS datasets, we conducted a two-sample MR study to examine the correlation between sleep characteristics and the likelihood of developing DN. For MR research to ensure robust possible causal inference, three prerequisites must be fulfilled: 1) genetic instruments must have strong associations with the exposure, 2) genetic instruments must not be associated with potential confounding variables, and 3) genetic instruments should only influence the outcome through the exposure, excluding alternative mechanisms [13]. Figure 1 depicts a schematic for a bidirectional MR analysis.

This research conducted a two-sample MR analysis using summary data on the relationship between single nucleotide polymorphisms (SNPs) and phenotypes. The data were gathered from published GWAS conducted on the European population in the UK Biobank (https://www.ukbiobank.ac.uk/) and FinnGen (https://www.finngen.fi/en). The present study incorporated eleven sleep-related characteristics as factors of interest, including self-reported total sleep duration [14], short sleep (≤6 h) [14], long sleep (≥9 h) [14], insomnia [15], chronotype [16], daytime napping [17], daytime sleepiness [18], sleep duration measured by an accelerometer [19], timing of least activity during a 5-hour period [19], sleep efficiency [19], and the number of sleep episodes [19]. The research outcome for DN was derived from the ninth data release of the FinnGen collaboration [20]. The sample included 4,111 cases and 308,539 controls for DN. This study utilized data from FinnGen and UK Biobank. Ethical approval for the FinnGen study was obtained from the Coordinating Ethics Committee of the Hospital District of Helsinki and Uusimaa (HUS), with the ethical approval number HUS/990/2017. For the UK Biobank, ethical approval was granted by the North West Multi-centre Research Ethics Committee (MREC), with the approval reference number 11/NW/0382. All participants provided informed consent prior to participation in both FinnGen and UK Biobank studies. The GWAS datasets used in this study are publicly available and can be freely accessed by researchers. Therefore, no additional ethical approval was required for this secondary analysis.

IVs were selected based on a GWAS significance threshold of p<5×10^-8. SNPs that met this criterion were considered strong candidates for IVs. To ensure the independence of the selected SNPs and to avoid bias due to linkage disequilibrium (LD), we utilized PLINK software (https://www.cog-genomics.org/plink/) to prune the SNPs, excluding those with a pairwise LD r² <0.001 and ensuring a distance of more than 10,000 kb between SNPs. Additionally, SNPs were selected if they had an F-statistic>10 to ensure strong instrument strength. The F-statistic is calculated using the formula: F=R² (N-K-1)/(K[1-R²]), where K is the number of instrument, N is the sample size, and R² is the proportion of variance in the exposure explained by the SNP [21].

MR analyses were conducted using the TwoSampleMR package in R (https://github.com/MRCIEU/TwoSampleMR). We employed five different methods to ensure robust possible causal inference: inverse variance weighted (IVW), MR-Egger, simple mode, weighted mode, and weighted median [22-25]. The IVW method is our primary analysis tool as it combines the Wald ratios of individual SNPs into a single causal estimate, providing high statistical power when the instrumental variables are valid and there is no horizontal pleiotropy. We used MR-Egger regression to assess and correct for potential horizontal pleiotropy, which can bias the IVW estimates. MR-Egger provides a valid causal estimate even in the presence of pleiotropy, albeit with reduced statistical power. Simple mode and weighted mode methods offer alternative approaches to account for pleiotropy by identifying the most frequent causal estimate among the instruments. The weighted median method offers robustness to invalid instruments by providing a consistent causal estimate if at least 50% of the weight comes from valid instruments, thus offering protection against outliers and pleiotropic effects. To assess the presence of heterogeneity, which can indicate variability in the causal estimates derived from different SNPs, we used Cochran’s Q test. A significant Cochran’s Q value suggests heterogeneity among the SNPs, implying potential violations of MR assumptions. Conversely, a p-value >0.05 from Cochran’s Q test indicates no significant heterogeneity. Pleiotropy was evaluated using MR-Egger’s intercept; a non-significant intercept (p>0.05) suggests no evidence of directional pleiotropy. Sensitivity analyses were performed to examine the robustness of our findings. These included leave-one-out analyses, where each SNP is removed in turn to assess the influence of individual SNPs on the overall causal estimate. A consistent estimate across these analyses further supports the robustness of the causal inference.

Following the application of certain criteria, a total of 455 SNPs associated with eleven sleep-related characteristics were selected from the GWAS database. These SNPs were then used as IVs in the analysis (Supplementary Table 1). The details about the eleven sleep-related characteristics and DN are succinctly outlined in Supplementary Table 2. The SNPs used in the outcomes of both the forward and reverse MR analyses are shown in Supplementary Tables 3 and 4. All SNPs exhibited F statistics over 10, suggesting the absence of weak instrumental bias.

The results of the MR study are shown in Figures 2 and 3. The main IVW analysis revealed a possible causal link between genetically determined sleep efficiency and a reduced risk of DN (odds ratio [OR]: 0.384; 95% confidence interval [CI] 0.205 to 0.717; p=0.003) (Figures 3 and 4). Similarly, the weighted median approach yielded similar findings with an OR of 0.409, a 95% CI ranging from 0.190 to 0.881, and a p-value of 0.022. Heterogeneity was identified using the heterogeneity test. The test results for MR-Egger showed a Cochran’s Q value of 0.950 and a p-value of 0.813, while the results for IVW showed a Q value of 2.571 and a p-value of 0.631, as shown in Table 1. However, despite the use of the random-effects IVW technique as the main strategy, the findings produced from the method were not affected by the presence of heterogeneity. The SNPs for DN did not exhibit any signs of imbalanced horizontal pleiotropy. This is evident from the intercept of MR-Egger, which was -0.037 with a p-value of 0.292, as seen in Table 2. The funnel plots demonstrated the lack of directional pleiotropy for the outcomes, as seen in Figure 4D. The leave-one-out analysis conclusively showed that the MR estimates were not affected by any specific SNP (Figure 4A). Supplementary Table 5 presents the results of the IVW, weighted median, MR-Egger, simple mode, and weighted mode analyses for all eleven sleep-related characteristics on DN. A OR value greater than 1 suggests a higher odds of the outcome with the exposure, while a value less than 1 suggests lower odds. The CI provides a range of values within which the true OR is likely to fall, with a 95% CI indicating that we are 95% confident that the true OR lies within this range. If the CI does not include 1, the result is considered statistically significant.

The functional outcomes of DN are considered as exposures, whereas the eleven sleep-related characteristics are considered as outcomes. The results of the reverse MR analysis are shown in Figure 5. The findings from the IVW analysis suggest that worse functional outcomes after DN may be associated with a higher probability of experiencing a greater number of sleep episodes. The OR is 1.015, with a 95% CI ranging from 1.003 to 1.028, and a p-value of 0.016 (Figures 5 and 6). The findings obtained using the weighted median (OR: 1.015; 95% CI 1.000 to 1.031; p=0.047) were consistent. The heterogeneity test indicated no existence of heterogeneity (Cochran’s Q=8.064, p=0.427 for MR-Egger; Q=8.069, p=0.527 for IVW, as shown in Table 3). The SNPs for the number of sleep episodes did not exhibit any evidence of imbalanced horizontal pleiotropy, as shown by the MR-Egger intercept of 0.00025 (p= 0.944, as shown in Table 4). The funnel plots provided no indication of directional pleiotropy for the outcomes (Figure 6D). The leave-one-out analysis verified that the MR estimates were not affected by any single SNP (Figure 6A). The results for the five methodologies for all DN on eleven sleep-related characteristics are shown in Supplementary Table 6.

This research investigated the possible causal relationship between sleep characteristics and the risk of DN using a bidirectional MR approach. The results indicated that higher sleep efficiency was linked to a reduced chance of developing DN in the European population. Furthermore, we discovered a compelling cause-and-effect link suggesting that genetic predisposition to DN may elevate the likelihood of experiencing a greater number of sleep episodes, as shown by reverse MR. These findings might provide data for the prevention of DN and the screening of individuals with a heightened risk of DN.

Sleep has a vital role in preserving human health, as it profoundly affects several essential physiological processes [26,27]. Sleep regulation is an intricate process closely governed by physiological systems [28]. Our investigation revealed a possible causal relationship between sleep efficiency and a reduced risk of DN. The correlation discovered between sleep efficiency and DN risk is consistent with previous research indicating that individuals with diabetes and associated comorbidities, such as DN, often have poor sleep quality and disruptions. Sleep efficiency, a measure derived from actigraphy, combines estimates of awake and sleep [29,30]. The inclusion of wake time following sleep onset and sleep latency in the computation allows for precise assessment of sleep quality. Sleep efficiency is believed to be an indicator of the uninterrupted nature of sleep [31,32]. Research has shown a connection between sleep fragmentation and cognitive impairment, as well as decreased insulin sensitivity in healthy individuals [33,34]. Decreased sleep efficiency has been linked to hypertension, the noninvasive evaluation of brachial artery flow-mediated dilation in the general adult population, and the likelihood of developing cardiovascular disease [35,36]. Inadequate sleep quality has also been associated with other negative health consequences, such as metabolic dysfunction, inflammation, and altered glucose metabolism [37,38]. Poor sleep efficiency can lead to chronic inflammation, which is a known contributor to DN. Inflammatory cytokines such as interleukin 6 and tumor necrosis factor-α can promote renal inflammation and fibrosis, leading to impaired kidney function. Sleep efficiency is closely linked to glucose metabolism and insulin sensitivity. Poor sleep can result in insulin resistance and impaired glucose control, both of which are key factors in the development and progression of DN. Hormonal regulation also plays a critical role. Disrupted sleep patterns can affect the secretion of hormones such as cortisol and melatonin. Elevated cortisol levels can increase blood pressure and glucose levels, exacerbating the risk of DN. Additionally, melatonin has antioxidant properties and its dysregulation can lead to increased oxidative stress and kidney damage. These factors are all significant to the development of DN. This work expands upon these discoveries by presenting evidence of a possible cause-and-effect relationship, indicating that therapies targeting the enhancement of sleep efficiency may have positive effects on individuals at risk of DN. The OR of 0.384 (95% CI 0.205-0.717; p=0.003) suggests a significant decrease in the incidence of DN when sleep efficiency is improved. This highlights the crucial role of sleep health in managing diabetes.

On the other hand, the discovery that genetic predisposition to DN may raise the likelihood of experiencing a greater number of sleep episodes (OR: 1.015; 95% CI 1.003-1.028; p=0.016) complicates the connection between sleep and DN. This finding indicates a bidirectional relationship where DN may cause sleep disruption or more frequent sleep episodes. This might be due to symptoms like nocturia, pain, or metabolic imbalances associated with diabetes and DN. Sleep fragmentation, which refers to numerous interruptions and awakenings during sleep, may result in low sleep quality and has been linked to negative metabolic consequences such as insulin resistance and heightened inflammation. These results emphasize the need for a comprehensive strategy to address sleep difficulties in patients with DN in order to enhance overall health outcomes.

The strength and reliability of our results were reinforced by conducting sensitivity analyses, which verified that the observed connections were not influenced by horizontal pleiotropy or outlier SNPs. This enhances the soundness of our causal deductions and diminishes the probability of extraneous variables affecting the outcomes. Employing several sensitivity analyses, such as MR-Egger, weighted median, and leave-one-out analysis, instills confidence in the stability and dependability of our findings.

Although our work has notable features, such as the use of extensive GWAS data and rigorous MR techniques, it is important to highlight several limitations. Our results can only be applied to people of European descent, and more research is necessary to investigate similar connections in other ethnic groups. Genetic differences and environmental variables that affect sleep and DN may vary across different groups, and it is important to comprehend these dynamics in a wider perspective. Furthermore, while MR analyses are effective in reducing the impact of confounding and reverse causation, their reliability depends on the accuracy of instrumental variables, which may not include all genetic differences associated with sleep characteristics and DN. The use of SNPs as instrumental factors presupposes that these genetic variations are highly correlated with the exposure (sleep characteristics) and do not influence the outcome (DN) via other means. Sleep efficiency and other sleep characteristics were derived from self-reported data and actigraphy measures. While these are commonly used and validated methods, they are not as precise as polysomnography. Future studies could benefit from using more accurate and detailed measures of sleep. Deviation from these assumptions has the potential to introduce bias into the findings. Finally, the specific biological processes responsible for the observed connections are still not understood, which justifies the need for more experimental and clinical investigations to clarify the pathways involved.

Future studies should focus on investigating the underlying processes by which sleep efficiency affects the likelihood of developing DN. Examining the influence of inflammation, metabolic control, and hormonal pathways in this context might provide a deeper understanding of the fundamental mechanisms involved. Furthermore, it is necessary to conduct longitudinal research and clinical trials to evaluate the effects of sleep therapies on DN outcomes in order to confirm our results and apply them in clinical settings. It is vital to comprehend how including strategies to enhance sleep efficiency in diabetes care regimens might minimize the risk of DN.

To summarize, our research indicates that improving sleep efficiency may lower the likelihood of developing DN, underscoring the need to integrate sleep management into preventative measures for diabetic complications. Moreover, the possible causative impact of DN on the number of sleep episodes highlights the need for comprehensive treatment strategies that address both metabolic and sleep health. Subsequent studies should prioritize investigating the underlying mechanisms of these connections and assessing the efficacy of sleep therapies in decreasing the risk of DN. By simultaneously addressing sleep health and metabolic management, we can enhance the quality of life and clinical outcomes for those with diabetes who are at risk of developing DN.

This research, which examines the relationship between genetics and sleep efficiency, reveals that those with a greater genetic predisposition for excellent sleep efficiency have a lower chance of developing DN. This emphasizes the importance of good sleep efficiency in avoiding DN. Moreover, there is compelling evidence that genetic predisposition to DN may elevate the likelihood of experiencing a greater number of sleep episodes, suggesting a possible two-way interaction. These results emphasize the need to include sleep health in diabetes management strategies. Additional investigation is required to validate these connections and delve into the underlying biological processes.