Type 2 diabetes mellitus (T2DM) is an increasingly prevalent metabolic disorder, characterised by rising incidence and prevalence(Reference Zheng, Ley and Hu1). Its progression can lead to serious health complications, including retinopathy, nephropathy, neuropathy and ketoacidosis(Reference Galiero, Caturano and Vetrano2,Reference Anders, Huber and Isermann3) . Thus, identifying novel protective factors to prevent the onset and progression of T2DM is crucial(Reference Magkos, Hjorth and Astrup4).
Dried fruits, being concentrated forms of fresh fruits with reduced moisture content, are easier to store. Primarily composed of fructose, they are also abundant in trace elements, vitamins and bioactive compounds(Reference Alasalvar, Salvadó and Ros5–Reference Bennett, Singh and Clingeleffer8). The influence of dried fruit intake on T2DM has been examined in several studies. For instance, a meta-analysis revealed that the high dietary fibre content in dried fruits can slow carbohydrate absorption, thereby reducing postprandial blood glucose levels. The fructose in dried fruits can moderate the increase in blood sugar, and their low glycaemic index aids in controlling blood sugar by reducing insulin secretion(Reference Qi, Chiavaroli and Lee9–Reference Nyambe-Silavwe and Williamson11). Observational studies suggest that the rich nutrient and bioactive substance content in dried fruits can offer antioxidant and anti-inflammatory benefits, enhance insulin sensitivity and assist in preventing and managing T2DM(Reference van der Merwe, Moore and Hill12,Reference Rajkowska, Otlewska and Broncel13) . Furthermore, research by Effie Viguiliouk indicates that dried fruits, due to their low glycaemic index, can lower glycaemic responses by replacing half of the available carbohydrates in white bread(Reference Viguiliouk, Jenkins and Blanco Mejia14). However, most of these studies are observational and, despite addressing various confounding factors, cannot entirely negate residual confounding. Thus, there is an evident gap in genetic evidence from large population-based studies regarding the potential role of dried fruit intake in mitigating the risk of T2DM, both with and without complications, and its effect on glycaemic traits.
Mendelian randomisation (MR) leverages genetic study data to assess the unconfounded relationship between exposure factors and outcomes. This method, by randomly allocating genetic variants during meiosis, effectively minimises confounding factors(Reference Pierce, Ahsan and Vanderweele15). As far as we are aware, no research to date has applied the MR method to investigate the causal link between dried fruit intake and T2DM, covering cases both with and without various complications, and glycaemic traits. In this study, we utilise a two-sample MR approach, drawing on comprehensive genome-wide association study (GWAS) summary data, to explore the potential protective impact of dried fruit intake on the risk of T2DM. This investigation offers new insights and strategies for T2DM prevention and management, potentially enhancing public health and life quality.
Materials and methods
Research design and data sources
The methodology of our study is depicted in Fig. 1. We utilised a two-sample MR approach to investigate the causal relationship between dried fruit intake and T2DM, as well as associated glycaemic traits. This two-sample MR methodology, which employs distinct and independent GWAS datasets, offers enhanced effectiveness and accuracy compared with single-sample MR. Instrumental variables (IV) for this analysis were selected based on single nucleotide polymorphisms (SNP) that align with the three essential principles of a two-sample MR design: (1) a strong association with the exposure (P < 5 × 10–8, F-statistic >10); (2) independence from confounding factors between the exposure and the outcome; and (3) influence on the outcome solely via the exposure(Reference Minelli, Del Greco and van der Plaat16). In our study, dried fruit intake constituted the exposure factor, and the various T2DM and glycaemic traits represented the outcomes.
Three key assumptions of the Mendelian randomisation study. (1) SNP are strongly associated with dried fruit intake; (2) SNP are independent of confounders; (3) SNP must only affect T2DM and glycaemic traits via dried fruit intake. SNP, single-nucleotide polymorphism; T2DM, type 2 diabetes mellitus.
The GWAS summary data regarding dried fruit intake were derived from the UK Biobank, encompassing 500 000 samples. Dried fruit intake was assessed via a questionnaire that inquired about the frequency of consumption, leading to data on dried fruit intake as an exposure factor for 421 764 European-descent participants. The GWAS summary data for diverse clinical phenotypes of T2DM originated from the FinnGen database, with glycaemic traits data (fasting glucose, 2-h glucose, fasting insulin and HbA1c) sourced from Meta-Analyses of Glucose and Insulin-related traits Consortium consortium. Given the public availability of these summary data, ethical review approval was not required for this study.
Selection of instrumental variables
For the MR analysis, adherence to the criteria of relevance, independence and exclusion restriction was crucial. SNP significantly associated with dried fruit intake (P < 5 × 10–8) were initially selected as IV. To minimise high linkage disequilibrium among the chosen SNP, clumping was performed using stringent parameters (r 2 < 0·001 and physical window = 10 000 kb), ensuring the independence of the IV. The r² threshold of 0·001 ensures very low correlation between SNP, while the 10 000 kb window captures potential long-range LD, both based on established genetic epidemiology practices. The F-statistic was calculated to gauge the strength of the IV, with an F-statistic greater than ten suggesting the absence of significant weak IV bias. The F-value was determined using the equation F = R 2 × (N-2)/(1-R 2), where R 2 represents the variance in exposure explained by each IV, calculated as 2 × EAF × (1-EAF) × Beta2, with beta being the allele effect size and EAF the effect allele frequency. Additionally, the IV identified in this study were scrutinised using the PhenoScanner database to detect any pleiotropic effects (P < 1E-5, r² < 0·8). SNP associated with outcomes related to our exposure of interest were subjected to further scrutiny. For instance, if during our analysis, we identified some IV that were associated with potential confounding factors, these confounders were included as covariates in the multivariable MR model.
Statistical analysis
In the univariate two-sample MR analysis, three methodologies were deployed to investigate the genetic link between dried fruit intake and T2DM, inclusive of complications: MR-Egger regression, the weighted median method and the inverse variance weighted (IVW) approach. The IVW method served as the principal analytical tool in this study, complemented by MR-Egger regression and the weighted median method. The robustness of the findings was ensured through the application of Cochrane’s Q test for assessing IV heterogeneity and the use of funnel plots to exhibit this heterogeneity symmetrically. To detect multicollinearity, both the MR-Egger intercept test and the MR-PRESSO global test were utilised. Furthermore, the leave-one-out test was conducted to ascertain the stability of the results. Statistical analyses were executed using R software, with a P value threshold of less than 0·05 denoting statistical significance.
Multivariable MR Analysis: The analysis utilised data from Chen Jin’s study, which included a table ‘Details of phenotypes related to 43 IV found on the Phenoscanner website’. This table demonstrated a pronounced correlation between the IV and BMI. Therefore, these IV were integrated into the multivariable MR analysis to mitigate their impact on the causal relationship between exposure and outcomes. The comprehensive results are delineated in online Supplementary Table S1 (Reference Jin, Li and Deng17). The multivariable MR analysis, inclusive of BMI, was executed to refine the assessment of causal effects between exposure and outcomes.
Results
Selection of instrumental
Variables following thorough screening, forty-three SNP were chosen as IV for our analysis. The cumulative F-statistic for these IV stood at 42·3651, with each individual IV’s F-statistic surpassing the threshold of ten, and the lowest observed F value exceeding 17·4989. Further specifics regarding these IV are detailed in online Supplementary Table S2.
Mendelian randomisation
In this study, three methods – IVW, weighted median and MR-Egger – were utilised to explore the causal relationships between dried fruit intake and T2DM, covering both uncomplicated cases and with five types of complications, as well as four types of glycaemic traits. The results consistently demonstrated a causal link between higher dried fruit intake and a lower risk of T2DM, including cases with neurological, ophthalmic and renal complications. Importantly, the IVW method indicated a causal relationship between dried fruit intake and reduced fasting insulin levels, a result not supported by the weighted median and MR-Egger methods. However, no significant associations were found for T2DM involving ketoacidosis, peripheral circulatory complications, HbA1c, fasting glucose or 2-hour glucose using any of the methods. Detailed findings are presented in Fig. 2 and Table 1. The scatter plots reveal the estimated impact of IV on both exposure and outcomes, with an upward trend indicating a negative correlation between dried fruit intake and the risk of T2DM in various categories, including fasting insulin (online Supplementary Figs. S1–10). Additionally, the diversity in the number of IV used for analysing the causal relationship between dried fruit intake and outcome is due to variations in the outcome datasets and the exclusion of palindromic SNP with moderate allele frequencies.
Associations of genetically predicted dried fruit intake T2DM and glycaemic traits. No. of SNP, number of single nucleotide polymorphisms. SNP, single-nucleotide polymorphism; T2DM, type 2 diabetes mellitus.
Two-sample Mendelian randomisation (MR) analyses of the association between dried fruit intake and T2DM and glycaemic traits
IVW, inverse-variance weighted; MR, Mendelian randomisation; T2DM, type 2 diabetes mellitus; WM, weighted median.
Genetic pleiotropy was assessed using MR-Egger regression analysis. In cases of T2DM without complications and fasting insulin, all intercepts were near zero (P > 0·05), suggesting minimal bias influence on the results. Similarly, the MR-PRESSO method corroborated the MR-Egger regression findings, indicating an absence of genetic pleiotropy. For T2DM without complications, the MR-PRESSO analysis yielded a P value of less than 0·05, but the MR-PRESSO outlier test results supported no horizontal pleiotropy (P > 0·05). In the case of fasting insulin, the MR-PRESSO analysis showed a P value below 0·05, and most SNP had P values near 1 in multiple outlier detection tests, with only a few exceptions and no significant outliers detected. These findings further reinforce the robust causal relationship between dried fruit intake and fasting insulin levels. Comprehensive results are presented in Table 2. The funnel plot (online Supplementary Figs. S11–15) illustrates that causal effects are symmetrically distributed when individual SNP serve as IV, implying reduced potential bias and enhanced stability and reliability of the results. Nonetheless, both the MR-Egger and MR-PRESSO methods indicate the existence of horizontal pleiotropy in T2DM with neuropathy, diabetic retinopathy and nephropathy, as detailed in Table 2.
Horizontal pleiotropic test between dried fruit intake and T2DM and glycaemic traits
DT, distortion test; T2DM, type 2 diabetes mellitus.
The leave-one-out analysis was employed to scrutinise the results obtained via the IVW method. Following the individual exclusion of each SNP, the outcomes consistently aligned with those of the IVW method in the causal effect analysis. This consistency implies that no single SNP exerted a significant impact on the causal estimations. These findings are illustrated in online Supplementary Figs. S16–17. Notwithstanding the observed heterogeneity in cases of T2DM without complications and fasting insulin (Q value < 0·05), the results from the multiple random effects model concurred with the MR estimates. This concurrence points to a causal link between dried fruit intake and both T2DM without complications and fasting insulin levels (P < 0·05), as detailed in Table 3.
Heterogeneity test between dried fruit intake and T2DM and glycaemic traits
df, degrees of freedom; IVW, inverse-variance weighted; MR, Mendelian randomisation; Q, Cochran’s Q statistic; T2DM, type 2 diabetes mellitus.
In the context of multivariable MR analysis, the analysis addressing T2DM without complications and fasting insulin yielded the following insights: For T2DM without complications, post-BMI adjustment (OR = 0·6056, 95 % CI: 0·3472, 1·0562, P = 7·7146e-02), the causal link between dried fruit intake and T2DM without complications was no longer significant. Regarding fasting insulin, after BMI adjustment (OR = 0·8065, 95 % CI: 0·7438, 0·8744, P = 1·8549e-07), the relationship between dried fruit intake and fasting insulin continued to exhibit a causal connection, with a marginal increase in the causal effect in the multivariable MR relative to the univariate MR, as detailed in Table 4.
Causal relationships of dried fruit intake and BMI on T2DM and glycaemic traits estimated by multivariable MR
IVW, inverse variance weighted; MR, Mendelian randomisation; SNP, single nucleotide polymorphism; T2DM, type 2 diabetes mellitus.
Discussion
This research employed a two-sample MR method to investigate the causal link between dried fruit intake and T2DM, including both the presence and absence of various complications and glycaemic traits. The findings reveal a causal relationship between dried fruit intake and a lower incidence of T2DM without complications, as well as a decrease in fasting insulin levels. However, such a causal relationship was not established for T2DM with complications. Based on the search results from the PhenoScanner website, we used multivariable MR analysis to adjust for the causal effect between exposure and outcome. We found that after adjusting for BMI, the causal effect between dried fruit intake and type 2 diabetes without complications was no longer significant. Additionally, we observed that the relationship between dried fruit intake and fasting insulin levels became more significant after adjusting for BMI. These results imply that the intake of dried fruits might contribute to a reduced occurrence of T2DM in its uncomplicated form and lower fasting insulin, and BMI may have some influence on these relationships, which should be carefully considered when developing prevention strategies for T2DM and its associated clinical manifestations.
As mentioned, causal estimation is effective when the three assumptions in the MR model are satisfied. First, forty three significantly correlated and independent SNP were selected that were closely related to dried fruit intake. The lowest observed F-statistic is 17·4989, indicating that the selected SNP were robust IV. Second, to evaluate the bias caused by pleiotropy in MR, we used the MR-Egger regression and MR-PRESSO methods, no pleiotropy was observed in the results. Third, the heterogeneity analysis indicates potential heterogeneity in certain instrumental variables, but the validity tests of these IV show that their strength is sufficient (all F-statistics are ≥ 15; online Supplementary Table 2). Overall, the sensitivity analysis confirms the reliability of these results. Moreover, when heterogeneity was present in the results, we re-estimated the causal relationship using a multiplicative random effects model. Heterogeneity was observed in cases of type 2 diabetes without complications and in fasting insulin (Q value < 0·05), but the results of the multiplicative random effects model were consistent with the MR estimates.
Although only a limited number of studies have explored the relationship between dried fruit intake and T2DM, some research has identified a close connection between dried fruit intake and key T2DM indicators, such as postprandial glucose levels, insulin secretion and insulin sensitivity(Reference Hernández-Alonso, Camacho-Barcia and Bulló18). Dried fruits, known for their high dietary fibre content, can effectively slow carbohydrate absorption, thereby reducing postprandial blood glucose levels(Reference Nyambe-Silavwe and Williamson11,Reference Alasalvar, Chang and Kris-Etherton19) . Additionally, the fructose content in dried fruits may attenuate the rate and magnitude of blood sugar spikes, and their low glycaemic index contributes to decreased insulin secretion, thus facilitating blood sugar regulation(Reference Sullivan, Petersen and Kris-Etherton20,Reference Zhu, Fan and Dong21) . Observational studies have highlighted that dried fruits, being nutrient-rich and containing various bioactive compounds, offer antioxidant and anti-inflammatory benefits, enhancing insulin sensitivity and aiding in the prevention and management of T2DM on multiple fronts(Reference Hamauzu and Suwannachot7,Reference Rajkowska, Otlewska and Broncel13,Reference Magiera, Czerwińska and Owczarek22) . Furthermore, the fatty acids present in certain dried fruits have been suggested to improve lipid metabolism, diminish lipid accumulation and alleviate insulin resistance(Reference Jardim, Domingues and Alves23). To our knowledge, this study represents the inaugural use of the MR method to evaluate the effect of dried fruit intake on T2DM.
The relationship between dried fruit intake and the manifestation of T2DM without complications, as well as its impact on fasting insulin levels, involves a variety of factors. First, inflammation, a critical factor in the development of T2DM(Reference Hotamisligil24–Reference Del Bo’, Martini and Porrini26), may be alleviated by the antioxidants and anti-inflammatory agents present in numerous dried fruits, potentially reducing T2DM risk(Reference van der Merwe, Moore and Hill12). Moreover, the onset of T2DM is closely connected to both insulin secretion and sensitivity(Reference Galicia-Garcia, Benito-Vicente and Jebari27). The antioxidants and micronutrients found in dried fruits could offer protection to pancreatic cells, thus aiding in the preservation of efficient insulin secretion and function(Reference Viguiliouk, Jenkins and Blanco Mejia14,Reference Hernández-Alonso, Camacho-Barcia and Bulló18,Reference Alasalvar, Chang and Kris-Etherton19) . Additionally, dried fruits are rich in unsaturated fatty acids, particularly monounsaturated and polyunsaturated fats, believed to enhance insulin sensitivity and decrease insulin resistance(Reference Matemu, Adeyemi and Nyoni28). This improvement in insulin sensitivity can lead to more effective cellular responses to insulin, thereby contributing to better blood sugar management(Reference Martins and Conde29). Finally, the high dietary fibre, healthy fats and plant proteins in dried fruits slow the digestive process, reducing the glycaemic response. This decelerated digestion helps stabilise blood sugar levels, minimising abrupt spikes in blood sugar and insulin fluctuations and consequently promotes improved glycaemic control(Reference Noronha, Braunstein and Glenn30–Reference Jia, Xuan and Dai33).
Our study’s findings reveal that the dried fruit intake can notably decrease the risk of developing T2DM without complications, in contrast to its forms with complications. Furthermore, our analysis showed that dried fruit intake is associated with lower fasting insulin levels, although it does not significantly impact fasting plasma glucose, 2-hour glucose or glycated Hb levels. Notably, the antioxidants and anti-inflammatory compounds present in dried fruits are likely to enhance insulin sensitivity, contributing to the long-term regulation of blood sugar metabolism. This suggests that dried fruits could be more beneficial in preventing milder forms of T2DM. Nevertheless, additional research is required to fully understand the underlying mechanisms through which dried fruits exert their protective effects on various clinical manifestations of T2DM.
This study boasts several notable strengths. Foremost, to the best of our knowledge, it is the inaugural study to leverage large-scale GWAS for exploring the causal link between dried fruit intake and T2DM. Our application of the two-sample MR approach effectively addresses certain limitations inherent in observational studies, such as the issue of reverse causation. Furthermore, our research adhered to a stringent P value criterion of <5 × 10–8 for selecting SNP associated with dried fruit intake as IV. We also successfully eliminated linkage disequilibrium among these IV. Significantly, the F-statistics for all our IV exceeded the threshold of 10. Lastly, we employed a variety of methods to conduct sensitivity analyses, assess horizontal pleiotropy and evaluate heterogeneity. These comprehensive tests affirm that the observed associations between dried fruit intake and both T2DM without complications and fasting insulin levels are stable and robust.
Limitation
Our study, while comprehensive, is not without its limitations: (i) The participant pool in the GWAS was exclusively of European descent. This uniformity in genetic background may have contributed to the heterogeneity observed in our results, thus casting uncertainty on the generalisability of our findings to other ethnicities and regions. (ii) Despite conducting extensive sensitivity analyses to validate the stability of our results, conclusively proving their independence from horizontal pleiotropy remains a challenge. (iii) The data on dried fruit intake were sourced from questionnaire surveys conducted by the UKB, which could potentially introduce classification bias. However, the considerable size of the sample population likely reduces the impact of such bias. (iv) The precise mechanisms underlying the impact of dried fruits on health outcomes, particularly in relation to T2DM, are not entirely elucidated. In summary, our research indicates that dried fruit intake may lower the risk of developing T2DM without complications.
Conclusion
Our research, employing MR analysis, has determined that the dried fruit intake may lower the risk of T2DM that manifests without complications. This effect is potentially linked to a decrease in fasting insulin secretion. Besides, BMI may have some influence on these relationships. Interestingly, our findings do not suggest a similar risk reduction for T2DM with various complications. This disparity could be attributed to the typically milder nature of T2DM when it occurs without complications, as opposed to the more severe forms that present with complications. These results underscore the prospective health benefits of dried fruits and offer valuable insights for the primary prevention of T2DM in everyday health practices.
Acknowledgements
The authors thank all contributors within the MAGIC consortium. Data on glycaemic traits were obtained from the Meta-Analyses of Glucose and Insulin-related Traits Consortium investigators, accessible via https://www.magicinvestigators.org/. Additionally, sincere appreciation is extended to the GWAS consortia, the UKB and the FinnGen database for their invaluable provision of summary-level data.
The present study was funded by the National Natural Science Foundation of China (No. 8 217 012) and the Key Project of Research and Development Plan (No. 2022YFF1202600).
The authors’ contributions are as follows: L. G. was responsible for the study design, data analysis and manuscript drafting. Z. W. and L. G. provided assistance in data analysis. X. L. and J. W. were involved in revising the manuscript critically. G. W. conducted a thorough review and editing of the manuscript. All authors made substantive contributions to the article and have given their approval for the final version to be submitted.
The authors have no financial or personal conflicts of interest to declare.
Supplementary material
For supplementary material/s referred to in this article, please visit https://doi.org/10.1017/S0007114524001879
