The impact of type 1 diabetes mellitus on intellectual status: A Mendelian randomization study Authors: Rong Bao, Zhangyan Shi, Hanteng Wang, Puyan Zhu, Bingyu Zhang, Haixiang Zhang, Kejin Zhang Published: 2025 ABSTRACT BACKGROUND:The status and changes of intelligence are closely related to the treatment and quality of life of patients with type 1 diabetes mellitus (T1DM), but the relationship between the disease and the decline of patients’ intelligence and the rise of the risk of intellectual disability (ID) is still uncertain.MATERIALS AND METHODS:A Mendelian randomization (MR) analysis was conducted using the inverse-variance-weighted (IVW) method. Additional MR methods, including MR-pleiotropy residual sum and outlier (MR-PRESSO), radial MR, and other extended methods, were employed to address potential confounding factors.RESULTS:The MR analysis demonstrated a causal association between T1DM and lower intelligence (β = −0.030, 95% CI: −0.040 to − 0.020, P = 1.27E-08) as well as an increased risk of ID (β = 0.062, 95% CI: 0.037 to 0.087, P = 7.10E-07). Consistent results were observed with extended MR methods (all Ps < 0.05). In addition, T1DM significantly reduced the rate of structural changes in the brain during development and aging, including the changes in total brain volume, surface area, and volume of the hippocampus and thalamus, which also supported the causal relationship observed.CONCLUSIONS:This MR study provided robust evidence that T1DM may lead to a decline in intelligence and an increased risk of ID. Future research, analyzing the mechanism of T1DM occurrence on patients’ intellectual status and the characteristics of intelligence changes, would help improve the treatment effectiveness and quality of life of T1DM patients. Introduction Type I diabetes mellitus (T1DM), also called autoimmune diabetes, generally shows up during childhood and adolescence, usually at around 13–14 years of age or even earlier. According to the 2021 International Diabetes Federation Diabetes Atlas, there are more than 1,211,900 children and adolescents under 20 years old suffering from this disease globally,[1] its incidence is still on the rise.[2] For most patients with T1DM, the main approach in treatment is to carry out complex daily management to control hyperglycaemia, which makes it highly reliant on their intelligence condition.[3] Increasing emphasis is being placed on interventional therapy and psychological approaches in future treatment of patients. Therefore, for diabetic children and adolescents to lead a healthy life and effectively manage the disease from the time of diagnosis, they need to leverage the connection between T1DM and their intelligence. The crucial role of the patient’s intellectual status in managing this disease is mainly reflected in the following aspects: (1) daily management, such as subcutaneous insulin injection, blood glucose monitoring, and dosage adjustment according to food intake, requires a certain level of intelligence and understanding[4]; (2) the implementation level of the complex treatment plan is a significant factor affecting the treatment outcome and requires a strong execution ability,[5] which is a manifestation of an individual’s cognitive ability; additionally, (3) knowledge about the disease and diabetic education is important for treatment.[6] However, the relationship between T1DM and the patients’ intellectual status remains uncertain. Previous studies have demonstrated correlations between T1DM, decline in intelligence, and increased risk of intellectual disability (ID, previously known as mental retardation).[7] Patients with early-onset diabetes (<5 years at diagnosis) may have lower scores on cognitive tests compared to those with later-onset, and this will persist into adulthood.[8] Early-onset patients (mostly with T1DM) are at a critical stage of central nervous system (CNS) development. Frequent occurrences of extreme glucose (e.g. hypoglycemia, hyperglycaemia, and ketoacidosis) can affect the structure and function of the brain, resulting in decline and impairment,[9] especially in children and adolescents. Such abnormal changes in the brain structure of patients from the time of diagnosis to adulthood will have an impact on their psychological and mental performance.[10] However, negative correlations and reverse associations have also been reported. For example, Ohmann[11] believed that cognitive impairment might be independent of metabolic quality or disease duration, and deficits in general intelligence (g), psychomotor efficiency, and cognitive flexibility might precede the onset of T1DM.[12] Therefore, whether this relationship stems from a real causal connection, confounding factors, or a reverse causal relationship remains a topic of ongoing debate. Mendelian randomization (MR) presents a novel approach that leverages the random process of allele assortment at conception leading to an effective “randomization” of individuals to higher or lower genetic risk of an exposure condition such as a history of T1DM, to estimate the causal effect of such exposure status on the outcomes of interest.[13] With the recent proliferation of large-scale genetic data, this method has been widely used in the medical literature to investigate the causative pathways between complex exposure and outcome traits.[14] Here, to determine the causal impact of T1DM on the risk of intelligence decline and increased risk of ID, a meticulous MR study was conducted using available summary-level statistical data from genome-wide association tests (GWASs). Methods and Materials Study design and setting In this study, MR analysis was conducted to determine the causal relationship between T1DM and intellectual status, using the summary-level statistic data of GWASs. The study was conducted from January to July 2024. The MR study followed the STROBE-MR guidelines[15] and the MR analysis handbook.[16] To accurately infer the causal relationship between exposure and outcome, the study must meet the three assumptions (to see, the Supplementary Files): relevance, independence, and exclusion [Figure 1]. The flow diagram of Mendelian randomization analysis. IVs, instrumental variables; SNP, single nucleotide polymorphisms; T1DM, type 1 diabetes mellitus; IQ, intelligence quotient; ID, intellectual disability Study participants and sampling The population samples in this study were all Europeans, according to Haycock’s recommendation[17] that participants with shared ancestry can weaken the potential impact of population stratification on MR analysis results. The summary-level statistical data for T1DM were derived from 15 GWAS studies. Patients in these studies (963 to 6,729 cases) were diagnosed by the International Classification of Diseases (ICD) – 10 criteria. Four studies involved T1DM patients without any complications (codes: E109), nine involved those with complications (codes: E100 ~ E108), and two were unmentioned (codes: E109). Approximately 773,135 (including 45,812 cases) participants were involved in total. The data of intelligence extracted from Benyamin’s study,[18] in which the intelligence measurement scores of 17,989 children aged 6–18 years were investigated. The data of ID were from the FinnGen Biobank dataset,[19] which includes 761 ID patients and 218,031 controls. In Brouwer’s study,[20] eight global brain measurements and seven subcortical structures were used as indicators to describe the rates of brain structural changes. Genetic variants independent of age and associated with longitudinal changes in brain structure were identified in more than 15,000 participants (to see Supplementary File). Data collection tool and technique The summary-level data of GWAS studies for T1DM, intelligence levels, and ID were downloaded from publicly available resources listed in the Supplementary File [Table S1]. The data of the changes in brain structure covering the lifespan were downloaded from the ENIGMA plasticity working group (https://enigma.ini.usc.edu). In the first step, to construct a robust instrumental variable (IV) for MR analysis, variants (usually single nucleotide polymorphism [SNP]) genetically associated (P < 5E-8) with T1DM were extracted from the summary-level statistical data of GWASs. Then, clumping processing (r2 < 0.001 within 10,000 kb windows) was employed to obtain independent SNPs. Then, these independent SNPs [Table S2] were included to construct IV and used to investigate the genetic association between T1DM and outcomes of interest. The possibility of false positive associations due to weak instrument bias was assessed by calculating the sum of the R2 values of the variance of all SNPs and the F statistics. The detailed calculation of these parameters was outlined in the Supplementary File. Additionally, to access the influence of the presence of T1DM complications on the relationship between T1DM, intelligence level, and ID risk, two sub-IVs were constructed with SNPs associated with T1DM patients with and without complications respectively, for following multivariant MR analysis. Considering the potential confounding effects of linkage disequilibrium and over-counting of SNPs, the tag-SNP selection and new MR analysis will be conducted as described in the Supplementary File and Table S2. In the second step, intelligence level, risk of ID, and the rates of change of 15 brain structural measures were used as the outcomes of interest in this MR analysis. The effect of each SNP included in IVs on these outcomes was extracted, respectively. To ensure that effect estimates were aligned to the same allele, the independent SNPs associated with T1DM were harmonized with variants from these outcomes (described in the Supplementary File). Ethical consideration This study was conducted following the principles of the Declaration of Helsinki and received approval from the institutional review board at Medical School, Northwest University (Decision No: 240305006-1). Although the analysis plan was not preregistered, all analytical methods were developed before the start of the study. Statistical analysis The primary MR analysis used the inverse-variance-weighted (IVW) method. This method combines the Wald ratios of the causal effects of each SNP through meta-analysis, providing the most precise estimates and having greater statistical power. Cochrane’s Q value and Rucher’s Q statistics were used to assess the heterogeneity among these genetic variants. If there was significant heterogeneity, the random effects model of the IVW method would be adopted. RadialMR analysis was used to identify outliers; after removing the outliers, the MR analysis was re-conducted. For sensitivity analysis, five extended methods of MR analysis, including the MR-Egger, weighted median, simple mode, weighted mode, and MR-PRESSO methods, were used to validate the results analysed by the IVW method and detect whether there was pleiotropic bias in our genetic instruments.[16] The consistency of these MR methods was considered and visualized using a forest plot. The “leave-one-out” method (by sequentially removing an SNP from IV and running MR analysis) was used for sensitivity analysis. In addition, a multivariate MR analysis was performed to assess the impact of mediation caused by different complications of T1DM.[21] All analyses were conducted using the “TwoSampleMR” package (version 0.5.2)[22] and “RadialMR” package in R software (version 4.1.2).[23] Statistical significance was set at P < 0.05. Results Strength of IVs In the primary MR analysis, IV, which included 171 genome-wide SNPs associated with T1DM, explained 46.90% of the variance in the corresponding log odd ratio (OR) risk of T1DM. The mean F-statistic, another parameter for measuring the strength of this instrument, was 518 (range 30–4106), indicating that this instrument was strong for MR analyses with a low risk of weak instrument bias [Table S2]. Effect of T1DM on intelligence The IVW method showed a significant association between T1DM and intelligence scores (β = −0.030, 95% CI: −0.040 to −0.020, P = 1.27E–08; Figure 2). This finding indicates that T1DM can lead to substantially lower intelligence. Neither Cochrane’s Q for IVW (Q = 63.67, df = 50, P = 0.093) nor Rucher’s Q statistics for MR-Egger (Q = 68.18, df = 51, P = 0.054) provided evidence of significant heterogeneity among the genetic instruments [Table S3]. The intercept from the Egger regression was 0.006 (se = 0.003, P = 0.066), indicating that no significant pleiotropic bias was found in this MR analysis. The “leave-one-out” sensitivity analysis also showed that the direction and precision of the summary association remained largely unchanged when each SNP was sequentially removed. Effect of T1DM on childhood intelligence estimated by Mendelian randomization. Two-sample was performed using a genetic instrument containing 109 independent SNPs associated with T1DM. (a) Each point represents the SNP effects on T1DM and childhood intelligence score. Colored lines represent the results analysed with five different MR methods: inverse-variance-weighted (IVW, in light blue), weighted median (WMe, in deep green), MR-Egger (ME, in blue), Weighted mode (WMo, in red), and Simple mode (SM, in green), which described the association between a 1-unit log odd ratios (ORs) of T1DM event and 1-SD of childhood intelligence score; and (b) the effects (beta values), 95% confidence intervals (CIs), and P values of each method had been visualized with a forest plot The consistent results estimated by five extensions of the MR test indicated the stability of our primary MR analysis [Figure 2]. Although eight outliers (rs11651753, rs12665429, rs13716, rs2111485, rs2395360, rs2816313, rs722988, and rs7772682; Table S2) were found by RadialMR analysis, the same genetic effect (β = −0.032, se = 0.005, t = −5.54, P = 1.14E–10) was still observed after removing these outliers [Figure S1]. Effect of T1DM on ID risk The results of IVW method showed that the occurrence of T1DM increased the risk of ID in patients (β = 0.062, 95% CI: 0.037 to 0.087, P = 7.10E–07; Figure 3). Compared with healthy participants, T1DM patients had a higher risk of ID (OR = 1.064, 95%CI: 1.037–1.091). Both Cochrane’s Q for IVW and Rucher’s Q statistics for MR-Egger did not provide evidence of significant heterogeneity among our genetic instruments [Table S3]. The intercept of Egger regression also indicated no significant pleiotropic bias in this MR analysis. In addition, the results of the other five MR methods were consistent [Figure 3]. The “leave-one-out” sensitivity analysis also showed that the estimated effect of T1DM on ID was stable. Furthermore, eight outliers (rs116371893, rs1233386, rs12927355, rs56994090, and rs7772682; Table S2) were detected from 171 SNPs by the RadialMR method. After removing these outliers, the results also showed the same trend (β = 0.059, se = 0.013, t = 4.68, P = 3.0E–06; Figure S1). Effect of T1DM on ID by Mendelian randomization. (a) Each point represents the SNP effects on T1DM and the risk of ID. Colored lines represent the results analysed with five different MR methods: inverse-variance-weighted (IVW, in light blue), weighted median (WMe, in deep green), MR-Egger (ME, in blue), Weighted mode (WMo, in red), and simple mode (SM, in green), which described the association between a 1-unit log odd ratios (ORs) of T1DM event and the risk of ID; and (b) the effects (ORs), 95% confidence intervals (CIs), and P values of each method had been visualized with a forest plot Multivariable MR and reverse association estimation As described in Table S2, some SNPs (n = 63) were associated with T1DM patients with complications (such as renal, ophthalmic, or other complications), some (n = 67) associated with strict T1DM patients, and some identified with mixed T1MD sample. To determine whether the relationship between T1DM, intelligence, and ID was influenced by the presence of complication (s), a multivariable MR analysis was performed. The analysis revealed that both exposure conditions, whether T1DM patients had complications or not, showed a significant causal association with intelligence decline and the risk of ID (all Ps < 0.05; Table S4). No any reverse association was observed by followed reversive MR analysis (all Ps > 0.05; Table S5). Tag-SNP selection and new MR analysis Figure S2 listed 11 tag-SNP loci selected by the SNPinfo Web Server (https://snpinfo.niehs.nih.gov/), which covered 55 T1DM-associated SNPs located in the 6th chromosome (chr6: 26,359,866 to chr6: 33,917,860 bp). Combined with other SNPs located in other chromosomes region, a more rigorous and reliable IV was constructed, and a new MR analysis was performed. The results with IVW and MR-Egger methods yielded a consistent causal relationship between T1DM, intelligence level, and ID risk (all Ps < 8.2E-04; Table S6) Effect of T1DM and the rates of brain changes To estimate the relationship between T1DM and the rates of brain changes in patients’ brain development and aging, an additive MR analysis was performed. The results revealed that T1DM was age-independent and associated with a significant reduction rate of change in total brain volume (β = −52.51, se = 19.67, P = 7.6E–03), surface area (β = −8.70, se = 2.56, P = 6.9E–04), and volume of the thalamus (β = −2.08, se = 0.43, P = 1.2E–06) and hippocampus (β = −1.00, se = 0.29, P = 4.9E–04; Table S7). Discussion An increasing number of researchers and clinicians are exploring the impact of T1DM on cognitive development due to its significance in diagnosis, treatment, monitoring, and quality of life. Whether T1DM causes intelligence decline, increased risk of ID, or even senile dementia is an attractive topic.[22] This MR study firstly provides consistent evidence that T1DM may lead to lower intelligence and a higher risk of ID in patients by affecting brain structure. The findings illuminate the complex relationship between T1DM and cognitive function, offering new directions for treatment, care, and cognitive intervention. Causality analysis is crucial in modern medicine, particularly for treating and managing common and complex diseases (e.g. T1DM), as it allows healthcare professionals to recommend appropriate therapies or interventions. However, for traditional epidemic, case-control, and clinical trial design, accessing the causal relationship between two complex variables is challenging due to unmeasured confounding (s), reverse causality, and other potential factors. Therefore, until now, there was still no consensus on the causal relationship between T1DM and the risk of decreased intelligence and cognitive impairment in patients.[89101112] To estimate the causal relationship between T1DM, intelligence decline, and the risk of ID, in this study, a two-sample MR analysis was conducted, because of its robustness to aforementioned issues.[23] In this method, a series of genetic markers are used as instrumental variables for analysis, which can effectively reduce the interference of potential problems (mentioned above) on the analysis results compared to traditional analysis methods.[16] The current MR study not only observed the impact of T1DM on the decline of patients’ intelligence level and increased risk of ID, also emphasized the crucial role of cognitive intervention for T1DM patients. The impact of T1DM on patients’ intelligence level and ID risk has already been indicated in epidemiological studies, clinical observations, and pathological research. Experts have put forward a widely accepted viewpoint that early-onset T1DM can negatively impact the development of CNS during childhood and adolescence due to the brain’s susceptibility to extreme glycemia,[24] leading to decreased intelligence, reduced psychomotor efficiency, and the occurrence of ID. Furthermore, complications in T1DM patients have also drawn the attracted of clinicians, as well as their impact on lifestyle changes, glycemic control, and intelligence changes in T1DM patients. For instance, more than 50% of T1DM patients may also develop chronic kidney disease over their lifetime,[25] which is well known to impair cognitive function.[26] Therefore, the occurrence of proliferative diabetes retinopathy has been suggested to estimate the risks for cognitive impairment in people with T1DM.[27] A single diabetic ketoacidosis episode was also reported associated with sizable IQ declines in newly diagnosed T1DM patients.[28] However, the causal relationship between them was still unclear. In this study, two-step MR analyses were involved to estimate the causal relationship between T1DM and patients’ cognitive dysfunction. The primary MR analysis showed that T1DM can substantially lead to a decrease in patients’ intelligence level and an increase in the risk of ID. The followed multivariable MR confirmed that this causal relationship was not affected by factors like T1DM comorbidities and extreme values [Table S4 and Table S5]. This two-step MR analysis reveals a robust causal relationship between the onset of T1DM and cognitive impairment, independent of comorbid conditions. An understanding of the functions of genes tagged by genetic variants included in IV also contributes to the interpretation of the causal relationship identified in this MR analysis. Lots of genetic variants included in IVs of current MR analysis were also identified as linked to cognitive impairment, neurotransmitter transmission, synapse formation, brain function, neurodegenerative diseases, dementia, and Alzheimer’s disease (to see the Supplementary Files). More intriguingly, one-third (55/171) of the SNPs were located in the major histocompatibility complex/human leukocyte region (MHC/HLA; Table S2). Genetic variation in this chromosomal region was also considered to have the most significant impact on the pathology of T1D.[29] Lots of genes in this region also regulate brain function and volume and are linked to memory consolidation, sensory processing, language impairment, cognitive decline, and other mental health phenotypes.[30] Regarding the potential confounding effects of linkage disequilibrium between SNPs and over-counting in the MHC/HLA region, only tag-SNP was included in IV [Table S2], and a consistent causal relationship was still observed. Findings of investigation between T1DM and the rate of changes in brain structure, which indicate that T1DM can impact the rate of changes in brain structure throughout patients’ development and aging, offered a more nuanced understanding of the causal relationship between T1DM and cognitive dysfunction. Hyperglycemia and disturbances in glucose metabolism, frequently associated with diabetes, can have substantial effects on brain structure and function.[31] Consequently, cognitive dysfunction is increasingly recognized as both a complication and comorbidity of diabetes.[32] The results of this MR study showed that the impact of T1DM was particularly evident in global brain volume, surface area, and volumes of thalamus and hippocampus [Table S7]. To our knowledge, there are significant phenotypic correlations between brain volume and cortical surface area and between human intelligence and ID.[33] The subcortical structures of the thalamus and hippocampus are also regarded as critical brain regions influencing patients’ intelligence, ID,[34] as well as spatial and emotional cognitive ability.[35] It offered a new perspective on the relationship between T1DM and cognitive dysfunction in patients and provided new targets for the diagnosis, treatment, and clinical monitoring of T1DM patients, such as focusing on diabetic education and spatial and emotional control abilities training. This study possesses several strengths. Firstly, its key advantage lies in the MR design, which not only provides causal inference regarding the association between T1DM and declining intelligence and increased ID risk but also avoids widespread biases typically present in observational and epidemiological studies. Secondly, the summary-level statistical datasets utilized in the T1DM and outcome GWASs were derived from several European populations and other larger available GWASs, thus preventing potential bias caused by differences in genetic backgrounds. Thirdly, the consistency of multiple MR analyses and sensitivity tests boosts the credibility of these findings. Limitation One should be cautious when interpreting the findings of this MR study, given the following concerns. To strengthen the statistical power of MR analysis, this study incorporated 15 publicly available GWAS datasets. While this approach enhanced the robustness of the findings, it also introduced substantial heterogeneity into the analysis. Although statistical assessments suggest that this heterogeneity did not substantially affect the MR results, its potential impact cannot be entirely ruled out. Considering the potential impact of population stratification on MR analysis, this study’s findings regarding the relationship between T1DM and cognitive dysfunction are currently limited to the European population. To establish more universal conclusions, future research should integrate MR studies conducted across diverse ethnic populations. Recommendation Clarifying the causal relationship between T1DM and decreased intelligence level, and increased risk of ID was crucial for the long-term treatment and disease control of T1DM. Meanwhile, the impact of T1DM on the rate of brain structure changes in patients provided detailed targets for our future treatment and intervention measures.[36] In the long-term treatment of T1DM, in addition to conventional drug therapy and disease monitoring, intervention in the patients’ intellectual level was also essential. In the long-term treatment of T1DM, in addition to conventional drug therapy and disease monitoring, intervention in the patients’ intellectual level was also essential. In the process of intelligence intervention, in addition to providing diabetes knowledge education, attention should also be paid to training and improving the spatial cognitive ability and emotional control ability of patients. In the process of intelligence intervention, in addition to providing diabetes knowledge education, attention should also be paid to training and improving the spatial cognitive ability and emotional control ability of patients. This can not only improve the therapeutic effect of T1DM but also reduce or delay the risk of cognitive impairment and even diabetic dementia. Conclusions In conclusion, this study provides evidence supporting a causal relationship between T1DM and declining intelligence and increased risk of cognitive impairment. In the daily management of T1DM patients, implementing routine intellectual measurement, monitoring intervention may help improve treatment effectiveness and patients’ quality life, slowing down the progression of dementia. Further investigation of this underlying link could also provide valuable insights into the development of intelligence and cognitive dysfunctions. Conflicts of interest There are no conflicts of interest.