Causal associations between HbA1c and multiple diseases unveiled through a Mendelian randomization phenome-wide association study in East Asian populations

Abstract

Most analyses of hemoglobin A1c (HbA1c) and multiple common diseases have focused on European populations, thus there is a need for Mendelian randomization phenome-wide association study (MR-PheWAS) in East Asian populations. We used MR-PheWAS to investigate the potential causal associations between HbA1c and 159 types of diseases in the Biobank Japan dataset, employing the inverse variance weighted as the primary statistical approach, supplemented by MR-Egger and weighted median analyses. Additionally, multiple sensitivity analyses were conducted to assess heterogeneity and pleiotropy. High HbA1c levels are associated with an increased risk of type 1 diabetes (odds ratio [OR] = 4.07; 95% confidence interval [CI]: 2.34~7.07), type 2 diabetes (OR = 4.76; 95% CI: 3.01~7.55), cataract (OR = 1.33; 95% CI: 1.18~1.51), diabetic nephropathy (OR = 5.70; 95% CI: 2.24~14.46), and peripheral arterial disease (OR = 1.62; 95% CI: 1.29~2.04). Conversely, elevated HbA1c levels are associated with a reduced risk of asthma (OR = 0.76; 95% CI: 0.67~0.86), breast cancer (OR = 0.75; 95% CI: 0.65~0.87), and cerebral aneurysm (OR = 0.71; 95% CI: 0.57~0.88). The results of the causal association between HbA1c and numerous diseases in East Asian populations provides insights for the region’s specialized glycemic control and disease prevention programs, as well as new preventive and treatment options.

1. Introduction

The development of hemoglobin A1c (HbA1c) happens through a non-enzymatic interaction between glucose and hemoglobin, offering insight into the average blood sugar levels maintained over the preceding 2 to 3 months.^[1–3] HbA1c is widely measured in clinical studies^[4–6] and serves as a key biomarker for diagnosing diabetes and monitoring patient progress, a condition linked to a range of complications.^[7–14] Furthermore, extensive studies in European populations have unveiled possible relationships between HbA1c and diabetes-related genetic variants and various conditions like cardiovascular and cerebrovascular diseases,^[15,16] cancers,^[17,18] bone diseases,^[19–21] gastrointestinal diseases,^[22,23] eye diseases,^[24,25] and others. Single nucleotide polymorphisms (SNPs) that affect HbA1c in Europeans have similar correlations in Asian populations.^[26,27] However, the degree of association between allele frequencies and SNPs linked to HbA1c may be impacted by variations in genetic makeup across different racial and ethnic groups.^[28] Meanwhile, some hereditary kinds of anemia are more common in Asian people, indicating a higher sensitivity to HbA1c.^[29] Given the different reactions of populations to genetic alterations in HbA1c, additional data is needed to better understand its genetic effects on various illnesses in East Asian cultures.

Mendelian randomization (MR) establishes causation and overcomes analytical problems in typical observational research by using genetic variation.^[30–32] This method permits precise estimation of the causal effects of specific exposures on the risk of disease. Integrating MR methods with hypothesis-free phenome-wide association study (PheWAS) procedures allow for the identification of relationships between exposures and various disease outcomes or features.^[33–35] Thus, we investigate the potential association between HbA1c and multiple clinical outcomes using Mendelian randomization phenome-wide association study (MR-PheWAS) techniques in persons of East Asian heritage.

2. Materials and methods

2.1. Study design

Using MR-PheWAS, we examined the potential relationships between genetic variations related to HbA1c and different disease traits in an East Asian population. The effectiveness of MR analysis depends on 3 essential assumptions: the genetic variants selected as instrumental variables (IVs) must be strongly associated with the exposure. They should remain independent of any confounders, and their influence on the outcomes should occur only through the exposure.^[36–38]

2.2. Data sources

We conducted a 2-sample MR analysis using summary-level data from the Taiwan Biobank for exposure and the BioBank Japan for outcomes, ensuring independent, non-overlapping populations for exposure and outcome. The HbA1c data were obtained from 92,615 participants in Taiwan Biobank,^[39] with a mean (SD) HbA1c level of 5.764% (0.743). Outcome data, covering 159 International Classification of Diseases (ICD)-coded disease phenotypes, were sourced from BioBank Japan (https://pheweb.jp/downloads).^[40] Only summary-level data were used, with no access to individual-level information. We selected genetic instruments based on genome-wide significant SNPs (P < 5 × 10⁻⁸) associated with HbA1c. In total, 39 SNPs were identified as IVs, mapped to 31 genes. On average, each disease phenotype included around 13 million SNPs. For further details, refer to Table S1, Supplemental Digital Content, http://links.lww.com/MD/O529.

2.3. Selection of instrumental variables

In MR studies, SNPs serve as IVs to examine causal relationships. Each SNP strictly follows inclusion criteria, with a significance threshold set at a P value less than 5 × 10⁻⁸ to ensure a significant association between the IV and the exposure. Additionally, to avoid ambiguity, we excluded palindromic SNPs with intermediate allele frequencies using the default parameters of the TwoSampleMR package in R. To mitigate concerns about linkage disequilibrium, we set an r² threshold of 0.001 and a separation distance of 10,000 kb, as described previously.^[41–44] To quantify the strength of the IVs, we calculated the F statistic for each SNP using the formula^[45]:

$${\displaystyle \begin{array}{l}{\displaystyle F=\frac{R^2(\mathrm{n}-1-k)}{(1-R^2)k}}\end{array}}$$

where R² is the proportion of variance explained. k is the number of SNP, and n is the sample size. An F statistic >10 indicates a low risk of weak instrument bias.^[46]

2.4. MR-PheWAS analysis

In contrast to being confined to a single trait, MR-PheWAS enables the concurrent investigation of the relationship between a single exposure and multiple phenotypic outcomes. MR-PheWAS was employed to unveil potential correlations between HbA1c and a range of East Asian phenotypes. To reduce the probability of type I errors, we employed the false discovery rate method with a 5% threshold for multiple comparison correction.

The primary step in our analysis was the harmonization of HbA1c data with clinical outcomes. We employed the inverse variance weighted (IVW) method for MR estimation, supported by the weighted median and MR-Egger methods to improve the reliability and precision.^[30] Specifically, the IVW method was conducted using a multiplicative random effects model. For sensitivity analyses, we used: MR-Egger intercept to assess horizontal pleiotropy; Cochran’s Q statistic to evaluate heterogeneity among SNPs; funnel plots to visually assess potential directional pleiotropy; leave-one-out analysis to determine the influence of individual SNPs on causal estimates; MR-PRESSO to detect and correct for outliers.

MR analyses were conducted using the TwoSampleMR package in R. The primary outputs included odds ratios (ORs) with 95% confidence intervals (CIs), along with heterogeneity statistics and pleiotropy assessments. The input data comprised genome-wide association studies (GWAS) summary statistics, including effect sizes (β), standard error, P values, and effect allele frequencies (EAF), ensuring a robust analytical framework.

3. Results

3.1. Results of the MR-PheWAS analysis

We utilized 39 SNPs as IVs for HbA1c. All of the IVs exhibited an F statistic exceeding 10 (Table S2, Supplemental Digital Content, http://links.lww.com/MD/O530). After applying a 5% false discovery rate correction, 8 outcomes remained viable (Fig. 1). According to the IVW method, a significant positive association was observed between HbA1c and the following conditions: “type 1 diabetes (T1D)” (OR = 4.07; 95% confidence interval [CI]: 2.34~7.07), “type 2 diabetes (T2D)” (OR = 4.76; 95% CI: 3.01~7.55), “cataract” (OR = 1.33; 95% CI: 1.18~1.51), “diabetic nephropathy (DN)” (OR = 5.70; 95% CI: 2.24~14.46), and “peripheral arterial disease (PAD)” (OR = 1.62; 95% CI: 1.29~2.04). Moreover, an elevation in HbA1c levels was concurrently linked with decreased risks of asthma (OR = 0.76; 95% CI: 0.67~0.86), breast cancer (OR = 0.75; 95% CI: 0.65~0.87), and cerebral aneurysm (OR = 0.71; 95% CI: 0.57~0.88). Other than the opposite direction of beta for cerebral aneurysms observed in MR-Egger compared to IVW, the beta directions for the remaining 7 diseases were consistent across all 3 statistical methods (see Figs. 2 and 3 and Table S3, Supplemental Digital Content, http://links.lww.com/MD/O531).

Figure 1.

The distribution of P values for the associations between HbA1c and 159 phenotypes in the MR analysis. Line 1 indicates the significance threshold adjusted for the false discovery rate, while line 2 represents the significance threshold set at P = .05. HbA1c = hemoglobin A1c, MR = Mendelian randomization.

Figure 2.

Causal effects of HbA1c on 8 diseases in the MR analysis, showing ORs and corresponding 95% CIs. CIs = confidence intervals, HbA1c = hemoglobin A1c, IVW = inverse-variance weighted, MR = Mendelian randomization, ORs = odds ratios, WM = weighted median.

Figure 3.

Scatter plots indicating the causal associations between HbA1c and 8 diseases. HbA1c = hemoglobin A1c, MR = Mendelian randomization, SNP = single nucleotide polymorphism.

3.2. Results of the sensitivity analysis

Table S4, Supplemental Digital Content, http://links.lww.com/MD/O532 and Figure 4 depict the results of Cochran’s Q statistic and funnel plots, revealing heterogeneity (P < .05) in several analyses. Table S5, Supplemental Digital Content, http://links.lww.com/MD/O533 and Figure 3 present the results of the MR-Egger test, which did not detect the presence of horizontal pleiotropy. Outliers were detected (P_{Global Test} < .05) within the MR-PRESSO analysis (refer to Table S6, Supplemental Digital Content, http://links.lww.com/MD/O534). Nevertheless, these outliers did not significantly influence the final outcomes. Moreover, as depicted in Figure 5, the outcome remained unaffected when SNPs were eliminated one by one in the leave-one-out analysis.

Figure 4.

Funnel plots of the analyses. MR = Mendelian randomization, SE = standard error.

Figure 5.

Leave-one-out sensitivity analyses using the IVW method to investigate the causal estimates of HbA1c on 8 diseases after excluding a particular SNP from the analysis. HbA1c = hemoglobin A1c, IVW = inverse-variance weighted, SNP = single nucleotide polymorphism.

4. Discussion

Our MR-PheWAS analysis focused on the East Asian population to elucidate the causal relationships between HbA1c and various diseases. The findings indicate that elevated HbA1c levels were associated with an increased risk of diabetes, PAD, DN, and cataracts. Notably, genetically higher HbA1c levels were found to be negatively associated with the risk of asthma, breast cancer, and cerebral aneurysms.

4.1. HbA1c and diabetes and associated complications

This study found that elevated HbA1c levels were significantly associated with an increased risk of diabetes and its complications in the East Asian population. HbA1c is a well-established diagnostic marker for diabetes and a predictor of diabetes-related complications,^[47–49] with similar clinical applications reported in both Western and Eastern populations.^[50–54] However, its accuracy as a diagnostic tool for T2D varies across ethnic groups. A large multiethnic GWAS identified 42 novel and 18 previously known genetic variants associated with HbA1c, explaining 4% to 14% of trait variation. Notably, the genetic contributions to HbA1c levels differed significantly across populations, with particularly pronounced variations in African American individuals, which substantially affected the accuracy of HbA1c as a diagnostic marker for T2D.^[55]

The association of HbA1c and DN is supported by previous studies, including multiple European studies that have demonstrated a significant correlation between HbA1c levels and DN risk.^[56–58] For instance, a Swedish cohort study involving 10,398 children and adults found no statistically significant difference in the risk of retinopathy or nephropathy between individuals with an average HbA1c level of <6.5% (<48 mmol/mol) and those with HbA1c levels of 6.5% to 6.9% (48–52 mmol/mol) after 8 to 20 years of follow-up following a diagnosis of T1D. However, when the average HbA1c level exceeded 7.0% (≥53 mmol/mol), the prevalence of any retinopathy (defined as simplex or more severe forms) and microalbuminuria became slightly more common.^[56] Additionally, a Cox regression analysis of 3220 Chinese patients with T2D aged ≥ 30 years without DN found that annual fluctuations in HbA1c were predictive of DN risk in patients aged 30 to 89 years. Compared with an HbA1c coefficient of variation ≤ 6.68%, an HbA1c–coefficient of variation > 13.44% was associated with a 1.58-fold increased risk of DN.^[59] The potential mechanisms underlying this association may involve HbA1c-induced oxidative stress. Elevated HbA1c levels stimulate the polyol pathway, promote the formation of advanced glycation end products (AGEs), and activate protein kinase, all of which contribute to increased reactive oxygen species levels and oxidative stress.^[60–63] Elevated renal reactive oxygen species levels can result in damage to crucial cellular components and DNA, along with endothelial impairments,^[64,65] a characteristic of T2D and diabetic kidney disease.^[66–68]

In Taiwan, a case-control study demonstrated an association between high-risk PAD and significant variability in HbA1c readings.^[69] Similarly, in Koreans diagnosed with T2D, HbA1c was significantly correlated with PAD.^[70] However, the UK Prospective Diabetes Study found that improving HbA1c management did not reduce the incidence of severe vascular disorders, such as PAD.^[71] This observation is complex, as HbA1c only reflects average blood glucose over the previous 8 to 12 weeks, without accounting for glucose fluctuations,^[3,72] suggesting that blood glucose fluctuations beyond HbA1c levels may also contribute to vascular complications.^[73–76]

A study based on the Korean KoGES cohort (≥50 years old; 1972 age-related cataracts [ARC] cases and 38,290 healthy controls) investigated the association between metabolic syndrome and its components with ARC. The results showed that elevated HbA1c levels were significantly associated with an increased risk of ARC (OR = 1.92),^[77] consistent with the findings of the present study. The potential mechanism may involve the accumulation of AGEs, which increase with rising HbA1c levels and accumulate in the lens, leading to structural and functional abnormalities.^[78–81] AGEs may also induce oxidative stress, further damaging lens proteins and accelerating cataract development.^[82–85] However, a community-based cross-sectional study in Saudi Arabia (334 patients with T2D) found no association between HbA1c and cataracts.^[86] Similarly, Esteves et al^[87] in Brazil reported comparable findings, attributing the discrepancy to study design or good metabolic control in their sample.

4.2. HbA1c and asthma, cerebral aneurysm, breast cancer

This study found a negative association between HbA1c levels and the risk of asthma and cerebral aneurysms, which is not entirely consistent with previous studies. For instance, a prior cross-sectional study of 47,606 UK adults with asthma found a significant positive correlation between HbA1c levels, prediabetes or diabetes (defined by elevated HbA1c), and increased asthma-related hospitalization rates.^[88] Additionally, a retrospective cohort study of 5722 US adults aged 18 to 64 with obesity and asthma also reported that higher HbA1c levels were associated with an increased frequency of asthma attacks.^[89] These discrepancies may be explained by factors such as inadequate treatment of individuals with HbA1c levels in the diabetic range, potentially exacerbating asthma due to other chronic disease factors. Moreover, the lack of data on potential confounding or modifying factors, such as asthma medications and dyslipidemia treatment, may contribute to these differences. A cross-sectional study conducted in China involving 223 patients with a single ruptured intracranial aneurysm and diabetes found a non-linear relationship between HbA1c levels and aneurysm rupture, with inflection points at 5.5 and 8.9.^[90] A retrospective study in the United States, analyzing medical records of 4701 patients with 6411 intracranial aneurysms from 2 hospitals, examined the relationship between aneurysmal subarachnoid hemorrhage and HbA1c levels using logistic regression and propensity score weighting, and found no significant correlation between HbA1c levels and aneurysmal subarachnoid hemorrhage.^[91] These differences may also be attributed to the observational nature of the studies, which cannot fully exclude the influence of confounding factors.

Breast cancer primarily affects women and is a leading cause of death globally, with a significant proportion of fatalities occurring in developing countries.^[92–95] The relationship between HbA1c levels and breast cancer risk has been supported by previous studies. A prospective research found an inverse relationship between HbA1c levels and estrogen receptor-negative breast tumor risk in postmenopausal women.^[96] Concomitantly, HbA1c is significantly inversely related to stage I breast tumor risk but might increase the risk for stage II–IV tumors.^[97] A possible underlying mechanism may involve insulin-like growth factor 1 (IGF-1), which has been implicated in cancer development.^[98] Higher IGF-1 levels have been positively correlated with HbA1c,^[99,100] and IGF-1 gene polymorphisms have been associated with breast cancer risk, with potential variations based on menopausal status or tumor subtype.^[101] Moreover, the role of insulin in breast cancer progression may be attenuated in individuals with low or absent estrogen levels,^[96] suggesting that estrogen modulates IGF-1 function in breast cancer. The complex interplay between female hormones and HbA1c levels contributes to the heterogeneity of breast cancer subtypes, warranting further investigation.

While previous studies have examined the association between HbA1c and diabetes-related diseases, our study systematically applies an MR-PheWAS approach in an East Asian population. Unlike traditional observational studies, MR analysis mitigates confounding and reverse causality, providing stronger evidence for causal relationships. By leveraging public GWAS data, our study efficiently explores the causal effects of genetically predicted HbA1c on a broad spectrum of diseases, uncovering both established and potentially novel associations. This approach not only enhances the reliability of our findings but also offers a cost-effective and scalable framework for future research.

Nonetheless, studies limited to East Asian populations may fail to generalize persuasively to other ethnic groups. Furthermore, MR studies’ linear causation assumption may oversimplify the complex link between HbA1c and illness. Finally, because of the sample size limits inherent in GWAS datasets, there is still a danger of selection bias, since they may not completely reflect the larger population.

5. Conclusion

The MR-PheWAS analysis generally revealed a significant positive causal association between HbA1c and T2D, T1D, cataract, PAD, and DN, while a negative correlation was observed between HbA1c and asthma, breast cancer, and cerebral aneurysms. These findings highlight the complex role of HbA1c in various diseases in East Asian populations, offering potential guidance for tailored medical interventions and disease prevention strategies aimed at improving public health outcomes.

Author contributions

Writing – original draft: Li Han, Shuling Xu, Rumeng Chen, Zhiwei Zheng, Yining Ding, Zhu Wu.

Writing – review & editing: Sen Li, Binsheng He, Meihua Bao.