Association between haemoglobin A1c and cerebral microbleeds in community‐based stroke‐free individuals: A cross‐sectional study

The association between haemoglobin A1c (HbA1c) and cerebral microbleeds (CMBs) remains unclear. We aimed to investigate the association between HbA1c and CMBs in community‐based individuals without stroke or transient ischaemic attack (TIA) and whether the association differs between individuals with and without diabetes mellitus (DM).


| INTRODUCTION
Cerebral microbleeds (CMBs), a neuroimaging marker of cerebral small vessel disease (CSVD), are characterised by small foci of blood cell leakage. 1,2 CMBs can predict stroke and increase the risk of death from all causes. 3 Therefore, it is critical to identify the risk factors for CMBs and facilitate the primary prevention of cerebrovascular diseases.
Increasing evidence has shown that diabetes mellitus (DM) can damage microvessels, 4 and several studies have suggested that patients with DM were more likely to have CMBs. [5][6][7][8][9][10][11][12][13] Haemoglobin A1c (HbA1c) has been recognised as a biomarker that reflects the longterm status of glycaemic control in patients with DM, 14 and reduction in HbA1c levels can reduce the risk of microvascular diseases. 15 However, it is unclear whether HbA1c contributes to the development of CMBs. Furthermore, most of the previous studies on the association between HbA1c and CMBs focussed more on hospitalbased patients with stroke. 16,17 However, few studies have been conducted in stroke-free populations.
The present study aimed to investigate the association between HbA1c levels and CMBs in individuals free of stroke or transient ischaemic attack (TIA) and whether the association varies between individuals with and without DM.

| Study population
All individuals were recruited from a community study known as Cardio-and cerebrovascular Accident Monitoring, Epidemiology, and caRe quAlity System (CAMERA), which aimed to investigate the cerebrovascular disease risk in a community-based population. 18 The present study specifically recruited individuals aged 18-85 years who participated in the CAMERA study from January 2015 to September 2019. Individuals with the following conditions were excluded from this study: (1) known malignant tumours; (2) severe clinical conditions (heart failure, hepatic failure or renal failure); (3) stenting therapy history; (4) contraindications to magnetic resonance imaging (MRI); (5) pregnancy; (6) absence of susceptibility-weighted imaging (SWI) images or MRI with poor image quality; (7) history of stroke or TIA (the purpose of excluding patients with stroke or TIA is to avoid overestimating the prevalence of CMBs); and (8) unsuitability for HbA1c in the assessment of glycaemic status, such as haemoglobinopathies or red blood cell disorders.
We did not exclude individuals on medication or therapy affecting platelet counts, and prothrombin time or partial thromboplastin time was not tested in our study population.

| Data collection
The individuals were interviewed face-to-face by trained coordinators who were blinded to the MRI data. General demographic characteristics, behavioural lifestyle and medical history were collected through a questionnaire. Hypertension was defined as self-reported medical history or having taken hypertension agents in the previous 2 weeks. DM was defined according to the American Diabetes Association criteria in 2022 14 as fasting blood glucose (FBG) ≥ 7.0 mmol/L or HbA1c ≥ 6.5% or self-reported or use of oral hypoglycemic agents or insulin in the last 2 weeks. Dyslipidemia was defined as self-reported medical history or having taken lipid lowering agents in the previous 2 weeks. Measurements of weight, height and blood pressure were performed by trained nurses. Body mass index was calculated as the weight in kilograms divided by the square of the height in metres. The average of two measurements recorded in the right arm with a rest period of 5 minutes was used to determine the systolic and diastolic blood pressure.

| Measurement of haemoglobin A1c (HbA1c) and other biochemical parameters
The following fasting blood sample parameters were evaluated: HbA1c, FBG, low-density lipoprotein-cholesterol (LDL-C), highdensity lipoprotein-cholesterol and high-sensitivity C-reactive protein. All measurements were conducted in the central laboratory of the Beijing Tiantan Hospital.

| Reproducibility study
The MR images of 40 individuals were randomly selected for the intra-observer and inter-observer reproducibility study. Two observers blinded to clinical information independently evaluated the presence and number of CMBs on SWI images. One observer interpreted the SWI images again after a time interval of 3 months for the purpose of minimising memory bias.

| Statistical analyses
The normal distribution of the data was determined using the Kolmogorov-Smirnov test. Quantitative variables with normal distribution were presented as mean � standard deviation, and variables with non-normal distribution were summarised as median and interquartile ranges. Qualitative data were described as frequencies and percentages. Quantitative data were compared using an independent t-test (normal distribution) or Wilcoxon test (non-normal distribution). Qualitative data were compared using the Chi-square test, and Fisher's exact test was used if ≤20% of the expected cell counts were less than five. Multivariable logistic regression was performed to estimate the odds ratio (OR) and corresponding 95% confidence interval (CI) of HbA1c in discriminating the presence of CMBs by adjusting for potential confounders which were significant in univariate analysis (p < 0.05) and a history of DM. Multinomial logistic regression was used for assessing the association between HbA1c and the location of CMBs. The individuals without CMB were used as a reference. A subgroup analysis of individuals with and without DM was performed. Cohen's kappa analysis was utilised to determine the intra-observer and inter-observer agreement in identifying the presence of CMBs. The intraclass correlation coefficient was calculated to determine the intra-observer and interobserver agreement in evaluating the number of CMBs. A twotailed p value < 0.05 was considered statistically significant. All statistical analyses were performed using SAS software version 9.4 (SAS Institute Inc., Cary, NC, USA).

| General characteristics of the study population
In total, 626 individuals were recruited from the Tsinghua community in the CAMERA study between January 2015 and September 2019.
After excluding individuals who had MRI contraindications (n = 32) or a history of stroke or TIA (n = 50), a total of 544 individuals were finally included in the statistical analysis ( Figure 2 Table 2).

| DISCUSSION
This study investigated the association between HbA1c levels and CMBs in community individuals without stroke or TIA. We found that HbA1c was significantly associated with CMBs after adjusting for potential confounders. In the subgroup analysis, HbA1c was significantly associated with CMBs in individuals with DM, whereas this association was not statistically significant in individuals without DM.
Our findings indicate that the association between HbA1c and CMBs was more pronounced in individuals with high blood glucose levels, suggesting that poor blood glucose control may increase the risk of CMBs in individuals with DM.
The prevalence of CMBs in our study was 21.88%, which is higher than the result of the Taizhou study (18.51% 20 ) and two-fold higher than that of the Shunyi study (  There was a significant association between HbA1c and CMBs in our study, suggesting an elevated HbA1c level increases the risk of CMBs. Our findings are consistent with some previous studies.  28 and enhancing the production of mitochondrial reactive oxygen species (ROS), 29 leading to the formation and deposition of advanced glycation end products, 4 which are associated with the onset and progression of diabetes. 30 Second, the overproduction of ROS and increased glycosylation of haemoglobin due to the suppression of the oxygen-carrying capacity of haemoglobin lead to tissue hypoxia and microbleeds. 31 Third, high HbA1c levels impair myogenic response by diminishing the contractile capability of cerebral vascular smooth muscle cells. 29 After the arteriole loses tension, the elasticity of the vessel wall decreases, and cerebral blood flow is enhanced with transient hypertension, leading to vascular rupture and CMB formation.
The major strengths of our study are as follows: First, our participants were recruited from a community without stroke or TIA, indicating the significant importance of glycaemic control in strokefree individuals. Second, compared to previous studies investigating the association between history of DM or fasting glucose and CMBs, [32][33][34][35][36][37] HbA1c was used to investigate the association between glucose status and CMBs in the present study. It may provide more reliable and precise results on the association between long-term glycaemic status and CMBs. Third, the use of SWI guaranteed the accurate detection of CMBs. 1 However, our study has some limitations. First, unmeasured variables may have residual confounding effects on our findings. A potential confounding effect from anticoagulant use could not be adjusted in the analysis because only one participant in our study used an anticoagulant, which was warfarin.
Second, only the association, but not a causal effect, between HbA1c and CMBs can be investigated based on this cross-sectional study.
Third, all individuals were recruited from the community, which may indicate a possible selection bias. Despite a moderate sample size of 544 individuals, the findings may not be more representative than those inferred from a larger sample size. These findings should be further verified in future multicentre, large-scale, prospective cohort studies.

| CONCLUSION
In our study, HbA1c was associated with CMBs in participants without stroke or TIA, particularly in individuals with DM. This finding suggests that the status of glycaemic control warrants attention for the prevention of CMBs.

AUTHOR CONTRIBUTIONS
Yongjun Wang, Gaifen Liu and Xihai Zhao designed the study. Xihai Zhao and Gaifen Liu made critical revisions to the manuscript.