Distinct eGFR trajectories are associated with risk of myocardial infarction in people with diabetes or prediabetes

The relationship between estimated glomerular filtration rate (eGFR) trajectories and myocardial infarction (MI) has so far been unclear in people with diabetes or prediabetes. We aimed to identify common eGFR trajectories in people with diabetes or prediabetes and to examine their association with MI risk.


Highlights
• Distinct trajectories of estimated glomerular filtration rate (eGFR) were associated with myocardial infarction (MI) risk in a diabetic or prediabetic population.
• The MI risk was higher in the high-decreasing group although the eGFR levels were similar to the moderate-stable group during the last exposure period.
• The moderate-increasing group still had a significantly increased MI risk when reaching the normal range.

| INTRODUCTION
Chronic kidney disease (CKD) is affecting 10% to 16% of the global adult population 1 and was a significant cause of several adverse clinical outcomes such as kidney failure 2 , cardiovascular disease, 3 and all-cause mortality. 4 Estimated glomerular filtration rate (eGFR) is a common and convenient indicator of CKD. Associations of eGFR and myocardial infarction (MI) were less consistent; some studies showed eGFR was related to a significantly increased risk of MI, [5][6][7] and others observed an unchanged risk with lower eGFR. 8,9 Since these studies measured eGFR at only one single time point, there has been no consideration of how eGFR varies within individuals over time and of its potential impact on the future risk, which may not be enough for characterizing longterm MI risk prediction.
Recently, changes in eGFR were measured with two major methods: percentage change using two measurements and slope-based approaches using multiple measurements, [10][11][12] both of which may largely ignore trajectory of eGFR over time. Accordingly, the African American Study of Kidney Disease and Hypertension (AASK) used eGFR trajectories to reflect the change of eGFR during long-term exposure, 13 to further determine the associations between annual change in eGFR and subsequent clinical outcomes. Although some currently emerging reports take note of the eGFR trajectories, they only focus on the risk of all-cause mortality. 14,15 In addition, a meta-analysis suggested that insulin resistance might aggravate the decrease in eGFR, leading to renal impairment in participants with diabetes or prediabetes. 16 Simultaneously, diabetes or prediabetes were associated with increased MI risk. Thus, it is important to estimate the association between eGFR trajectories and risk of MI in the diabetes and prediabetes population.
Therefore, in the present study, we aimed to identify distinct trajectories of eGFR in a diabetes or prediabetes population to have more clarity about the association between eGFR and the risk of MI thereby providing effective prevention strategies for MI.

| Study population
This study consisted of 101 510 participants (81 110 males and 20 400 females, 18-98 years old) at baseline (2006) in the Kailuan study which is a community-based longitudinal cohort study located in Tangshan City, China. 17 The design, methods, rationale, and examination details of the study have been previously described. 8 All participants provided their written informed consent and finished questionnaire interviews, anthropometric measurements, clinical examinations, and laboratory assessments in 2-year cycles until the present day. This investigation was approved jointly by the ethics committee of Kailuan Gen-

| Assessment of eGFR
A serum creatinine sample was collected in the morning after an overnight fast and was analyzed using an automatic biochemical analyzer (Hitachi 747, Hitachi, Tokyo, Japan). eGFR was calculated using the creatinine-based Chronic Kidney Disease Epidemiology Collaboration (CKD-EPI) equation, 18 and at least two measurements during the exposure period (from 2006 to 2012) were used to establish the eGFR trajectory.

| Assessment of outcome
After the baseline examination, participants were followed up for the first occurrence of MI. The incidence of MI was identified by referring to medical records from all 11 Kailuan hospitals and the Municipal Social Insurance Institution from the baseline examination to 31 December 2017. The diagnostic criteria of MI were based on chest pain symptoms, electrocardiogram changes, and cardiac enzyme levels, collectively. 19

| Assessment of covariates
Demographic and clinical characteristics, including age, gender, smoking status, alcohol status, physical activity, history of disease, and medical history were obtained using questionnaires. Smoking and alcohol status were classified as never, former, or current according to self-reported information. Physical activity was classified as inactive (none), moderate active (<80 minutes per week), and very active (≥ 80 minutes of activity per week). Dyslipidemia was defined as any self-reported history. Hypertension was defined as any self-reported hypertension or blood pressure (BP) ≥140/90 mm Hg. Diabetes mellitus was defined as any self-reported diabetes mellitus or FBG ≥7 mmol/L. Anthropometric measurements were measured by trained doctors and nurses, including weight, height, waist circumference (WC), and BP. Body mass index (BMI) was calculated as weight (kg)/height (m) 2 . Systolic blood pressure (SBP) and diastolic blood pressure (DBP) were measured three times with the participants in seated position F I G U R E 2 Flowchart of the study cohort. eGFR, estimated glomerular filtration rate; MI, Myocardial infarction using a mercury sphygmomanometer, and the average of three readings was used in the analyses.
Blood samples were collected in the morning after an overnight fast and were analyzed using an auto-analyzer (Hitachi 747, Tokyo, Japan). FBG and lipid profile, including total cholesterol (TC), triglycerides, low-density lipoprotein cholesterol, and high-density lipoprotein cholesterol, were measured using a standardized method. Using a color scale, the results of the urine strip test were semi-quantified as absent, trace, 1+, 2+, or 3+, and proteinuria was defined as 1+ or higher. All the above items were measured or updated in 2-year cycles (from 2006 to 2012).

| Statistical analyses
In the current study, we used eGFR trajectory groups as the exposure. These models were identified by latent mixture modeling (SAS Proc Traj). 20,21 The model with five patterns was identified to fit the best. We then qualitatively examined the trajectory groups and named each trajectory group based on their visual patterns of change in eGFR levels.
Continuous and categorical variables were described as medians with interquartile ranges and percentages, and Wilcoxon and chi-square tests were used to examine the significance of differences between trajectory groups. Person-years were calculated from the data in 2012 to the date when MI occurred, the date of death, or the date of participating in the last examination in this analysis.
Cox proportional hazards models were used to investigate the association of trajectory groups with the risk of MI. Cox proportional hazard models were built to adjust for different confounding factors: Model 1 was unadjusted. Model 2 was adjusted for age and gender at baseline. Model 3 was further adjusted for current smoker, current alcohol, physical activity, proteinuria, angiotensin-converting enzyme inhibitor (ACEI) medication, angiotensin II receptor blocker (ARB) medication, antidiabetic medication, antihypertension medication, lipid-lowering medication, and BMI, FBG, SBP, and TC at baseline. Because these potential confounders may change during the follow-up, we conducted several sensitivity analyses with adjustment for all potential confounders in 2012 or average BMI, FBG, SBP, and TC during the exposure period (from 2006 to 2012). Considering that insulin resistance is an important index for diabetes and prediabetes, we used WC to replace BMI in model 3 in order to represent the state of insulin resistance more accurately. Besides, we also performed adjusted survival curves of relative hazards from the Cox proportional hazards models. 22 It is unclear whether the eGFR trajectories in special populations might exhibit differential risks of MI. We further explored potential interaction between eGFR trajectories and age (<55 years vs ≥55 years), gender, diabetes status (diabetes vs prediabetes), hypertension status (yes vs no), and proteinuria (yes vs no).
Statistical analyses were conducted by using SAS version 9.4 (SAS Institute, Cary, North Carolina). All reported P values were based on two-sided tests of significance, and P < .05 was deemed statistically significant.

| Characteristics of the trajectory groups
In this prospective cohort of 24 723 participants with diabetes or prediabetes, we identified five eGFR distinct trajectories during a 6-year follow-up exposure period. Each of the trajectory groups was named according to its range and visual pattern of change in eGFR levels. A total of 9.4% of participants had an eGFR consistently <60 mL/ min/1.73m 2 (referred to as low-stable pattern); 31.4% of participants had a stable eGFR of 60 to 90 mL/ min/1.73m 2 (referred to as moderate-stable pattern); 29.5% of participants had an eGFR of 60 to 90 mL/ min/1.73m 2   .03 Abbreviations: ACEI, angiotensin-converting enzyme inhibitor; ARB, angiotensin II receptor blocker; BMI, body mass index; DBP, diastolic blood pressure; eGFR, estimated glomerular filtration rate; FBG, fasting blood glucose; HDL, high-density lipoprotein cholesterol; IQR, interquartile range; LDL, low-density lipoprotein cholesterol; SBP, systolic blood pressure; TC, total cholesterol; TG, triglyceride; WC, waist circumference.  (Figure 3). Characteristics of the five trajectories groups at baseline (2006) were compared and are summarized in Table 1. Compared with the high-stable group, participants in the low-stable groups were more likely to be older and women, have a higher BMI, SBP, and DBP, less drinking, less smoking, and higher concentrations of TC and FBG (Table 1).

| eGFR trajectories associated with incident MI
There was a total of 235 incidents of MI during a mean follow-up of 4.61 ± 0.82 years. Table 2 shows the association between eGFR trajectories and the risk of MI in participants with diabetes or prediabetes. Figure 4 shows that the risk of MI increased significantly in accordance with the eGFR trajectory groups from high-stable to lowstable (P < .01). We observed that eGFR trajectories were significantly associated with future MI risk. After adjustment for potential confounders, compared with participants in the high-stable group, those in the low-stable group (hazard ratio [HR], 4.35; 95% CI,1.95-9.69) experienced the highest risk of MI among all five trajectory groups. Although participants in the moderate-stable group had eGFR levels similar to the participants in the high-decreasing group during the last exposure period (76.3 mL/min/1.73m 2 vs 76.8 mL/min/1.73m 2 ), the risk of MI was much higher in the participants with an eGFR decreasing over time than in those with an eGFR which was stable over time (HR, 2.88; 95% CI, 1.36-6.08 vs HR, 3.57; 95% CI, 1.63-7.85). Although eGFR levels increased to the normal range, participants in the moderateincreasing group still had a significantly increased risk of MI (HR, 2.63; 95% CI, 1.24-5.55) as compared with those in the high-stable group after adjustment for potential confounders.

| Sensitivity analysis
Considering these potential confounder changes from 2006 to 2012, we added sensitivity analyses by adjusting for covariates in 2012 or average BMI, FBG, SBP, and TC during the exposure period (from 2006 to 2012)-and the results remained the same (Table 2). Moreover, we used T A B L E 3 Hazard ratios of trajectories of eGFR for incident myocardial infarction with stratification by baseline characteristics WC to replace BMI in model 3 in order to represent the state of insulin resistance more accurately in sensitivity 3-and the analysis yielded the same pattern of results. Taken together, the results of our study indicated that the trajectories of eGFR were important for the development of MI.

| Subgroup analysis
The associations between eGFR trajectories and incident MI stratified by age, gender, hypertension, diabetes, and proteinuria are shown in Table 3. We also tested the interactions between eGFR trajectories and age, gender, hypertension, diabetes, and proteinuria in relation to MI and subtypes. We did not find significant interactions between eGFR trajectories and age, gender, hypertension, diabetes, and proteinuria in relation to both subtypes of MI (P for interaction >.05 for all).

| DISCUSSION
In this large prospective cohort of people with diabetes and prediabetes, we identified five distinct eGFR trajectory groups, in which participants shared a similar pattern of change in eGFR levels during a 6-year exposure period.
The results show that continuous low levels are independently associated with an about four-fold higher risk of MI in people with diabetes and prediabetes after adjusting for major confounding factors, whereas participants who maintained high levels of eGFR throughout the exposure period had the lowest risk of incident MI. Moreover, participants kept moderate levels of eGFR which had like those in participants from high levels of eGFR to moderate levels at last exposure period, the risk of MI was much higher in eGFR decreasing group. Notably, compared with the lowest risk group, participants who started with moderate levels of eGFR that increased substantially also had an increased risk of incident MI, although their eGFR levels were within the normal range (Table 4).
There is growing awareness that eGFR progression is an important risk factor for end-stage renal disease (ESKD), cardiovascular disease, and all-cause mortality. Previous studies have illustrated large variation in diabetes-dependent decline in eGFR with or without albuminuria. 23 In past studies, a rapid decline in kidney function was reported as an independent risk factor for all-cause mortality. 24 Accordingly, recent studies have shown the relationship between eGFR slopes, which were estimated by multiple measurements of eGFR, and the subsequent risk of ESKD, 25 cardiovascular disease, 12 and all-cause mortality. 24,26,27 In 2019, Megumi Oshima et al reported that eGFR slopes were a prognostic factor for identifying individuals at high risk of cardiovascular disease and all-cause mortality among participants with type 2 diabetes. 10 Similarly, our study identified individuals' eGFR trajectories and found that a rapid decrease in eGFR levels may indicate a higher risk of MI. However, we also found participants with increasing normal eGFR levels cannot restore MI risk.
In this study, we assessed the associations between distinct trajectories of eGFR and the risk of MI with diabetes or prediabetes. The group-based trajectory modeling is a powerful statistical approach to describe the trajectories of eGFR over years. This approach could estimate the average, variability, and the direction of variability simultaneously in one model and allow us to investigate the population heterogeneity in longitudinal changes in eGFR levels, which may provide additional information. For example, among participants with a similar cumulative average of eGFR levels throughout a long period, those with a rapid decrease in eGFR levels may have a higher risk of MI than those who keep stable levels. Moreover, participants with increasing eGFR levels which even reach normal also had a risk of MI. We think these findings are essential because they would identify different trajectories, encompass strategies reducing eGFR to prevent MI, and provide critical implications for intervention for MI. Mechanisms potentially underpinning the association between eGFR and the incidence of MI may be that eGFR reflects an impaired kidney function which has risk factors in common with cardiovascular disease. Furthermore, various mechanisms have been suggested to explain the reason. Decrease in eGFR may indirectly influence cardiovascular risk factors, such as lower levels of BP and lipids. 28 There are also other possible factors including activation of the renin-angiotensin-aldosterone system, endothelial dysfunction, inflammation, and oxidative stress. 29,30 Accordingly these factors may lead to progression of cardiovascular disease. In addition, worsening kidney function may cause decreased appetite and overall frailty, and indirectly result in decreased physical function. 27 The strengths of our study include the prospective design, large cohort, a long follow-up period, repeated measurements of eGFR levels, and use of eGFR changes to estimate the risk of MI. Our study also has several limitations. First, the study design is observational. Although we carefully adjusted for potential risks factors, the possibility of residual confounding remained. Second, this study lacks information about glycosylated hemoglobin and 2-hour glucose, which are important indexes for diagnosis of diabetes and prediabetes, due to the high cost for our large study cohort. Third, the disease duration and records of medications were not included in this study, which may influence the results. Finally, the population in the Kailuan cohort study were mostly male coal miners. The unbalanced distribution of gender may have different demographic characteristics and risk factors, and it is unknown whether our findings could be generalized to other ethnic groups. Further studies are warranted to verify our findings.

| CONCLUSIONS
We found that distinct trajectories of eGFR were associated with MI risk in a diabetic or prediabetic population.
Our findings indicate that change in eGFR levels might play to some extent a reversible role in the early stage of the development of MI and highlight the importance of monitoring longitudinal changes in eGFR levels in the primary prevention of MI. With the development of medicine, therapies targeted at eGFR remission for MI need further investigation.