Both long- and short-term glycemic variability have been associated with incident diabetes complications. We evaluated their relative and potential additive effects on incident renal complications in the Action to Control Cardiovascular Risk in Diabetes trial. A marker of short-term glycemic variability, 1,5-anhydroglucitol (1,5-AG), was measured in 4,000 random 12-month postrandomization plasma samples (when hemoglobin A1c [HbA1c] was stable). Visit-to-visit fasting plasma glucose coefficient of variation (CV-FPG) was determined from 4 months postrandomization until the end point of microalbuminuria or macroalbuminuria. Using Cox proportional hazards models, high CV-FPG and low 1,5-AG were independently associated with microalbuminuria after adjusting for clinical risk factors. However, only the CV-FPG association remained after additional adjustment for average HbA1c. Only CV-FPG was a significant risk factor for macroalbuminuria. This post hoc analysis indicates that long-term rather than short-term glycemic variability better predicts the risk of renal disease in type 2 diabetes.
The relative and potential additive effects of long- and short-term glycemic variability on the development of diabetic complications are unknown.
We aimed to assess the individual and combined relationships of long-term visit-to-visit glycemic variability, measured as the coefficient of variation of fasting plasma glucose, and short-term glucose fluctuation, estimated by the biomarker 1,5-anhydroglucitol, with the development of proteinuria.
Both estimates of glycemic variability were independently associated with microalbuminuria, but only long-term glycemic variability remained significant after adjusting for average hemoglobin A1c.
Our findings suggest that longer-term visit-to-visit glucose variability improves renal disease prediction in type 2 diabetes.
Introduction
Chronic elevations in blood glucose have long been recognized as a main driver of renal disease in patients with both type 1 and type 2 diabetes (T2D). Recent reports from intensive glucose-lowering studies in T2D have shown that long-term glycemic variability may contribute to the risk of microvascular events beyond the overall level of glycemic control (1,2).
Measurement of 1,5-anhydroglucitol (1,5-AG) has been proposed as a valid estimate of short-term glucose variability (3,4). Because glucosuria interferes with 1,5-AG renal reabsorption, lower plasma 1,5-AG may reflect hyperglycemic episodes over recent weeks. A single plasma 1,5-AG level of <6 μg/mL was associated with cardiovascular risk in populations with and without diabetes (5,6) and with microvascular events in patients with diabetes (7).
Long-term glycemic variability assessment requires collecting glucose or hemoglobin A1c (HbA1c) values and calculating various variability metrics over multiple visits. Assessing short-term glycemic variability estimated by a single 1,5-AG measurement may be simpler and easier to interpret and provide similar clinical relevance. Combining short-term with long-term estimates of glycemic variation may improve the prediction of outcomes.
In the current study, we assessed the individual and combined relationships of 1,5-AG and visit-to-visit glucose variability with the development of renal complications in T2D.
Research Design and Methods
Plasma samples and data were from the Action to Control Cardiovascular Risk in Diabetes (ACCORD) trial. ACCORD was a two-by-two factorial, randomized, parallel trial of an intensive (HbA1c <6.0% [42.1 mmol/mol]) versus standard (HbA1c 7.0–7.9% [53–62.8 mmol/mol]) glucose-lowering regimen, including distinct blood pressure and lipid intervention arms (8,9). Within 4 months, the median HbA1c level had fallen from 8.1% (65 mmol/mol) at baseline to 6.5% (47.5 mmol/mol) in the intensive and to 7.5% (58.5 mmol/mol) in the standard group and remained stable thereafter. Fasting plasma glucose (FPG) concentrations were measured every 4 months for a maximum of 84 months. Serum creatinine levels were measured every 4 months for the first year, followed by annual and study closeout measurements (8). Urine albumin and creatinine levels were measured annually and at study closeout. For this analysis, we randomly selected 4,000 participants (equally distributed across two glycemic treatment arms). We measured their 1,5-AG levels at 12 months postrandomization to avoid the substantial glycemic variability resulting from the initiation of glucose-lowering protocols in both treatment groups, to have time to capture a sufficient number of subsequent renal events, and to evaluate the association after a glucose-lowering intervention. Participants with missing outcomes, outcomes before the 12-month collection, and fewer than two glucose measures after the 4-month time point were excluded from the analysis.
Primary and Secondary Outcomes
Primary outcomes were the times (starting at 12-month time point) to the development of 1) microalbuminuria (ratio of urine albumin to creatinine ≥30 mg/g) and 2) macroalbuminuria (ratio of urine albumin to creatinine ≥300 mg/g) (8). The ACCORD study determined both primary outcomes. A secondary composite outcome included the time to the first of at least two consecutive values of estimated glomerular filtration rate <45 mL/min/1.73 m2 by the Chronic Kidney Disease Epidemiology Collaboration equation (10) or renal failure (initiation of dialysis, end-stage renal disease, renal transplantation, or serum creatinine >291.72 μmol/L in the absence of an acute reversible cause) classified by ACCORD investigators. Patients randomly assigned to active lipid-lowering treatment were excluded from this analysis because of the impact of fenofibrate on plasma creatinine (and thus estimated glomerular filtration rate) levels (11).
Glucose Variability
A single 12-month serum sample was used to measure 1,5-AG concentrations. We measured them in duplicate (intra-assay coefficient of variation [CV] 1.6%) and in eight batches (interassay CV 3.2%) using an enzymatic assay (Diazyme Laboratories, Poway, CA). Long-term visit-to-visit glucose variability was quantified by the CV of FPG (CV-FPG) using FPG measured longitudinally from 4 months postrandomization, when the initial separation between glucose treatment arms occurred (and FPG levels remained stable thereafter) (1,2).
Statistical Analysis
Differences between participants who did and did not develop an event were assessed by the Wilcoxon test for continuous variables and the χ2 or Fisher exact test, as appropriate, for categorical variables. Cox proportional hazards models were used to test the association of study outcomes with CV-FPG and 1,5-AG. CV-FPG was included as a continuous and time-dependent covariate (12). Based on a previous report (6) and confirmed here by penalized linear splines models, a 1,5-AG level of ≤6 μg/mL was categorized as low. The proportionality of all model predictors was confirmed by goodness-of-fit tests using Schoenfeld residuals (13). CV-FPG was scaled and log transformed. Analyses were sequentially adjusted for variables (models 1–4; see table legend for details). Models 3 and 4 included CV-FPG and 1,5-AG in the same model to test for their independent effects. Models including 1,5-AG were adjusted for a 1,5-AG assay batch effect.
If CV-FPG or 1,5-AG was significant in model 4, we tested for differential risks between the glucose treatment arms for microalbuminuria and the secondary composite outcome (macroalbuminuria events were too infrequent). Interaction P values were calculated between CV-FPG or 1,5-AG and the treatment arms in models 1 and 4. To evaluate the additive risk of the two glycemic variability measures, we compared the risk of developing microalbuminuria and the secondary composite outcome in those with high CV-FPG (defined by the median) and low 1,5-AG versus those with low CV-FPG and high 1,5-AG.
All statistical analyses were performed using R (version 4.2.2; https://www.r-project.org). A two-sided P value <0.05 was considered statistically significant.
Data and Resource Availability
The ACCORD database is available upon request from the NHLBI Specimen and Data Repository (https://biolincc.nhlbi.nih.gov/studies/accord/). The data supporting the findings of this study are available from the corresponding author upon reasonable request.
Results
Participants
Among 3,998 participants (two participants with missing identification codes were excluded from analysis), median (interquartile range) values of 1,5-AG for the intensive and standard treatment arms were 11.6 (7.6–16.4) and 7.4 (4.0–12.1) μg/mL, respectively. Microalbuminuria developed in 534 of 2,631 (median follow-up 60 months), macroalbuminuria in 189 of 3,589 (median follow-up 63 months), and the secondary composite outcome in 248 of 3,094 (median follow-up 61 months) qualifying participants. Baseline characteristics stratified by the events are listed in Table 1. Multiple risk factors for renal disease were present for each outcome.
Glucose Variability and Risk of Renal Outcomes
Individually, both CV-FPG and low 1,5-AG were associated with microalbuminuria in model 1 and remained significantly associated after adjusting for other risks factors (model 2; CV-FPG hazard ratio 1.16 [95% CI 1.07, 1.26]; P < 0.001; low 1,5-AG 1.29 [1.06, 1.56]; P = 0.011) (Table 2). Including both CV-FPG and low 1,5-AG in the model (model 3) yielded similar hazard ratios and CIs, indicating an independent association of CV-FPG and low 1,5-AG with the development of microalbuminuria. After additionally adjusting for average HbA1c (model 4), the association of low 1,5-AG with microalbuminuria was attenuated. CV-FPG, but not 1,5-AG, was significantly associated with macroalbuminuria in models 1 to 3 and remained so in model 4 (1.22 [1.03, 1.44]; P = 0.022). For the secondary composite outcome, CV-FPG, but not low 1,5-AG, was a significant risk factor after adjusting for age (model 1), but not after further adjusting for other covariates.
Stratified analyses were performed for CV-FPG, because only CV-FPG remained significant in model 4 (Table 3). The association between CV-FPG and microalbuminuria was significant in the standard glucose control group after full adjustment (model 4; interaction P = 0.139). Again, CV-FPG showed a stronger association with the secondary composite outcome in the standard glucose control group than in the intensive glucose control group (interaction P < 0.001). Participants within both high CV-FPG and low 1,5-AG categories did not show a significantly greater risk of developing microalbuminuria or the secondary composite outcome than those with other combinations of these variability metrics (results not shown).
Discussion
In this study, both CV-FPG and low 1,5-AG were independently associated with the development of microalbuminuria after adjusting for multiple risk factors. However, only the association of CV-FPG remained after adjusting for average glycemic control. Higher CV-FPG, but not low 1,5-AG, was a significant risk factor for macroalbuminuria. Higher CV-FPG was more harmful in those receiving less intensive glucose control, consistent with our previous findings (1,2). Combining both high CV-FPG and low 1,5-AG did not increase the risk of either primary or secondary outcomes. These results confirm the potential value of estimates of glycemic variability for determining the risk of renal complications. However, it does not seem that a single measure of 1,5-AG is sufficient to capture this risk, and determination of longer-term glucose variation is still needed. The trend of CV-FPG being associated with greater risk in participants with less strict diabetes control suggests that the combination of both higher HbA1c and glycemic variability may be more closely linked with the development of complications.
The limitations of our study include 1) the potentially lower accuracy and statistical power of a single 1,5-AG measurement at 12 months versus repeated 1,5-AG measurements or continuous glucose monitoring for recording short-term glycemic variability and 2) the post hoc nature of our analysis.
In summary, although long-term glycemic variability and short-term glycemic variability measured by 1,5-AG demonstrated the potential to predict renal complications in individuals with T2D, longer-term visit-to-visit glucose variability seems to provide superior prediction of renal disease compared with 1,5-AG.
Article Information
Acknowledgments. The authors thank the BioLincc repository for providing the samples.
Funding. This work was supported by the Veterans Affairs Cooperative Studies Program, Department of Veterans Affairs Office of Research and Development. Additional support was received from the National Heart, Lung, and Blood Institute (NHLBI), National Institutes of Health (NIH; R01-067690 and 5R01-094775), and the American Diabetes Association to P.R. J.J.Z. was supported by the National Institute of Diabetes and Digestive and Kidney Diseases (K01DK106116), the NHLBI (R21HL150374), and the National Human Genome Research Institute (R01HG006139), NIH; the National Science Foundation (NSF; DMS-2054253 and IIS-2205441); and the Arizona Biomedical Research Centre New Investigator Award. A.V. and E.Z. were supported by the NSF Mathematical and Physical Sciences–High Program.
Duality of Interest. No potential conflicts of interest relevant to this article were reported.
Author Contributions. T.O. analyzed data and wrote the first draft of the manuscript. A.V. and E.Z. analyzed data. H.Z. provided advice and edited the manuscript. J.K. and P.R. contributed to discussion and reviewed/edited the manuscript. J.J.Z. supervised the project, designed experiments, and wrote the manuscript. All authors reviewed, edited, and approved the manuscript. J.J.Z. is the guarantor of this work and, as such, had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.
