Abstract
Aims
The efficacy of intensive blood pressure (BP) control remains controversial, and the variability of HbA1c was a risk factor for macrovascular events in patients with type 2 diabetes. We investigated whether the HbA1c variability modifies the efficacy of intensive BP control.
Methods
Data from the Action to Control Cardiovascular Risk in Diabetes Blood Pressure (ACCORD‐BP) trial was utilized. K‐means clustering was used to cluster patients into three groups based on the HbA1c variability score and baseline HbA1c values. Cox proportional hazard models and generalized linear models were used to measure the subgroup differences in intensive BP control treatment effects. The primary outcome was a composite of nonfatal myocardial infarction (MI), stroke, or death from cardiovascular causes.
Results
In patients with low HbA1c variability rather than medium or high HbA1c variability, intensive BP control reduced the risk of the primary outcome on a relative scale (HR 0.60, 95%CI 0.40–0.90, p interaction was 0.03), non‐fatal MI (HR 0.61, 95% CI 0.37–1.00, p interaction was 0.04) and stroke (HR 0.19, 95%CI 0.05–0.64, p interaction was 0.02) or absolute scale. Regardless of the variability group, intensive BP control did not reduce the risk of cardiovascular or all‐cause mortality (p interaction >0.05) both on relative and absolute risk scales.
Conclusion
HbA1c variability had effect on the efficacy of intensive BP control and intensive BP control brought a significant macrovascular benefit in patients with type 2 diabetes and low HbA1c variability.
Keywords: HbA1c variability, intensive blood pressure control, machine learning algorithms, macrovascular outcomes, stroke
1. INTRODUCTION
Diabetes is associated with several cardiovascular risk factors, with high blood pressure (BP) being one of the most common in patients with diabetes. 1 Lowering BP is a well‐established strategy for preventing microvascular and macrovascular events in patients with diabetes. 2 The United Kingdom Prospective Diabetes Study and the Action in Diabetes and Vascular Disease Controlled Evaluation trial have shown a significant reduction in microvascular or macrovascular events when mean BP was decreased in patients with type 2 diabetes. 3 , 4
However, in the Action to Control Cardiovascular Risk in Diabetes Blood Pressure (ACCORD‐BP) study, intensive BP control (targeting systolic BP ≤120 mmHg) did not reduce the composite cardiovascular outcome or all‐cause mortality compared to standard BP control (targeting systolic BP ≤140 mmHg) in patients with type 2 diabetes. 5 In contrast, the Systolic Blood Pressure Intervention Trial (SPRINT), which employed a similar study design to the ACCORD study, excluded patients with type 2 diabetes and demonstrated that the risk of adverse cardiovascular events decreased significantly with intensive BP control. 6 One post hoc analysis of ACCORD‐BP found that the intensive glycaemic intervention confounds the interpretation of the BP trial. 7 These studies illustrated that glycaemic control in patients with type 2 diabetes may have an impact on the efficacy of blood pressure‐lowering therapy.
Currently, the core of common blood glucose management strategies for type 2 diabetes is achieving ideal haemoglobin A1c (HbA1c) levels. However, simply achieving an ideal HbA1c level does not fully explain the occurrence of macrovascular events in patients with type 2 diabetes. The level and variability of HbA1c were considered risk factors for macrovascular events in patients with type 2 diabetes. 8 , 9 , 10 , 11 The objective of our study was to investigate whether HbA1c variability modifies the efficacy of intensive BP control.
2. METHODS
2.1. Study participants and data collection
The ACCORD study is a randomized, multicenter, double 2 × 2 factorial clinical trial designed to test the effects of intensive glycaemic control, intensive lipid control and intensive BP control on CVD risk in middle‐aged and older patients with type 2 diabetes and either established CVD or additional CVD risk factors. 12 The design and results of the ACCORD study have been previously published. 13 The ACCORD‐BP trial enrolled 4733 high CVD risk participants with type 2 diabetes aged >40 years with a systolic blood pressure (SBP) between 130 and 180 mmHg. 5 After the first year of treatment, the average SBP in the intensive BP control group was 119.3 and 133.5 mmHg in the standard BP control group, resulting in an average between‐group difference of 14.2 mmHg.
2.2. Exposure variables and study outcomes
Our study measured HbA1c variability using the HbA1c variability score (HVS) and baseline HbA1c levels. The HVS was calculated as the percentage of the number of changes in HbA1c >0.5% (5.5 mmol/mol) among all HbA1c measurements for an individual. 11 The primary outcome was major cardiovascular adverse events (MACEs), defined as a composite outcome of nonfatal myocardial infarction, nonfatal stroke, or death from cardiovascular causes. The secondary outcome was all‐cause mortality. All outcomes were adjusted.
2.3. Statistical analysis
Baseline characteristics and crude outcomes of participants were presented as frequencies and percentages for categorical variables and as means and SD or medians and interquartile ranges for continuous variables, depending on their distribution type. The Chi‐square or Mann–Whitney U test was used to compare the baseline characteristics according to the variable's distribution type. The K‐means method was used to cluster the HVS and baseline HbA1c of participants into three groups: low, medium and high HbA1c variability levels. K‐means clustering is a method of vector quantization, originally from signal processing, that aims to partition “n” observations into ‘k’ clusters in which each observation belongs to the cluster with the nearest mean (cluster centres or cluster centroid), serving as a prototype of the cluster. We used the Silhouette, Calinski–Harabasz and Davies–Bouldin indices to describe the data dispersion within and across the clusters and to compute the ratio of similarity of the original data from the clusters. We calculated the internal validation metrics separately when dividing the total population into k (k = 2–10) groups, and after combining the two metrics (Davies‐Bouldin Index and Silhouette Indexes), we concluded that clustering performs best when dividing the population into three groups (Table S1).
We used three models to conduct the various analyses: model 1: unadjusted; model 2: adjusted for age, intensive blood glucose treatment arm, sex, CVD history and ethnicity; model 3: adjusted for age, intensive blood glucose treatment arm, sex, CVD history, ethnicity, body mass index, BP, heart rate, hyperlipidaemia, estimated glomerular filtration rate, comorbidity (heart failure, depression and albuminuria) and smoking status. We calculated outcome metrics on both relative and absolute scales. Graphical methods via scaled Schoenfeld residuals were used to examine the proportional hazard assumption; all models met the proportional hazard assumption. We used a Cox proportional hazards regression model to estimate the hazard ratios (HRs) of the primary and secondary outcomes. Adjusted absolute risk reduction of intensive BP control at 5 years, stratified by HbA1c variability, were obtained from generalized linear models (model 3).
We performed a variety of sensitivity analyses. First, we included the patient's time‐weighted mean value of HbA1c or the number of HbA1c measurements for each patient in our model to further clarify the effect of HbA1c variability. Moreover, to ascertain whether the differences in variability were attributable to variations in patient adherence, we introduced the SBP variability score (SBPVS). The algorithm for SBPVS is analogous to that of HVS, where a change greater than 10 mmHg in SBP from one follow‐up to the next is marked. The percentage of these marked instances over the total number of follow‐ups is denoted as SBPVS. We analysed using SBPVS to explore its correlation with HVS and whether its impact on intensive blood pressure treatment was analogous to that of HVS. If SBPVS does not manifest effects similar to HbA1c variability, it can indicate, to some extent, that HbA1c variability is not solely determined by adherence. To further ascertain the robustness of our findings, we conducted a sensitivity analysis using variation independent of the mean (VIM) as a complement to the HVS. 14 , 15 , 16 , 17 The calculation formula for VIM is as follows:
(x is the regression coefficient based on natural logarithm of standard deviation on natural logarithm of mean).
All analyses were performed using the R statistical software packages and Python. p values <0.05 (two‐sided) were considered statistically significant.
3. RESULTS
3.1. Baseline characteristics
Out of the 4733 participants from the ACCORD‐BP trial, 103 participants whose HVS could not be calculated were excluded. Therefore, our post hoc analysis of the ACCORD‐BP trial included 4630 participants, 2311 of which were randomized into the intensive BP control group and 2319 into the standard BP control group. The mean age of the participants was 62.7 years (SD 6.7), and 52.5% were men. There were no significant differences in baseline characteristics between participants in the standard and intensive BP control arms across the three different HbA1c variability groups. After a median follow‐up duration of 4.94 years, the primary outcome occurred in 428 participants (9.24%). The baseline characteristics and crude outcomes are presented in Table 1.
TABLE 1.
Baseline characteristics and crude endpoints of the study participants.
| Variability group | Low | Medium | High | Overall participants | ||||
|---|---|---|---|---|---|---|---|---|
| Standard arm | Intensive arm | Standard arm | Intensive arm | Standard arm | Intensive arm | Standard arm | Intensive arm | |
| N | 943 | 917 | 982 | 984 | 394 | 410 | 2319 | 2311 |
| Age | 63.69 ± 6.62 | 63.23 ± 6.46 | 62.44 ± 6.79 | 62.45 ± 6.45 | 61.18 ± 6.56 | 62.06 ± 7.04 | 62.73 ± 6.74 | 62.69 ± 6.57 |
| Female | 439 (46.55%) | 441 (48.09%) | 467 (47.56%) | 469 (47.66%) | 192 (48.73%) | 191 (46.59%) | 1098 (47.35%) | 1101 (47.64%) |
| Race or ethnic group (%) | ||||||||
| White | 582 (61.72%) | 593 (64.67%) | 567 (57.74%) | 591 (60.06%) | 192 (48.73%) | 200 (48.78%) | 1341 (57.83%) | 1384 (59.89%) |
| Black | 190 (20.15%) | 169 (18.43%) | 246 (25.05%) | 222 (22.56%) | 127 (32.23%) | 141 (34.39%) | 563 (24.28%) | 532 (23.02%) |
| Hispanic | 56 (5.94%) | 53 (5.78%) | 66 (6.72%) | 62 (6.30%) | 42 (10.66%) | 40 (9.76%) | 164 (7.07%) | 155 (6.71%) |
| Other | 115 (12.20%) | 102 (11.12%) | 103 (10.49%) | 109 (11.08%) | 33 (8.38%) | 29 (7.07%) | 251 (10.82%) | 240 (10.39%) |
| CVD History (%) | 277 (29.37%) | 276 (30.10%) | 347 (35.34%) | 335 (34.04%) | 144 (36.55%) | 166 (40.49%) | 768 (33.12%) | 777 (33.62%) |
| Education (%) | ||||||||
| Less than high school | 130 (13.79%) | 132 (14.41%) | 153 (15.60%) | 160 (16.26%) | 69 (17.51%) | 95 (23.23%) | 352 (15.19%) | 387 (16.76%) |
| High school graduate | 266 (28.21%) | 246 (26.86%) | 287 (29.26%) | 254 (25.81%) | 99 (25.13%) | 98 (23.96%) | 652 (28.13%) | 598 (25.90%) |
| Some college | 297 (31.50%) | 309 (33.73%) | 309 (31.50%) | 328 (33.33%) | 131 (33.25%) | 128 (31.30%) | 737 (31.79%) | 765 (33.13%) |
| College degree or higher | 250 (26.51%) | 229 (25.00%) | 232 (23.65%) | 242 (24.59%) | 95 (24.11%) | 88 (21.52%) | 577 (24.89%) | 559 (24.21%) |
| Blood pressure (mm Hg) | ||||||||
| Systolic | 139.09 ± 14.89 | 137.99 ± 15.66 | 138.74 ± 15.53 | 139.66 ± 16.07 | 141.23 ± 16.78 | 139.57 ± 16.64 | 139.30 ± 15.51 | 138.98 ± 16.02 |
| Diastolic | 75.51 ± 9.80 | 75.28 ± 10.09 | 75.80 ± 10.30 | 76.33 ± 11.38 | 77.19 ± 10.93 | 76.55 ± 11.23 | 75.92 ± 10.23 | 75.95 ± 10.53 |
| Mean systolic | 132.93 ± 8.31 | 119.67 ± 8.30 a | 133.83 ± 8.43 | 120.94 ± 8.36 a | 135.32 ± 10.00 | 123.51 ± 10.41 a | 133.72 ± 8.70 | 120.89 ± 8.83 a |
| Mean diastolic | 70.69 ± 7.48 | 65.13 ± 6.97 a | 71.55 ± 7.66 | 65.83 ± 6.93 a | 73.13 ± 7.96 | 67.60 ± 8.03 a | 71.47 ± 7.69 | 65.89 ± 7.20 a |
| HbA1c measures | ||||||||
| Baseline HbA1C (%) | 7.93 ± 0.88 | 8.01 ± 0.92 | 8.43 ± 1.06 | 8.46 ± 1.07 | 8.90 ± 1.22 | 8.89 ± 1.21 | 8.31 ± 1.08 | 8.36 ± 1.09 |
| Mean of HbA1C (%) | 6.87 ± 0.63 | 6.59 ± 0.68 | 7.47 ± 0.69 | 7.51 ± 0.72 | 8.22 ± 0.93 | 8.11 ± 0.88 | 7.36 ± 0.86 | 7.37 ± 0.86 |
| HVS | 17.77 ± 8.25 | 17.80 ± 8.14 | 43.39 ± 8.21 | 43.46 ± 8.40 | 75.78 ± 13.41 | 75.71 ± 13.43 | 38.48 ± 22.54 | 39.00 ± 22.68 |
| Intensive blood glucose management | 609 (64.58%) | 585 (63.79%) | 422 (42.97%) | 409 (41.57%) | 141 (35.79%) | 153 (37.32%) | 1172 (50.54%) | 1147 (49.63%) |
| Heart rate (BPM) | 72.11 ± 11.44 | 72.36 ± 11.61 | 73.42 ± 11.48 | 73.09 ± 11.38 | 74.56 ± 11.56 | 74.40 ± 12.73 | 73.08 ± 11.51 | 73.03 ± 11.74 |
| Body mass index | 31.62 ± 5.22 | 31.76 ± 5.43 | 32.38 ± 5.38 | 32.43 ± 5.69 | 32.37 ± 5.57 | 32.62 ± 5.64 | 32.07 ± 5.36 | 32.20 ± 5.59 |
| Cholesterol (mg/dL) | ||||||||
| Total | 188.42 ± 40.75 | 190.06 ± 41.98 | 193.06 ± 45.43 | 197.71 ± 46.70 | 193.71 ± 48.42 | 194.60 ± 47.12 | 191.29 ± 44.18 | 194.12 ± 45.08 |
| Low‐density lipoprotein | 107.22 ± 34.53 | 108.35 ± 35.37 | 109.46 ± 36.42 | 112.99 ± 37.37 | 110.03 ± 37.13 | 112.44 ± 40.71 | 108.65 ± 35.80 | 111.05 ± 37.26 |
| High‐density lipoprotein | 46.30 ± 13.41 | 46.83 ± 13.56 | 46.51 ± 14.76 | 45.56 ± 12.44 | 46.58 ± 14.34 | 45.81 ± 14.13 | 46.44 ± 14.15 | 46.11 ± 13.21 |
| Triglyceride (mg/dL) | 185.10 ± 148.75 | 181.94 ± 147.97 | 193.53 ± 183.10 | 206.88 ± 196.56 | 200.29 ± 257.99 | 195.93 ± 191.07 | 191.26 ± 185.77 | 195.04 ± 178.09 |
| Fasting serum glucose (mg/dL) | 164.92 ± 46.00 | 167.98 ± 49.66 | 175.13 ± 58.88 | 177.49 ± 58.57 | 188.89 ± 74.48 | 190.33 ± 68.57 | 173.33 ± 57.79 | 175.99 ± 57.73 |
| Estimated GFR, mL min−1 1.73 m−2 | 91.83 ± 26.02 | 91.10 ± 25.85 | 91.73 ± 28.59 | 92.14 ± 35.81 | 91.60 ± 26.70 | 90.81 ± 25.30 | 91.75 ± 27.24 | 91.49 ± 30.40 |
| History physical exam (%) | ||||||||
| Protein in urine | 170 (18.05%) | 166 (18.10%) | 193 (19.65%) | 186 (18.90%) | 93 (23.60%) | 88 (21.46%) | 456 (19.67%) | 440 (19.04%) |
| Heart failure | 31 (3.29%) | 36 (3.93%) | 40 (4.07%) | 45 (4.57%) | 21 (5.33%) | 24 (5.85%) | 92 (3.97%) | 105 (4.54%) |
| Neuropathy | 221 (23.46%) | 197 (21.48%) | 278 (28.31%) | 250 (25.41%) | 118 (29.95%) | 122 (29.76%) | 617 (26.62%) | 569 (24.62%) |
| Depression | 196 (20.81%) | 170 (18.54%) | 260 (26.48%) | 249 (25.30%) | 116 (29.44%) | 125 (30.49%) | 572 (24.68%) | 544 (23.54%) |
| Smoked cigarettes in last 30 days | 104 (11.03%) | 110 (12.00%) | 142 (14.46%) | 130 (13.21%) | 59 (14.97%) | 65 (15.85%) | 305 (13.15%) | 305 (13.20%) |
| Crude outcomes (%) | ||||||||
| Primary outcome | 65 (6.89%) | 38 (4.14%) a | 104 (10.59%) | 110 (11.18%) | 58 (14.72%) | 53 (12.93%) | 227 (9.79%) | 201 (8.7%) |
| All‐cause mortality | 27 (2.86%) | 31 (3.18%) | 61 (6.21%) | 47 (4.78%) | 42 (10.66%) | 59 (14.39%) | 130 (5.61%) | 137 (5.93%) |
| CVD‐mortality | 8 (0.85%) | 9 (0.98%) | 26 (2.65%) | 19 (1.93%) | 17 (4.31%) | 26 (6.34%) | 51 (2.20%) | 54 (2.34%) |
| Non‐fatal MI | 44 (4.67%) | 26 (2.84%) a | 70 (7.13%) | 75 (7.62%) | 31 (7.87%) | 24 (5.85%) | 145 (6.25%) | 125 (5.41%) |
| Total stroke | 19 (2.01%) | 3 (0.33%) a | 21 (2.14%) | 23 (2.34%) | 19 (4.82%) | 10 (2.44%) | 59 (2.54%) | 36 (1.56%) a |
Note: Plus–minus values are means ± SD. Race and ethnic group were self‐reported. The body mass index is the weight in kilograms divided by the square of the height in metres.
Abbreviations: CVD, cardiovascular diseases; GFR, glomerular filtration rate; HbA1c, haemoglobin A1c; HVS, HbA1c variability score; MI, myocardial infarction.
Statistically significant compared to the standard management approach, p < 0.05.
3.2. Clustering results and BP difference in each variability group
Compared with the quantitative partitioning method, the high variability group in the machine learning subgroup had fewer participants (17.4%) (Figure 1). The cut‐off values for the different grouping methods were presented at Table S2. Among the three evaluation indicators for clustering methods, K‐means had the best performance (Table S3). The difference in SBP between the intensive BP control group and standard BP control group was 13.26 , 12.89 and 11.81 mm Hg in low, medium and high variability groups, respectively. The differences in SBP at every visit under different treatment allocations in each variability group were presented in Figure S1. Mean blood pressure during follow‐up stratified according to variability is shown in Figure S2.
FIGURE 1.
Clustering results for K‐means algorithm and quantitative partitioning method. HbA1c, haemoglobin A1c; HVS, HbA1c variability score.
3.3. HbA1c variability and intensive BP control
Similar to other findings reported in the literature, the risk of macrovascular events increased significantly with an increase in HbA1c variability. 11 , 18 The risk of the primary outcome in the high HbA1c variability group was nearly three times higher (HR 3.06, 95% CI 2.30–4.07) compared to the low HbA1c variability group (Table S4).
There was a significant interaction between HbA1c variability and intensive BP control. In patients with low HbA1c variability, intensive BP control significantly reduced the risk of a primary outcome on both relative (HR 0.60, 95% CI 0.40–0.90, P for interaction was 0.03) and absolute scales (ARR 2.57, 95% CI 0.34–4.80, p for interaction was 0.05). In contrast, intensive BP control did not reduce the risk of outcomes in patients with medium or high HbA1c variability (Figures 2, 3 and 4). We found intensive BP control primarily focused on reducing the risk of non‐fatal MI (HR: 0.61, 95% CI 0.37–1.00, p for interaction was 0.04) and stroke (HR 0.19, 95% CI 0.05–0.64, p for interaction was 0.02) (Figures 2, 3 and 4).
FIGURE 2.
Time‐to‐event analysis for participants stratified by HbA1c variability levels. The time to event between intensive and standard blood pressure control in participants by using Cox proportional hazards regression. The primary outcome was a composite outcome of nonfatal myocardial infarction, stroke or death from cardiovascular causes.
FIGURE 3.
Outcomes in participants with intensive or standard blood pressure control arm, stratified by HbA1c variability. *Per 100 person‐year, model 3 was used, adjusted for age, intensive blood glucose treatment arm, sex, CVD history, ethnicity, body mass index, BP, heart rate, hyperlipidaemia, estimated glomerular filtration rate, comorbidity (heart failure, depression and albuminuria) and smoking status.
FIGURE 4.
Adjusted absolute risk difference of intensive versus standard blood pressure control on the study outcomes in participants stratified by HbA1c variability. MI, myocardial infarction. Model 3 was used, adjusted for age, intensive blood glucose treatment arm, sex, CVD history, ethnicity, body mass index, BP, heart rate, hyperlipidaemia, estimated glomerular filtration rate, comorbidity (heart failure, depression and albuminuria) and smoking status.
For all‐cause mortality, there was no significant interaction between HbA1c variability and intensive BP treatment and intensive BP treatment did not have a significant effect regardless of HbA1c variability, both on relative and absolute risk scales. We did not find a significant interaction between HVS, baseline HbA1c or VIM alone and intensive BP management after dividing the patients into three groups using the quantile partitioning method (Table S5).
We then conducted subgroup analyses within different glycaemic intervention groups. We found that the impact of HbA1c variability on the effectiveness of intensive BP control was primarily concentrated in the standard glycaemic control group. In the low HbA1c variability subgroup, intensive blood pressure control significantly reduced the risk of major outcomes (HR 0.38, 95% CI 0.21–0.71, p for interaction was <0.01). However, there was no significant difference in all‐cause mortality between the different glycaemic intervention groups (Table S6).
3.4. Sensitivity analysis
Our results remained robust after adjusting for the time‐weighted mean HbA1c values of the patients or the number of HbA1c measurements for each patient (Figures S3 and S4). To further validate the robustness of our results, we employed VIM as a complement to the HVS. VIM was utilized to assess the variability of HbA1c in another secondary analysis of ACCORD study. 14 After clustering HVS and VIM using the K‐means method, the findings, ranging from group sizes to the impact on risk of adverse outcomes, closely aligned with our main results (Tables S5 and S7). Then, we utilized various statistical methods to assess the correlation between HVS and SBPVS. Although there was a positive correlation between them, both linear and non‐linear associations were weak (Figure S5). Moreover, there was no significant interaction between the SBPVS grouping and intensive BP control (Table S8).
4. DISCUSSION
Our study revealed that HbA1c variability had a significant impact on the efficacy of intensive BP control in patients with type 2 diabetes. Intensive BP control reduced the risk of macrovascular events, including MI and stroke, in patients with low HbA1c variability, but not in those with medium or high HbA1c variability.
Based on the non‐ideal results of ACCORD‐BP, the latest American Diabetes Association (ADA) guidelines recommend a BP target of <140/90 mmHg or <130/90 mmHg for most patients with type 2 diabetes. 19 ADA advocates risk stratification to avoid overtreatment of patients with comorbidities and to reduce the potential for adverse drug events. 20 Interestingly, the ACCORD‐BP participants who met the eligibility criteria for SPRINT showed the potential benefit of intensive BP control, 21 which indicated that intensive BP control may interact with glycaemic management in patients with type 2 diabetes. The Action to Control Cardiovascular Risk in Diabetes Follow‐On (ACCORDION) trial continued to follow 3957 ACCORD participants for 54 to 60 months, during which the interaction between intensive BP control and glucose‐lowering therapy became more significant. Patients in the standard glucose‐lowering therapy group benefited from intensive BP control. 22 Our findings reveal that HbA1c variability modulates the efficacy of intensive blood pressure management primarily in the standard glycaemic treatment group. This aligns with previous studies suggesting cardiovascular benefits of intensive blood pressure control in patients under standard glycaemic treatment. 23 , 24 Notably, within this group, patients with low HbA1c variability demonstrated the most pronounced benefits from intensive blood pressure management. Furthermore, our data suggested that across different HbA1c variability groups, there were no significant differences in either the number of patients or baseline characteristics between the intensive and standard BP control groups. This indicated that blood pressure control strategies did not have impact on long‐term glycaemic control. Early identification of patients who will benefit from intensive BP control and the development of individualized therapy strategies may provide novel insights into managing patients with type 2 diabetes.
High HVS was associated with adverse events. 25 A retrospective cohort study reported that a high HVS had a increased risk of adverse macrovascular events in patients with type 2 diabetes. 26 Some previous studies have used the standard deviation or coefficient of variation values of multiple HbA1c measurements to assess long‐term glycaemic variability, 27 , 28 which are statistically indicative of discrete sample trends but are difficult to calculate and interpret in clinical practice. The HVS was much easier to interpret clinically. Our study incorporates both the HVS and baseline HbA1c values through a machine learning algorithm, which provided a comprehensive and systematic representation of glycaemic control status by combining multiple indicators. Meanwhile, the K‐means method used in this study outperformed traditional clustering methods. Our previous study found that intensive blood glucose lowering significantly reduced the risk of MACEs in patients with type 2 diabetes and low HbA1c variability. 17
Our study showed that intensive BP control reduced the risk of MACEs by 40% for participants with low HbA1c variability. This suggested that participants with more stable glycaemic control would benefit from intensive BP control. Moreover, we found that in the low variability group, intensive BP control primarily reduced the risk of non‐fatal MI and stroke. This aligned with the results of two previous meta‐analyses, where a reduction in SBP consistently led to a decreased risk of MI and stroke. 29 , 30 However, in the SPRINT study, intensive BP control did not reduce the risk of MI and stroke, and its reduction in the risk of adverse events in patients was mainly due to its reduction in the risk of all‐cause and CVD death. 6 In our study, intensive BP control reduced the risk of MI as well as stroke and did not reduce the risk of CVD or all‐cause mortality in the low variability group. This may be due to the accelerated rate of vascular aging in patients with type 2 diabetes, as evidenced by poor vascular compliance, increased blood pressure variability and impaired blood flow autoregulation. 2 , 31 A meta‐analysis showed that blood pressure control strategies based on predicted cardiovascular risk significantly reduced the CVD events compared with SBP‐based strategies, whereas subgroup analyses indicated that risk‐based blood pressure control strategies in patients with diabetes did not lead to ideal results.32 Patients with type 2 diabetes belong to a high‐risk group, and our study refined the classification of patients with diabetes, distinguishing between high HbA1c variability and low HbA1c variability.
Variability also served as a measure of adherence. Participants' adherence was not taken into account in this study, but our data suggested that across different variability groups, there were no significant differences in either the number of patients or baseline characteristics between the intensive and standard BP control groups. This demonstrates that there is no selection bias between the intensive and standard BP control groups across different variability groups. Moreover, we introduced SBPVS as a means to indirectly analyse patient adherence. If variability was a reflection of patient adherence, then not only would it impact the variability of HbA1c, but it would also influence the variability of other indicators. Therefore, we selected blood pressure variability, which is most relevant to our study. Additionally, we calculated SBPVS using an algorithm modelled after HVS to ensure the comparability of the two metrics. Our results indicate that although there is a positive correlation between HVS and SBPVS, the association between them is weak. Furthermore, SBPVS did not show a similar impact on intensive BP control as the HbA1c variability did. These findings suggest that patient adherence was not a major determining factor for variability.
This study has several limitations. As a post hoc analysis, our findings should be considered hypothesis‐generating rather than conclusive. The study population, while substantial, may limit result generalizability to all type 2 diabetes patients. Eliminating all potential confounding factors, particularly patient adherence, proved challenging. Subgroup analyses may be underpowered, and negative findings should be interpreted cautiously.
Importantly, HbA1c variability was calculated using follow‐up period values. While randomization can balance baseline variables, post‐randomization variables may not be similarly balanced, potentially introducing bias in interpreting the impact of HbA1c variability on treatment outcomes.
Future research should validate these results in prospectively designed studies with larger, more diverse cohorts and longer follow‐up periods. These studies should further explore the relationship between HbA1c variability and blood pressure control effects, focusing on long‐term outcomes of intensive BP control in patients with varying HbA1c variability levels. Additionally, methods to account for potential imbalances in post‐randomization variables should be considered when assessing the relationship between HbA1c variability and treatment efficacy.
Our study highlighted the importance of considering HbA1c variability in BP control strategies for patients with type 2 diabetes. Identifying patients with low HbA1c variability and subsequently implementing intensive BP control might be an effective strategy for reducing the risk of macrovascular events in patients with type 2 diabetes.
AUTHOR CONTRIBUTIONS
X. W., X. H. and J. P. designed the study and provided methodological expertise. X. W. and J. P. drafted the manuscript, while K. Z. and M. L. assisted with drafting tables and figures and performing statistical analyses. All authors have reviewed and approved the final manuscript.
CONFLICT OF INTEREST STATEMENT
The authors declare that they have no conflicts of interest.
PEER REVIEW
The peer review history for this article is available at https://www.webofscience.com/api/gateway/wos/peer‐review/10.1111/dom.16112.
Supporting information
Data S1. Supporting Information.
ACKNOWLEDGEMENTS
The authors gratefully acknowledge the ACCORD/ACCORDION study group and the National Heart, Lung, and Blood Institute Biologic Specimen and Data Repository Information Coordinating Center. The opinions expressed in this article are those of the authors and do not necessarily reflect the views of the ACCORD/ACCORDION study authors or the National Heart, Lung, and Blood Institute.
Wang X, Pei J, Zheng K, Liu M, Hu X. The effect of HbA1c variability on the efficacy of intensive blood pressure control in patients with type 2 diabetes. Diabetes Obes Metab. 2025;27(3):1208‐1216. doi: 10.1111/dom.16112
Contributor Information
Junyu Pei, Email: peijunyu123@outlook.com.
Xinqun Hu, Email: huxinqun@csu.edu.cn.
DATA AVAILABILITY STATEMENT
Data are available from the Biologic Specimen and Data Repository Information Coordinating Center (BioLINCC). The data can be accessed upon reasonable request at https://biolincc.nhlbi.nih.gov/studies/accord/.
REFERENCES
- 1. Adler AI, Stratton IM, Neil HA, et al. Association of systolic blood pressure with macrovascular and microvascular complications of type 2 diabetes (UKPDS 36): prospective observational study. BMJ. 2000;321(7258):412‐419. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 2. Emdin CA, Rahimi K, Neal B, Callender T, Perkovic V, Patel A. Blood pressure lowering in type 2 diabetes: a systematic review and meta‐analysis. JAMA. 2015;313(6):603‐615. doi: 10.1001/jama.2014.18574 [DOI] [PubMed] [Google Scholar]
- 3. UK Prospective Diabetes Study Group . Tight blood pressure control and risk of macrovascular and microvascular complications in type 2 diabetes: UKPDS 38. UK prospective diabetes study group. BMJ. 1998;317(7160):703‐713. [PMC free article] [PubMed] [Google Scholar]
- 4. Patel A, MacMahon S, Chalmers J, et al. Effects of a fixed combination of perindopril and indapamide on macrovascular and microvascular outcomes in patients with type 2 diabetes mellitus (the ADVANCE trial): a randomised controlled trial. Lancet. 2007;370(9590):829‐840. [DOI] [PubMed] [Google Scholar]
- 5. Cushman WC, Evans GW, Byington RP, et al. Effects of intensive blood‐pressure control in type 2 diabetes mellitus. N Engl J Med. 2010;362(17):1575‐1585. doi: 10.1056/NEJMoa1001286 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 6. Wright JT, Williamson JD, Whelton PK, et al. A randomized trial of intensive versus standard blood‐pressure control. N Engl J Med. 2015;373(22):2103‐2116. doi: 10.1056/NEJMoa1511939 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 7. Beddhu S, Chertow GM, Greene T, et al. Effects of intensive systolic blood pressure lowering on cardiovascular events and mortality in patients with type 2 diabetes mellitus on standard glycemic control and in those without diabetes mellitus: reconciling results from ACCORD BP and SPRINT. J Am Heart Assoc. 2018;7(18):e009326. doi: 10.1161/JAHA.118.009326 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 8. Currie CJ, Peters JR, Tynan A, et al. Survival as a function of HbA(1c) in people with type 2 diabetes: a retrospective cohort study. Lancet. 2010;375(9713):481‐489. doi: 10.1016/S0140-6736(09)61969-3 [DOI] [PubMed] [Google Scholar]
- 9. Grembowski D, Ralston JD, Anderson ML. Hemoglobin A1c, comorbid conditions and all‐cause mortality in older patients with diabetes: a retrospective 9‐year cohort study. Diabetes Res Clin Pract. 2014;106(2):373‐382. doi: 10.1016/j.diabres.2014.07.017 [DOI] [PubMed] [Google Scholar]
- 10. Gorst C, Kwok CS, Aslam S, et al. Long‐term glycemic variability and risk of adverse outcomes: a systematic review and meta‐analysis. Diabetes Care. 2015;38(12):2354‐2369. doi: 10.2337/dc15-1188 [DOI] [PubMed] [Google Scholar]
- 11. Forbes A, Murrells T, Mulnier H, Sinclair AJ. Mean HbA, HbA variability, and mortality in people with diabetes aged 70 years and older: a retrospective cohort study. Lancet Diabetes Endocrinol. 2018;6(6):476‐486. doi: 10.1016/S2213-8587(18)30048-2 [DOI] [PubMed] [Google Scholar]
- 12. Gerstein HC, Miller ME, Byington RP, et al. Effects of intensive glucose lowering in type 2 diabetes. N Engl J Med. 2008;358(24):2545‐2559. doi: 10.1056/NEJMoa0802743 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 13. Buse JB, Bigger JT, Byington RP, et al. Action to control cardiovascular risk in diabetes (ACCORD) trial: design and methods. Am J Cardiol. 2007;99(12A):21i‐33i. [DOI] [PubMed] [Google Scholar]
- 14. Sheng CS, Tian J, Miao Y, et al. Prognostic significance of long‐term HbA(1c) variability for all‐cause mortality in the ACCORD trial. Diabetes Care. 2020;43(6):1185‐1190. doi: 10.2337/dc19-2589 [DOI] [PubMed] [Google Scholar]
- 15. Dolan E, O'Brien E. Blood pressure variability: clarity for clinical practice. Hypertension. 2010;56(2):179‐181. doi: 10.1161/HYPERTENSIONAHA.110.154708 [DOI] [PubMed] [Google Scholar]
- 16. Rothwell PM, Howard SC, Dolan E, et al. Prognostic significance of visit‐to‐visit variability, maximum systolic blood pressure, and episodic hypertension. Lancet. 2010;375(9718):895‐905. doi: 10.1016/S0140-6736(10)60308-X [DOI] [PubMed] [Google Scholar]
- 17. Pei J, Wang X, Pei Z, Hu X. Glycemic control, HbA1c variability, and major cardiovascular adverse outcomes in type 2 diabetes patients with elevated cardiovascular risk: insights from the ACCORD study. Cardiovasc Diabetol. 2023;22(1):287. doi: 10.1186/s12933-023-02026-9 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 18. Ghouse J, Skov MW, Kanters JK, et al. Visit‐to‐visit variability of hemoglobin A1c in people without diabetes and risk of major adverse cardiovascular events and all‐cause mortality. Diabetes Care. 2019;42(1):134‐141. doi: 10.2337/dc18-1396 [DOI] [PubMed] [Google Scholar]
- 19. Cardiovascular Disease and Risk Management . Standards of medical Care in Diabetes‐2022. Diabetes Care. 2022;45(Suppl 1):S144‐S174. doi: 10.2337/dc22-S010 [DOI] [PubMed] [Google Scholar]
- 20. Joseph JJ, Deedwania P, Acharya T, et al. Comprehensive Management of Cardiovascular Risk Factors for adults with type 2 diabetes: a scientific Statement from the American Heart Association. Circulation. 2022;145(9):e722‐e759. doi: 10.1161/CIR.0000000000001040 [DOI] [PubMed] [Google Scholar]
- 21. Buckley LF, Dixon DL, Wohlford GF, Wijesinghe DS, Baker WL, Van Tassell BW. Intensive versus standard blood pressure control in SPRINT‐eligible participants of ACCORD‐BP. Diabetes Care. 2017;40(12):1733‐1738. doi: 10.2337/dc17-1366 [DOI] [PubMed] [Google Scholar]
- 22. Margolis KL, O'Connor PJ, Morgan TM, et al. Outcomes of combined cardiovascular risk factor management strategies in type 2 diabetes: the ACCORD randomized trial. Diabetes Care. 2014;37(6):1721‐1728. doi: 10.2337/dc13-2334 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 23. Shi S, Gouskova N, Najafzadeh M, Wei L‐J, Kim DH. Intensive versus standard blood pressure control in type 2 diabetes: a restricted mean survival time analysis of a randomised controlled trial. BMJ Open. 2021;11(9):e050335. doi: 10.1136/bmjopen-2021-050335 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 24. Critchley JA, Carey IM, Harris T, DeWilde S, Cook DG. Variability in glycated hemoglobin and risk of poor outcomes among people with type 2 diabetes in a large primary care cohort study. Diabetes Care. 2019;42(12):2237‐2246. doi: 10.2337/dc19-0848 [DOI] [PubMed] [Google Scholar]
- 25. Li S, Nemeth I, Donnelly L, Hapca S, Zhou K, Pearson ER. Visit‐to‐visit HbA variability is associated with cardiovascular disease and microvascular complications in patients with newly diagnosed type 2 diabetes. Diabetes Care. 2020;43(2):426‐432. doi: 10.2337/dc19-0823 [DOI] [PubMed] [Google Scholar]
- 26. Mo Y, Zhou J, Ma X, et al. Haemoglobin A1c variability as an independent correlate of atherosclerosis and cardiovascular disease in Chinese type 2 diabetes. Diab Vasc Dis Res. 2018;15(5):402‐408. doi: 10.1177/1479164118778850 [DOI] [PubMed] [Google Scholar]
- 27. Orsi E, Solini A, Bonora E, et al. Haemoglobin A1c variability is a strong, independent predictor of all‐cause mortality in patients with type 2 diabetes. Diabetes Obes Metab. 2018;20(8):1885‐1893. doi: 10.1111/dom.13306 [DOI] [PubMed] [Google Scholar]
- 28. MacMahon S, Peto R, Cutler J, et al. Blood pressure, stroke, and coronary heart disease. Part 1, prolonged differences in blood pressure: prospective observational studies corrected for the regression dilution bias. Lancet. 1990;335(8692):765‐774. doi: 10.1016/0140-6736(90)90878-9 [DOI] [PubMed] [Google Scholar]
- 29. Law MR, Morris JK, Wald NJ. Use of blood pressure lowering drugs in the prevention of cardiovascular disease: meta‐analysis of 147 randomised trials in the context of expectations from prospective epidemiological studies. BMJ. 2009;338:b1665. doi: 10.1136/bmj.b1665 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 30. Williams B. Treating hypertension in patients with diabetes: when to start and how low to go? JAMA. 2015;313(6):573‐574. doi: 10.1001/jama.2015.89 [DOI] [PubMed] [Google Scholar]
- 31. Karmali KN, Lloyd‐Jones DM, van der Leeuw J, et al. Blood pressure‐lowering treatment strategies based on cardiovascular risk versus blood pressure: a meta‐analysis of individual participant data. PLoS Med. 2018;15(3):e1002538. doi: 10.1371/journal.pmed.1002538 [DOI] [PMC free article] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
Supplementary Materials
Data S1. Supporting Information.
Data Availability Statement
Data are available from the Biologic Specimen and Data Repository Information Coordinating Center (BioLINCC). The data can be accessed upon reasonable request at https://biolincc.nhlbi.nih.gov/studies/accord/.