Improved prognosis prediction of all‐cause mortality and cardiovascular events by combined assessment of both ankle‐brachial index and brachial‐ankle pulse wave velocity in individuals with diabetes: The Kyushu prevention study for atherosclerosis, a prospective, multicenter survey

ABSTRACT

Aims/Introduction

People with diabetes exhibit early‐stage cardiovascular disease (CVD). In this study, we developed a new diagnostic algorithm to predict all‐cause mortality and CVD using both the ankle‐brachial index (ABI) and brachial‐ankle pulse wave velocity (baPWV) in individuals with diabetes.

Materials and Methods

This study utilized data from the Kyushu Prevention Study for Atherosclerosis, a prospective, multicenter survey. A total of 3,933 participants with diabetes were followed. The primary outcomes were all‐cause mortality and CVD. Risk was categorized based on ABI and baPWV, and the prognostic differences were evaluated. Additionally, the independence of this classification from the Framingham Risk Score (FRS) was tested.

Results

ABI was the strongest predictor of all‐cause mortality. Individuals with a low ABI (<0.70) had the highest risk, while those with an intermediate ABI (0.70–1.0) exhibited a risk similar to individuals with the extremely high baPWV (≥24 m/s), forming the second highest‐risk group. For coronary heart disease, ABI and baPWV had complementary predictive value. Individuals with a low ABI (<1.0) or high baPWV (≥19 m/s) were at the highest risk, whereas those with a low baPWV (<14 m/s) had the lowest risk. For cerebrovascular disease, only baPWV was a significant predictor, with risk cutoffs above 23 m/s and below 14 m/s.

Conclusions

We developed a new diagnostic algorithm incorporating both ABI and baPWV to predict all‐cause mortality and CVD in individuals with diabetes. This algorithm improved the accuracy of mortality and CVD risk assessment and was independent from the FRS.

A novel algorithm combining ABI and baPWV was developed to predict mortality and CVD risk in individuals with diabetes. This algorithm improved the accuracy of mortality and CVD risk assessment and was independent from the Framingham Risk Score.

INTRODUCTION

Just over half a billion people worldwide are living with diabetes, accounting for more than 10.5% of the global adult population, and this number is increasing rapidly ¹ . The risk of death from cardiovascular disease (CVD) in individuals with diabetes is more than 2.32 times higher than in those without diabetes ² . Among diabetic complications, peripheral artery disease (PAD) is one of the particularly concerning, as it often progresses silently and is associated with poor cardiovascular outcomes. In a study of Japanese subjects, the prevalence of PAD was 2.7% in individuals without diabetes ³ , while in a separate study we previously conducted, the prevalence in individuals with diabetes was reported to be 7.6% ⁴ . These findings suggest that diabetes significantly increases PAD risk, necessitating early detection and intervention.

Previous studies have shown that non‐invasive tests can identify individuals at high risk of all‐cause mortality by identifying PAD. The presence of PAD is associated with a worse prognosis, and its severity correlates with increased mortality ⁵ , ⁶ . A meta‐analysis by the Ankle Brachial Index Collaboration reported that measurement of ankle‐brachial index (ABI), a measurement of arterial stenosis, may improve the accuracy of cardiovascular risk prediction beyond the Framingham risk score (FRS) ⁷ . Meanwhile, we previously demonstrated that brachial‐ankle pulse wave velocity (baPWV), a marker of arterial stiffness, serves as a useful independent predictor of mortality and cardiovascular morbidity in individuals with diabetes ⁸ . Since baPWV and ABI can be measured simultaneously and assess different vascular functions, we hypothesized that their combined use would provide a more effective diagnostic algorithm than either marker alone.

In this study, we aimed to develop a new diagnostic algorithm incorporating both ABI and baPWV to improve the prediction of all‐cause mortality and CVD in individuals with diabetes.

MATERIALS AND METHODS

Study population

This study was based on data from the Kyushu Prevention Study for Atherosclerosis, a prospective, multicenter survey. Outpatients with diabetes (n = 4,272) and without diabetes (n = 2,166) who regularly visited Kyushu University Hospital and its 17 related hospitals, as well as University of the Ryukyus Hospital and its six related hospitals were enrolled in this survey from 2001 to 2003 ⁴ , ⁸ , ⁹ . Baseline assessment included measurements of height, weight, and systolic and diastolic blood pressure; 12‐lead electrocardiography; eye fundus examination; and laboratory tests of blood and urine. Medical records were reviewed for a history of cardiovascular events, current treatment for diabetes, and the use of medications such as angiotensin II receptor blockers (ARBs), angiotensin‐converting enzyme inhibitors (ACEIs), statins, antiplatelet agents (APAs), and/or anti‐thrombotic agents (ATAs). Participants were followed up for 3.2 ± 2.2 (mean ± SD) years. The protocol for this research project has been approved by a suitably constituted Ethics Committee, the Kyushu University Hospital Ethics Committee, and it conforms to the provisions of the Declaration of Helsinki. Written informed consent was obtained from all subjects.

Measurement of baPWV and ABI

At baseline, baPWV was measured using an oscillometric device (form PWV/ABI; Omron Colin Co. Ltd., Komaki, Japan). Four pneumatic pressure cuffs, two electrocardiogram electrodes, and a microphone for detecting heart sounds were attached at both arms and ankles, as well as the wrists and the left edge of the sternum, respectively, to record volume waveforms from the brachial and ankle arteries. ABI was also automatically calculated by the device as the ratio of systolic blood pressure in the leg to that in the arm on each side (Vascular Profiling System 2000, Omron Colin Co., Ltd., Komaki, Japan). The lower of the left and right ABI values was used for analysis ¹⁰ . ABI and baPWV were measured in all participants. Recursive partitioning analysis (RPA) was conducted using both indices, without prior exclusion based on ABI values. This approach allowed data‐driven identification of risk thresholds across the full spectrum of arterial status.

Clinical assessment

Hemoglobin A1c (HbA1c) levels were initially obtained as Japanese Diabetes Society (JDS) values. In this study, we present HbA1c values standardized according to the National Glycohemoglobin Standardization Program (NGSP), calculated as follows: JDS value +0.4 (%) ¹¹ and values were converted to International Federation of Clinical Chemistry and Laboratory Medicine (IFCC) mmol/mol units using the NGSP converter, available at http://www.ngsp.org/convert1.asp. Diabetic neuropathy was diagnosed by diabetologists based on typical symptoms and physical examination findings. Retinopathy was assessed by fundus examination by independent ophthalmologists. Clinical proteinuria was defined as a level of ≥1+ using the Albustix method or macroalbuminuria (≥300 mg/gCre). Hypertension was defined as systolic blood pressure ≥140 mmHg, diastolic blood pressure ≥90 mmHg, or the use of any antihypertensive medication. Dyslipidemia was on a prior diagnosis or if total cholesterol was ≥220 mg/dL, triglyceride was ≥150 mg/dL, or high‐density lipoprotein cholesterol was <40 mg/dL. While this differs from the latest Japan Atherosclerosis Society Guidelines for Prevention of Atherosclerotic Cardiovascular Diseases 2022 ¹² , we retained this definition to align with the FRS components used in our study. Smoking status was defined as current smoking through interviews.

Outcomes

The primary outcomes of this analysis were all‐cause mortality and the occurrence of major cardiovascular events, including coronary heart disease (CHD); fatal and nonfatal myocardial infarction, unstable angina, and cerebrovascular disease; fatal and nonfatal stroke and transient cerebral ischemic attack.

Statistical analysis

Continuous variables are presented as means ± SD or median (lower quartile–upper quartile), while categorical variables are expressed as frequencies and percentages. To classify subjects into high‐, intermediate‐, and low‐risk groups based on baPWV and ABI, cutoff values were determined using RPA to identify optimal thresholds for risk stratification. The RPA is a non‐parametric, data‐driven statistical method that constructs decision trees by repeatedly splitting the dataset into subgroups based on predictor variables. At each step, the algorithm selects the variable and corresponding cutoff that best separate the outcome of interest, optimizing classification or prediction. This method is particularly useful for identifying clinically meaningful threshold values in continuous variables ¹³ . Event‐free survival curves for each endpoint were estimated using the Kaplan–Meier method, and Cox proportional hazards regression models were employed to determine the independent predictive value of baPWV and ABI for mortality and cardiovascular disease, even after adjustment for the FRS. FRS was calculated using regression equations provided by the British Cardiac Society ¹⁴ , ¹⁵ . The FRS model for the prediction of coronary heart disease (including myocardial infarction, coronary heart disease‐related death, angina, and coronary insufficiency) and stroke (including transient ischemic attack) incorporates variables such as time period, age, sex, smoking history (including former smoker), diabetes status, total cholesterol, HDL cholesterol, systolic blood pressure, and the presence of left ventricular hypertrophy (LVH) on electrocardiography. Since LVH data was unavailable, it was assigned a value of zero. The time period for risk prediction was set to 5 years.

To determine the most effective predictive model among ABI alone, baPWV alone, or their combination, we performed receiver operating characteristic (ROC) analysis and compared the area under the curve (AUC) of each method.

A two‐tailed P‐value <0.05 was considered statistically significant. All analyses were conducted using JMP Pro version 11 (SAS Institute Inc., Cary, NC) and EZR (Saitama Medical Center, Jichi Medical University, Saitama, Japan), a graphical user interface for R (The R Foundation for Statistical Computing, Vienna, Austria). EZR is a modified version of R Commander designed to induce statistical functions frequently used in biostatistics ¹⁶ .

RESULTS

Baseline characteristics of the study population

A total of 4,272 participants with diabetes from the Kyushu Prevention Study for Atherosclerosis were enrolled, while 278 subjects were excluded due to safety concerns, including known severe PAD or diabetic foot, invalid measurements caused by incompressible arteries, or withdrawal of consent. A total of 3,994 participants were enrolled at baseline. After excluding 61 participants due to missing ABI and/or baPWV data, the eligible 3,933 participants with both ABI and baPWV were included in the analysis. The flow diagram of analysis is shown in Figure 1.

Figure 1.

Flow diagram of analysis: This flow diagram illustrates the study population selection process and outlines the exclusion criteria, which included individuals with known severe peripheral artery disease (PAD), diabetic foot, invalid measurements due to incompressible arteries, and those who withdrew consent. Ankle‐brachial index (ABI) and brachial‐ankle pulse wave velocity (baPWV) were measured in all participants.

The median age of the participants was 62 years, with 41% being female. The median ABI was 1.10, and the median baPWV was 17 m/s. Among the participants, 52% had hypertension, 49% had dyslipidemia, 26% had diabetic retinopathy, 24% had nephropathy, and 24% had diabetic neuropathy. Table 1 describes the demographic and clinical characteristics of the participants. During a mean follow‐up period of 3.3 ± 2.2 years, 258 subjects (6.6%) died from all causes, 181 subjects (5.0%) developed CHD, and 153 subjects (3.9%) experienced a new onset of cerebrovascular disease.

Table 1.

Characteristics at baseline

			n	3,994
Age (years old)			3,932	62 [53, 70]
Gender (%)	Female			1,595 (40.6)
	Male			2,338 (59.4)
BMI (kg/m2)			3,796	24.20 [21.90, 26.80]
Current smoker (%)			3,933	960 (24.4)
HbA1c (%, mmol/mol)			3,595	7.80 [6.60, 9.40], 62 [49, 79]
Serum creatinine (mg/dL)			3,416	0.70 [0.60, 0.90]
Uric acid (mg/dL)			3,156	5.25 [4.30, 6.30]
Hypertension (%)	Yes		3,933	2062 (52.4)
		Systolic blood pressure (mmHg)		135.00 [123.00, 150.00]
		Diastolic blood pressure (mmHg)		80.00 [72.00, 88.00]
Dyslipidemia (%)	Yes		3,933	1922 (48.9)
		Total cholesterol (mg/dL)		200.50 [178.00, 225.00]
		Triglycerides (mg/dL)		125.00 [88.00, 186.00]
		HDL cholesterol (mg/dL)		50.00 [41.00, 60.00]
Neuropathy (%)			3,933	948 (24.1)
Retinopathy (%)			3,933	1,003 (25.5)
Nephropathy (%)	Yes		3,933	949 (24.1)
		Proteinuria (%)	3,059	593 (19.4)
Lower ABI			3,933	1.10 [1.04, 1.15]
baPWV (m/s)			3,933	16.8 [14.5–19.9]

Data are presented as median [Q1, Q3] for continuous variables or as number (percentage) for categorical variables. ABI, ankle‐brachial index; baPWV, brachial‐ankle pulse wave velocity; BMI, body mass index, calculated as weight (kg) divided by height squared (m²); HbA1c, hemoglobin A1c; HDL cholesterol, high‐density lipoprotein cholesterol.

Risk classification based on ABI and baPWV

Participants were categorized into high‐risk, intermediate‐risk, and low‐risk groups, using RPA based on ABI and baPWV for each outcome.

For all‐cause mortality, the high‐risk group included participants with ABI <0.7. The intermediate‐risk group comprised participants with 0.7 ≤ ABI <1.0, or ABI ≥1.0 and baPWV ≥24 m/s. The low‐risk group included participants with ABI ≥1.0 and baPWV <24 m/s. For CHD, the high‐risk group included participants with ABI <1.0 or ABI ≥1.0 and baPWV ≥19 m/s. The intermediate‐risk group comprised participants with ABI ≥1.0 and 14 m/s ≤ baPWV <19 m/s. The low‐risk group included participants with ABI ≥1.0 and baPWV <14 m/s. For cerebrovascular disease, the high‐risk group included participants with ABI ≥1.1 and PWV ≥23 m/s. The intermediate‐risk group comprised participants with ABI <1.1, or ABI ≥1.1 and baPWV ≥14 m/s. The low‐risk group included the participants with ABI ≥1.1 and baPWV <14 m/s. The flowchart of risk classification using ABI and baPWV is shown in Figure 2.

Figure 2.

Risk classification based on ankle‐brachial index (ABI) and brachial‐ankle pulse wave velocity (baPWV) for all‐cause mortality and cardiovascular events using recursive partitioning analysis. As shown in the flowchart, participants were categorized into three groups: high‐risk, intermediate‐risk, and low‐risk, according to their incidence rates. Differences in incidence rates among the three groups were assessed using Pearson's χ² test. (a) All‐cause mortality: The high‐risk group was defined by a significantly lower ABI (<0.70). The intermediate‐risk group included individuals with ABI 0.7 or more, but less than 1.0, or those with ABI ≥1.0 and baPWV ≥24 m/s. (b) Coronary heart disease (CHD): Risk stratification was determined by both ABI and baPWV. The high‐risk group included individuals with ABI <1.0 or baPWV ≥19 m/s. The intermediate‐risk group comprised those with ABI ≥1.0 and baPWV 14 or more, but less than 19 m/s. (c) Cerebrovascular disease: Risk was primarily determined by baPWV. The high‐risk group consisted of individuals with baPWV ≥23 m/s, while the intermediate‐risk group included those with baPWV 14 or more, but less than 23 m/s.

Survival analysis and risk comparison

Kaplan–Meier survival analyses demonstrated significant differences among the three risk groups for each outcome (Figure 3). For all‐cause mortality, each of the three groups exhibited distinct prognosis. The hazard ratio of all‐cause mortality in the high‐risk group compared with the low‐risk group was 4.6 (95% CI: 2.7–7.6), P < 0.001 (Table 2a), independent of the FRS. For CHD, significant differences were observed between the high‐risk group and low‐risk groups and between the intermediate‐risk group and low‐risk groups but not between the high‐risk group and intermediate‐risk groups. The hazard ratio of CHD in the high‐risk group compared with the low‐risk group was 1.8 (95% CI: 1.0–3.2), P = 0.039, indicating an independent explanatory factor from the FRS (Table 2b). For cerebrovascular disease, the hazard ratio in the high‐risk group compared with the low‐risk group was 3.1 (95% CI: 1.7–5.8), P < 0.001, and in the high‐risk group compared with the intermediate‐risk group was 2.1 (95% CI: 1.4–3.0), P < 0.001. The incidence rate in the high‐risk group classified by ABI and baPWV was significantly different from other groups, whereas the FRS was not a significant explanatory factor for cerebrovascular disease. When the FRS was included in the multivariate analysis, the combination of ABI and baPWV was no longer a significant explanatory factor (Table 2c).

Figure 3.

Kaplan–Meier event‐free survival curves for all‐cause mortality and cardiovascular events, stratified by risk classification. The panels illustrate survival probabilities over time for all‐cause mortality, coronary heart disease (CHD), and cerebrovascular disease, categorized into high‐risk, intermediate‐risk, and low‐risk groups based on ankle‐brachial index (ABI) and brachial‐ankle pulse wave velocity (baPWV).

Table 2.

Cox proportional hazards regression models for all‐cause mortality and cardiovascular events stratified by risk classification

			Model 1		Model 2		Model 3
			Relative risk (95% CI)	P‐value	Relative risk (95% CI)	P‐value	Relative risk (95% CI)	P‐value
(a)	FRS		12.9 (5.1–29.5)	<0.001			4.6 (1.7–11.8)	0.004
	ABI‐PWV risk classification
		Intermediate vs. low			1.9 (1.4–2.5)	<0.001	2.1 (1.4–3.0)	<0.001
		High vs. low			4.3 (3.0–6.2)	<0.001	4.6 (2.7–7.6)	<0.001
		High vs. intermediate			2.3 (1.5–3.4)	<0.001	2.2 (1.3–3.7)	0.003
(b)	FRS		1.1 (1.0–1.1)	<0.001			5.3 (2.0–13.6)	<0.001
	ABI‐PWV risk classification
		Intermediate vs. low			1.8 (1.3–2.6)	<0.001	1.9 (1.1–3.5)	0.013
		High vs. low			1.8 (1.3–2.7)	<0.001	1.8 (1.0–3.2)	0.039
		High vs. intermediate			1.0 (0.8–1.3)	0.779	0.9 (0.7–1.2)	0.543
(c)	FRS		4.5 (0.9–17.7)	0.066			3.6 (0.7–15.2)	0.13
	ABI‐PWV risk classification
		Intermediate vs. low			1.5 (0.9–2.5)	0.086	1.0 (0.6–2.0)	0.898
		High vs. low			3.1 (1.7–5.8)	<0.001	1.5 (0.7–3.5)	0.379
		High vs. intermediate			2.1 (1.4–3.0)	<0.001	1.4 (0.7–2.6)	0.302

(a) Cox proportional hazards regression models for all‐cause mortality stratified by risk classification. (b) Cox proportional hazards regression models for coronary heart disease (CHD) stratified by risk classification. (c) Cox proportional hazards regression models for cerebrovascular disease stratified by risk classification.

Model 1 estimates the hazard ratio (HR) based on the Framingham risk score (FRS).

Model 2 incorporates risk classification using ankle‐brachial index (ABI) and brachial‐ankle pulse wave velocity (baPWV) thresholds determined through recursive partitioning analysis, with HR and P‐values calculated using the Cox model.

Model 3 adjusts for both ABI–baPWV risk classification and FRS to evaluate their combined predictive power. ABI, ankle‐brachial index; baPWV, brachial‐ankle pulse wave velocity; FRS, Framingham risk score; HR, hazard ratio; CI, confidence interval.

Predictive performance of ABI and baPWV

To evaluate the predictive power of ABI and baPWV, we conducted ROC analysis and compared the AUC for different predictive models (Figure 4). For all‐cause mortality, the AUC was significantly higher for the combination of ABI and baPWV (AUC: 0.62, 95% CI: 0.58–0.65) than for ABI alone (AUC: 0.56, 95% CI: 0.54–0.59, P < 0.01), or baPWV alone (AUC: 0.53, 95% CI: 0.51–0.55, P < 0.01) (Figure 4a). Similarly, for CHD, the AUC was significantly higher for the combination of ABI and baPWV (AUC: 0.58, 95% CI: 0.55–0.62) than for ABI alone (AUC: 0.52, 95% CI: 0.50–0.54, P < 0.01), or baPWV alone (AUC: 0.54, 95% CI: 0.52–0.57, P < 0.04) (Figure 4b). For cerebrovascular disease, the AUC for the combination of ABI and baPWV (AUC: 0.59, 95% CI: 0.55–0.62) was significantly higher than that for ABI alone (AUC: 0.51, 95% CI: 0.49–0.53, P < 0.01), but not significantly higher than that for baPWV alone (AUC: 0.59, 95% CI: 0.56–0.60, P = 0.74) (Figure 4c).

Figure 4.

Receiver operating characteristic (ROC) analysis comparing the predictive performance of ABI alone, baPWV alone, and their combined model for all‐cause mortality and cardiovascular events prediction. The figure illustrates the area under the curve (AUC) for each method, demonstrating the improved accuracy of risk stratification when both ankle‐brachial index (ABI) and brachial‐ankle pulse wave velocity (baPWV) are used together.

DISCUSSION

This study developed and validated a new diagnostic algorithm incorporating both the ABI and baPWV to predict all‐cause mortality and CVD in individuals with diabetes. The algorithm demonstrated improved accuracy in risk assessment compared to conventional models and remained independent from the FRS, which integrates and weights traditional cardiovascular risk factors. The clinical significance of this algorithm is notable for three reasons. First, it enhances risk prediction beyond the FRS. Second, ABI and baPWV can be measured simultaneously, allowing for efficient assessment. Third, both measurements are non‐invasive, making them practical for routine clinical use.

To define ABI–baPWV risk categories, we did not apply pre‐specified cut‐offs. Instead, we used RPA to identify data‐driven thresholds for ABI and baPWV that best discriminated all‐cause mortality, coronary heart disease, and cerebrovascular disease in this diabetic cohort.

In this large‐scale study, ABI was the strongest predictor of all‐cause mortality. Participants with a low ABI (<0.70) had the highest risk, while those with an intermediate ABI (0.70–1.0) had a risk level comparable to individuals with extremely high baPWV (≥24 m/s), forming the second highest‐risk group (Figure 2a). For CHD, ABI and baPWV played complementary roles in risk prediction. The highest‐risk group included individuals with either a low ABI (<1.0) or a high baPWV (≥19 m/s), whereas those with a low baPWV (<14 m/s) were classified as the lowest‐risk group (Figure 2b). For cerebrovascular disease, baPWV was the sole predictor, with significant risk differentiation at cut points of ≥23 m/s and < 14 m/s. These findings suggest that the combination of ABI and baPWV improves risk prediction for all‐cause mortality and CHD, but for cerebrovascular disease, baPWV alone appears to be the dominant predictor (Figure 2c). Previous epidemiological studies and expert reviews in Japanese populations have generally considered baPWV <14 m/s as low risk, 14–18 m/s as intermediate risk, and ≥ 18 m/s as high risk ¹⁷ , ¹⁸ , ¹⁹ . Our RPA‐derived thresholds therefore lie predominantly in the intermediate‐to‐high‐risk range, which is consistent with these prior reports. Our data‐driven algorithm is compatible with existing expert consensus, while providing optimized cut‐off values for prognosis in individuals with diabetes.

Brachial‐ankle pulse wave velocity (baPWV) is a well‐established marker of arterial stiffness and an independent predictor of CVD risk ⁸ , ²⁰ , ²¹ , ²² , ²³ . Although baPWV is susceptible to blood pressure fluctuations, it has shown a strong correlation with carotid‐femoral PWV ²⁴ and is widely utilized in Japanese epidemiological studies. Furthermore, its utility as a predictor of cardiovascular outcomes has been established in large Japanese cohorts ²⁰ , ²⁵ . ABI is also a well‐recognized indicator of arterial stenosis. However, the United States Preventive Service Task Force does not currently recommend ABI screening for PAD and CVD in asymptomatic adults due to insufficient evidence supporting its routine use in the general population ²⁶ . Nevertheless, multiple studies have demonstrated that abnormal ABI (≤0.90, 0.91–1.00 or ≥1.30 or 1.40) is predictive of mortality and CVD events, particularly in high‐risk populations such as individuals with diabetes ⁷ , ⁹ , ²⁷ , ²⁸ . While both baPWV and ABI are useful predictors, our findings demonstrate that combining these two markers significantly improves predictive accuracy, as reflected in the AUC. This enhancement is likely due to the complementary roles: ABI identifies arterial stenosis, while baPWV measures arterial stiffness and atherosclerosis severity. Together, they provide a more comprehensive assessment of vascular health and overall cardiovascular risk. A previous report found that individuals with diabetes and an ABI >0.90 exhibited reduced blood flow in the lower‐leg arteries, suggesting that ABI alone may not fully capture peripheral arterial disease severity in this population ²⁹ . Additionally, studies have indicated that an abnormally high ABI is associated with increased CVD risk and all‐cause mortality in the Japanese population ²⁷ , ³⁰ . These findings highlight the limitations of relying solely on ABI for risk stratification in diabetes.

Based on these findings, we advocate for the combined use of ABI and baPWV rather than relying on either test alone. This approach enhances the detection of high‐risk individuals, particularly those with diabetes, and provides a more accurate assessment of all‐cause mortality and CVD risk.

We used the FRS to estimate cardiovascular risk. Although it is known to overestimate CVD risk in Japanese individuals ³¹ we selected the FRS due to its widespread use and comparability across studies. Nonetheless, future research should consider incorporating Japanese‐specific risk models such as the Suita Score ³¹ , ³² or the prediction model of the Hisayama Study ³³ , ³⁴ to validate the findings more appropriately in this population.

Although ABI >1.3 is known to be associated with increased CVD risk in individuals with diabetes, the proportion of such cases in our cohort was 2.3%, limiting our ability to evaluate this subgroup. Future studies with larger sample sizes are warranted to investigate the utility of combined ABI and baPWV measurements in these high‐risk individuals.

Limitation

The present study has certain limitations. First, the relatively short follow‐up period (average of 3.2 years) may have limited our ability to capture long‐term cardiovascular events. As a result, the study may underestimate the long‐term predictive value of ABI and baPWV, particularly for cardiovascular events that develop over extended periods. Additionally, we were unable to collect sufficient data on periodic baPWV measurements during follow‐up, preventing an assessment of how baPWV progression over time correlates with outcomes. However, despite these limitations, the survival analyses consistently demonstrated the predictive power of the combined ABI and baPWV model. Second, we acknowledge that the reproducibility of baPWV was not assessed in our cohort. However, previous studies have demonstrated acceptable reproducibility of baPWV measurements ³⁵ . Third, although severe peripheral arterial disease may lead to pseudo‐low baPWV when ABI is ≤0.95 due to impaired pulse‐wave transmission, this phenomenon does not influence our outcome‐based risk classification. In the RPA model, all individuals with low ABI were consistently assigned to the high‐risk (or intermediate‐risk) groups across all endpoints. Thus, the combination of low ABI and high baPWV observed in our results reflects advanced arterial disease rather than a methodological contradiction. Forth, although the Framingham Risk Score was used to estimate cardiovascular risk, it was originally developed for individuals under 75 years of age. Given that only a small proportion of our cohort exceeded this age, the impact on our findings is likely limited. Nevertheless, future studies should consider employing age‐appropriate risk models, such as the Suita score ³¹ , ³² , particularly in cohorts with a higher proportion of elderly participants. Fifth, smoking history was included as a variable, but our definition was limited to current smokers, as data on ex‐smokers were not consistently available. This might have underestimated the impact of smoking on vascular health. While previous studies have demonstrated a strong link between smoking and baPWV elevation ³⁶ , ³⁷ , ³⁸ , our analysis did not find smoking to be a significant variable in the Cox proportional hazards models. This may be due to the exclusion of ex‐smokers or the potential confounding effects of other vascular risk factors. In clinical practice, individuals at high risk for atherosclerosis should still be advised to stop smoking, as the overall prevalence of current smoking in this cohort was low. Sixth, this study was conducted exclusively in Japanese subjects, which may limit the generalizability of our findings to other populations. Previous studies have demonstrated ethnic differences in both ABI and baPWV values ³⁹ , ⁴⁰ , suggesting that cut‐off values derived from Japanese cohorts may not be directly applicable to other ethnic groups. Additionally, the baPWV and ABI measurements in this study were obtained using the Omron device (Omron Colin Co. Ltd., Komaki, Japan), a system predominantly used in Asia. Validation in populations with different ethnic backgrounds, along with testing using alternative measurement devices, will be necessary before applying this diagnostic algorithm in a global context. Seventh, medication use—including antihypertensive, lipid‐lowering, antiplatelet, and glucose‐lowering agents—changed frequently during follow‐up, and detailed time‐varying exposure data were not available. Because adjusting only for baseline medication status would lead to misclassification and potential bias, we did not include medication variables in the multivariable models. This limitation is inherent to the real‐world observational design.

CONCLUSION

We demonstrated that the combination of ABI and baPWV improves the prediction of all‐cause mortality and cardiovascular disease (CVD) in individuals with diabetes more effectively than either measure alone. This is the first large‐scale cohort study in a population with diabetes to show that the combined use of ABI and baPWV independently predicts all‐cause mortality and CVD beyond traditional risk factors, using optimized cut‐off values for each risk category. Because ABI and baPWV can be measured simultaneously and are non‐invasive, their combined assessment provides a practical and valuable tool for evaluating CVD risk and overall prognosis in individuals with diabetes.

DISCLOSURE

The authors declare no conflict of interest.

Approval of the research protocol: The protocol for this research project has been approved by a suitably constituted Ethics Committee of Kyushu university hospital, Approval No. 14‐246, and it conforms to the provisions of the Declaration of Helsinki.

Informed Consent: Written informed consent was obtained from all subjects.

Approval date of Registry and the Registration No. of the study/trial: UMIN Clinical Trial Registry, UMIN000011245.

Animal Studies: N/A.

DATA AVAILABILITY STATEMENT

The data that support the findings of this study are available on request from the corresponding author. The data are not publicly available due to privacy or ethical restrictions.