Use of the Combination of Ankle-brachial Index and Percentage of Mean Arterial Pressure at Ankle for Improving Prediction of All-cause Mortality in type 2 Diabetes Mellitus

Background: Peripheral artery disease (PAD) in lower extremities is a common complication in type 2 diabetes and has shown to be associated with mortality. The ankle-brachial index (ABI) is a simple noninvasive method to screen PAD, but has limited sensitivity. We hypothesized that using the percentage of mean arterial pressure (%MAP) and the ABI in combination would improve prediction of mortality. Methods: We retrospectively collected the data of patients with type 2 diabetes who had undergone measurement of ABI and %MAP at our hospital. We separated the cohort into four groups according to the ABI and %MAP values, and examined these indices were associated with mortality. Results: A total of 5101 patients (mean age, 65 ± 11 years) were enrolled. During the follow-up period (median, 22.9 months), 266 (4.8%) of enrolled patients died. The combination of ABI and %MAP was signicantly better at predicting mortality than ABI alone. (C index: 0.62 [95% CI: 0.57, 0.65] vs. 0.57 [95% CI: 0.53, 0.62], P = 0.038). In multivariate analysis (with ABI >0.90 and %MAP ≤ 45% as the reference group), the highest risk of mortality was seen in patients with ABI ≤ 0.90 and %MAP >45% (hazard ratio = 1.983 [95% CI: 1.380, 2.848], P < 0.001). Conclusions: Adding %MAP to ABI appears to signicantly improve the predictive ability for all-cause mortality in patients with type 2 diabetes. ltration rate; HbA1c: hemoglobin A1c; HDL: high-density lipoprotein; HR: hazard ratio; IDI: integrated discrimination improvement; NRI: net reclassication improvement; P4P: pay-for-performance; PAD: peripheral artery disease; SD: standard deviation; UACR: urinary albumin-to-creatinine ratio.

The %MAP value was automatically determined based on the ankle pulse volume waveform during ABI measurement. The reproducibilities of the ABI and %MAP have been demonstrated using Bland-Altman plots in our previous study [15]. The lower ABI value, and the higher %MAP and baPWV values between lower limbs in an individual were used for the analyses. ABI ≤ 0.90 and %MAP > 45% were de ned as abnormal [12,15].

Statistical analysis
Continuous data were summarized as the mean ± standard deviation; differences among four study subgroups were analyzed using the one-way analysis of variance, and the Scheffe post hoc test was conducted to examine the differences between the high %MAP and normal %MAP subgroups in patients with a normal ABI group or a low ABI group. Categorical data were summarized as number with percentage (%) and compared among groups using the chi-square test. The primary endpoint was all-cause mortality. Information on deaths registered up to August 31, 2019 was obtained from the Ministry of Health and Welfare, Executive Yuan, Taiwan. Improvement in prediction of mortality caused by addition of the %MAP to ABI was assessed by examining the increments in the area under the receiver operating characteristic curve (AUC). The performances of the model containing the combination of ABI and %MAP and the model with ABI alone were evaluated by the C index. Integrated discrimination improvement (IDI) and continuous net reclassi cation improvement (NRI) were also assessed.
Cumulative risk for the all-cause mortality was assessed using Kaplan-Meier analysis; the log-rank test was used to determine if the differences between groups were signi cant. Multivariable Cox proportional hazards regression analysis was conducted to identify the independent predictors of mortality; hazard ratio (HR) and 95% con dence interval (CI) were calculated. Two-sided P value < 0.05 was considered statistically signi cant. Statistical analysis was performed using SPSS v22.0 (IBM Corp., Armonk, NY, USA), and R software v3.4.

Results
A total of 5569 patients were enrolled in this study, and %MAP was inversely correlated with ABI (Pearson correlation coe cient: -4.70, P < 0.001). Based on the ABI value, all patients were rst separated into two groups: a normal ABI group and a low ABI group. Each group was then separated into two subgroups according to the %MAP value. Thus, there were four subgroups: patients with normal ABI and normal %MAP (n = 4601); patients with normal ABI but high %MAP (n = 500); patients with low ABI but normal %MAP (n = 130); and patients with low ABI and high %MAP (n = 338, Fig. 1). Table 1 showed the baseline characteristics of patients in the different subgroups. Patients with high %MAP were signi cantly older than patients with normal %MAP in both the normal ABI group (70 ± 12 vs. 64 ± 10 years, P < 0.001) and the low ABI group (73 ± 12 vs. 65 ± 12 years, P < 0.001). The baPWV was signi cantly higher in the high %MAP subgroup than in the normal %MAP subgroup in both the normal ABI group (P < 0.001) and the low ABI group (P < 0.001). BMI, ABI and eGFR were signi cantly lower in the high %MAP subgroup than in the normal %MAP subgroup in both the normal ABI group (P = 0.027, P < 0.001, and P < 0.001; respectively) and the low ABI group (all P values < 0.001). Prevalence of CVD, albuminuria and use of antiplatelet drugs were signi cantly higher in the high %MAP subgroup than in the normal %MAP subgroup in both the normal ABI group (all P values < 0.001) and the low ABI group (P = 0.015, P = 0.004, and P < 0.001; respectively). The proportions of patients using oral antihyperglycemic drugs was signi cantly lower in the high %MAP subgroup than in the normal %MAP subgroup in both the normal ABI group (P < 0.001) and the low ABI group (P = 0.008). Patients with high %MAP were signi cantly more likely to be female, to have hypertension and higher systolic blood pressure, to be using antihypertensive drugs and insulin therapy, and to have longer diabetes duration than those with normal %MAP in the normal ABI group (all P < 0.001), but not in the low ABI group. Continuous data are presented as the mean ± SD, and categorical data are presented as numbers (percentages).
*: low ABI was de ned as an ABI value ≤ 0.90 and normal ABI > 0.90; high %MAP was de ned as a %MAP > 45% and normal %MAP ≤ 45%. # P: denotes a signi cant difference across the four subgroups. † P: post hoc analysis between two subgroups in patients with normal ABI; ‡ P: post hoc analysis between two groups in patients with low ABI. %MAP = percentage of mean arterial pressure, ABI = ankle-brachial index, ACE = angiotensin-converting enzyme, ARB = angiotensin II receptor antagonist, baPWV = brachial-ankle pulse wave velocity, BMI = body mass index, BP = blood pressure, CVD = cardiovascular disease, DPP4 = dipeptidyl peptidase-4, eGFR = estimated glomerular ltration rate, HbA1c = hemoglobin A1c, HDL = high-density lipoprotein, SD = standard deviation, SGLT2 = sodium glucose cotransporter 2, UACR = urine albumin-to-creatinine ratio. Continuous data are presented as the mean ± SD, and categorical data are presented as numbers (percentages).
Over median follow-up of 22.9 months (interquartile range: 13.2-29.7 months), 266 (4.8%) of the 5569 enrolled patients died. The incidence rates of mortality were 2.0 deaths/100 person-years in the normal ABI and normal %MAP subgroup, 5.0 deaths/100 person-years in the normal ABI but high %MAP subgroup, 4.8 deaths/100 person-years in the low ABI but normal %MAP subgroup, and 8.3 deaths/100 person-years in the low ABI and high %MAP subgroup, respectively; the difference. The survival rates were signi cantly different across these four subgroups (log-rank test P < 0.001, Fig. 2).
To evaluate how addition of the %MAP result to ABI affected prediction of all-cause mortality, we analyzed the increments in the AUC. We used ABI as the standard risk factor, AUC increased signi cantly from 0.

Discussion
The main nding of our study was that high ankle %MAP acted synergistically with low ABI to improve prediction of all-cause mortality in patients with type 2 DM. Using a combination of the two indices, ABI ≤ 0.90 and %MAP > 45%, predicted an approximately two-fold mortality risk than ABI > 0.90 and %MAP ≤ 45%. These results support our previous study which showed that high %MAP was a signi cant predictor of all-cause mortality in subjects with normal ABI [12]. A recent study has also shown that %MAP was associated with cardiovascular mortality in patients receiving hemodialysis [16]. The strength of the present study is that we demonstrated the synergistic effect of ABI and %MAP for prediction of mortality in a large sample of more than 5000 patients with type 2 DM.
Low ABI indicates a reduced systolic blood pressure at the ankle relative to that in the brachial artery, and this suggests partial occlusion of the ankle arteries [17]. Since the systolic blood pressure will be elevated in a non-compressible artery at the ankle, a false negative PAD diagnosis may occur when ABI alone is used for screening [18,19]. In the study by Wukich, et al., 42.7% of patients with DM and con rmed PAD had normal ABI value [20].
The %MAP represents the percentage difference between the mean and maximum amplitude of the ankle pulse volume waveform [11]. An occluded artery with a atted waveform will result in an increased in %MAP value [10]. Therefore, the pulse volume recording at the ankle might be a sensitive indicator of an occlusive artery with a non-compressible pattern, which is frequently observed in patients with DM [21].
The prevalence of PAD is increasing worldwide, and DM is an important risk factor for PAD [22,23]. Most patients with PAD are asymptomatic, but they have elevated risk for mortality [22][23][24]. In Taiwan, annual screening for foot complications is recommended in the clinical guidelines and in the P4P program for patients with DM [13,25]. In previous studies that have used the cutoff value of ABI ≤ 0.90, the prevalence of PAD in type 2 DM was about 10.0% in patients with a mean age of 63 years in Taiwan, 10.4% in Malay patients (mean age, 63 years) who living in Singapore, and 9.5% in patients (age > 40 years) in the US [26][27][28]. According to the real-world database, PAD was reported in 18.7% of patients with type 2 DM (mean age, 65 years) in the UK and in 13.6% patients with type 2 DM (mean age, 66 years) in the US [29,30]. In the present cohort, PAD prevalence was 8.4% when ABI ≤ 0.90 was the only criterion used, but increased to 17.4% when the combination of ABI ≤ 0.90 and %MAP > 45% were used. In the Taiwan National Health Insurance database, less than 2.2% patients with DM and age ≥ 65 years have a diagnosis of PAD, indicating that the condition is greatly underdiagnosed in clinical practice [31]. Thus, using ABI along with the automatically reported ankle %MAP is an effective and convenient method for PAD screening and for prediction of mortality [9,12].
The risk factors for abnormal ABI have been well investigated, but the risk factors for high %MAP are still not speci ed [32,33]. In the present study, the risk factors signi cantly associated with %MAP in both the different ABI groups, included age, CVD history, BMI, HbA1c, eGFR, UACR, baPWV, use of antiplatelet agents, type of oral antihyperglycemic drug, and type of hypertensive drug (Table 1). However, we did not include all cardiovascular risk factors in the present study; for example, a higher HbA1c variability has been previously reported to be associated with a higher %MAP [15]. Furthermore, this study has several limitations. First, all participants were from a single teaching hospital, and the results may not be generalizable to all population with type 2 DM. Second, this was a retrospective study and so we could not control the risk factors and treatments received during the follow-up period. Third, the cutoff value of 45% for %MAP is based on the ndings of previous studies [12]; we did not assess the normal range of %MAP in the present study.
In conclusion, the use of %MAP along with ABI appears to improve prediction of all-cause mortality in patients with type 2 DM. The %MAP is automatically reported during ABI measurement and so can conveniently be used for improving prognosis prediction in clinical practice.

Declarations
Ethics approval and consent to participate The study complied with the Declaration of Helsinki, and was approved by the Institutional Review Board of Taichung Veterans General Hospital, with a waiver of the need for informed consent.

Consent for publication
Not applicable.

Availability of data and materials
The datasets used and/or analyzed during the current study are available from the corresponding author on reasonable request.

Competing interests
The authors declare that they have no competing interests.

Funding
This work was supported by grants from Taichung Veterans General Hospital, Taichung, Taiwan (TCVGH-1093505D) and National Health Research Institute (grant number NHRI-EX109-10927HT). The funding bodies had no role in the decision to submit the manuscript for publication.
Author contributions YL participated in the data collection and writing of the manuscript. WS contributed to the study design. IL contributed to the study design, the data collection, interpretation of the data, and revision of the manuscript. IL is the guarantor of this work and 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.