Participants and baseline procedures
This was a prospective study, the RIO-T2D, with 668 individuals with type 2 diabetes enrolled between August 2004 and December 2008 and followed until December 2017 in the diabetes outpatient clinic of our tertiary-care university hospital. All participants gave written informed consent, and the local ethics committee had previously approved the study protocol. The characteristics of this cohort, baseline procedures and diagnostic definitions have been detailed elsewhere [34,35,36,37]. In brief, inclusion criteria were adults with type 2 diabetes, up to 80 years old, with any microvascular or macrovascular complication, or with at least two other modifiable cardiovascular risk factors. Exclusion criteria were morbid obesity, advanced renal failure (serum creatinine >180 μmol/l or eGFR <30 ml min−1 [1.73 m]−2) or the presence of any serious concomitant disease limiting life expectancy. For this specific analysis, individuals with clinical PAD, defined by a history of typical intermittent claudication, limb revascularisation procedures, foot ulceration or lower-extremity amputations, were excluded.
All participants were submitted to a standard baseline protocol that included a thorough clinical examination (including ABI measurement), laboratory evaluation and 24 h ambulatory BP monitoring. Diagnostic criteria for diabetes-related chronic complications have been detailed previously [34,35,36,37]. In brief, coronary heart disease was diagnosed by clinical and electrocardiographic criteria, or by positive ischaemic stress tests; and cerebrovascular disease by history and physical examination. A diagnosis of nephropathy needed at least two albuminuria measurements of at least 30 mg/24 h or a confirmed reduction in the GFR (eGFR ≤60 ml min−1 [1.73 m]−2 or serum creatinine >130 μmol/l). Peripheral neuropathy was determined by clinical examination (knee and ankle reflex activities, and foot sensation using the Semmes–Weinstein monofilament, vibration with a 128 Hz tuning fork and pinprick and temperature sensations) and neuropathic symptoms were assessed using a standard validated questionnaire . Clinic BP was measured three times using a digital oscillometric BP monitor (HEM-907XL, Omron Healthcare, Kyoto, Japan) with a suitably sized cuff on two occasions 2 weeks apart at study entry. The first measure of each visit was discarded and the BP used was the mean of the last two readings from each visit. Arterial hypertension was diagnosed if mean SBP was 140 mmHg or higher or diastolic BP (DBP) was 90 mmHg or higher, or if anti-hypertensive drugs had been prescribed. Ambulatory BP monitoring was conducted in the following month using Mobil-O-Graph version 12 (Dyna Mapa, Cardios, Sao Paulo, Brazil), and average 24 h SBP and DBP values were registered [36, 37]. Laboratory evaluations included fasting glycaemia, HbA1c, serum creatinine and lipids. Albuminuria was evaluated in two non-consecutive sterile 24 h urine collections.
After resting supine for at least 5 min, two BP readings were taken sequentially from each brachial and posterior tibial artery (total of eight measurements: four brachial and four tibial) using the same digital oscillometric BP monitor (HEM-907XL, Omron Healthcare) and were validated for ABI measurement [38, 39]. The ABI was calculated as the lowest tibial BP in either leg divided by the highest brachial BP in either arm . An ABI value of ≤0.90 was considered indicative of PAD. All ABI measurements were performed at baseline by a single independent examiner who was unaware of other clinical data.
Follow-up and outcomes assessment
The participants were followed regularly at least three times a year until December 2017 under standardised treatment. The observation period for each individual was the number of months from the date of the first clinical examination to the date of the last clinic visit in 2017 or the date of the first endpoint, whichever came first. The primary endpoints were the occurrence of any micro- or macrovascular outcomes. Macrovascular outcomes were total CVEs (fatal or non-fatal myocardial infarction, sudden cardiac death, new-onset heart failure, death from progressive heart failure, any myocardial revascularisation procedure, fatal or non-fatal stroke, any aortic or lower-limb revascularisation procedure, any amputation above the ankle and death from aortic disease or PAD), MACE (non-fatal myocardial infarction and stroke, plus all cardiovascular deaths) and all-cause and cardiovascular mortality . Microvascular outcomes were retinopathy development or worsening , renal outcomes  (new microalbuminuria development, new renal function deterioration [defined as doubling of serum creatinine to a value of ≥200 μmol/l, end-stage renal failure needing dialysis or death from renal failure] and a composite of the two) and peripheral neuropathy development or worsening . Retinopathy and renal outcomes were evaluated by annual examinations [36, 37], whereas peripheral neuropathy was evaluated during a second specific examination performed after a median of 6 years from the baseline examination . Peripheral neuropathy development or worsening was determined by a combination of signs (assessed using the Neuropathy Disability Score) and symptoms (assessed using the Neuropathy Symptom Score), as previously defined . The presence and severity of retinopathy was determined by a single retinal specialist, following the International Clinical Diabetic Retinopathy and Diabetic Macular Edema Disease grading .
Continuous data are described as means ± SD or as medians (interquartile range [IQR]). The baseline characteristics of individuals with an ABI of ≤0.90 and >0.90 were compared using unpaired t tests, Mann–Whitney U tests or χ2 tests, where appropriate. Kaplan–Meier curves of the incidence of cumulative endpoints during follow-up, compared with log-rank tests, were used to assess different incidences of outcomes between individuals with ABI values of ≤0.90 and >0.90. To assess the prognostic value of the ABI for each macro- and microvascular outcome, except for peripheral neuropathy, a time-to-event Cox analysis was undertaken. First, analyses were only adjusted for age and sex, and then further adjusted for other potential confounders and risk factors (diabetes duration, BMI, smoking, physical activity, diabetes treatment, arterial hypertension, number and classes of anti-hypertensive drugs in use, ambulatory 24 h SBP, presence of micro- and macrovascular complications at baseline, serum mean first-year HbA1c, HDL- and LDL-cholesterol levels, and use of statins and aspirin). Information on all of these covariates was obtained at baseline, except for HbA1c and HDL- and LDL-cholesterol, for which mean values obtained during the first year of follow-up were used. These results are presented as HRs with their 95% CIs. For peripheral neuropathy analyses, a multiple logistic regression model was used with the same statistical adjustments, except that height (instead of BMI) and the interval between the baseline and second neuropathy evaluations were included as adjusting covariates. These results are reported as ORs with their respective 95% CIs. In both analyses, the ABI was assessed as a continuous variable (with risks estimated for a 0.10 decrement in the ABI) and also as a categorical variable (with risks estimated for individuals with an ABI of ≤0.90 in relation to those with an ABI of >0.90). In separate analyses, individuals were further divided into three ABI subgroups (≤0.90, 0.91–1.00 and ≥1.01–<1.3), with the subgroup with ABI ≥1.01 considered as the reference.
To assess the improvement in discrimination performance after the addition of the ABI to the models, we used the C statistic (analogous to the area under the receiver operating characteristic curve applied to time-to-event analysis), compared by the method proposed by DeLong and colleagues  and the integrated discrimination improvement (IDI) index [43, 44]. The IDI is equivalent to the difference in discrimination slopes between models with and without the new variable, and its calculation is based on continuous differences in predicted risk in new and old models for individual participants with and without the outcome under study. Thus, the IDI is free from the dependence on empirical risk categories that is inherent to reclassification tables and can be used as an objective indicator of reclassification improvement. Both the absolute and the relative IDI were calculated. The relative IDI, reported as a percentage, facilitates interpretation of the IDI, and is defined as the increase in discrimination slope divided by the slope of the standard model including only traditional cardiovascular risk factors [43, 44]. In sensitivity and interaction analyses, interactions between the ABI and age (<65 vs ≥65 years), sex, diabetes duration (<10 vs ≥10 years), presence of micro- and macrovascular complications at baseline and glycaemic control (mean HbA1c <58.5 vs ≥58.5 mmol/mol; <7.5% vs ≥7.5%) were tested for all endpoints and whenever there was evidence of interaction (p < 0.10 for interaction term), the interaction term was kept within the whole model analysis and a further stratified analysis for that specific characteristic was performed. In addition, separate analyses were performed for each outcome after excluding those participants with this specific complication at baseline.
Finally, to determine the presence of possible reverse causality between the ABI and outcomes, separate analyses were performed excluding 35 individuals who had any of the endpoints (death or non-fatal CVEs) during the first 2 years of follow-up. Statistical analyses were performed using SPSS version 19.0 (SPSS, Chicago, IL., USA) and R version 3.4.1 (R Foundation for Statistical Computing, Vienna, Austria), and a two-tailed p value of <0.05 was considered significant.