Study design and sample size

This study was part of on-going prospective cohort investigation of peripheral vascular disease patients aimed at assessing risk predictors of peripheral vascular disease presence and outcome [15],[16]. Monte-Carlo simulations suggest that a multivariate regression model is adequately powered when using 10 outcome events per degree of freedom of the predictor variable [17]. We estimated that mortality at one year would be ~10% and planned to adjust for up to 15 variables in our regression model. This suggested that ~1600 patients would provide sufficient power to examine the association of diabetes with mortality.

Patients

Patients were recruited from in and out-patient vascular services at The Townsville Hospital, The Mater Hospital Townsville and The Royal Brisbane and Women’s Hospital, Australia. Patients with all types of peripheral vascular disease were considered for inclusion as previously described [15],[18]. Inclusion criteria for the current study included a diagnosis of PAD, assessment of fasting blood glucose and at least one follow-up assessment as an in or out-patient. This study was conducted in accordance with the Declaration of Helsinki. Ethical approval for the study was granted by the local Institutional Ethics Committees at The Townsville Hospital, The Mater Hospital Townsville, The Royal Brisbane and Women’s Hospital and James Cook University (61/05, MHS2006-01, H2196, 2007/004, 12/QTHS/202, MHS20140114-01, H5206, 13/ QTHS/125). Participants provided written informed consent for inclusion.

Definitions and diagnosis of PAD

The current study included patients with occlusive or aneurysmal disease of their peripheral arteries. Presenting complaints included asymptomatic carotid stenosis, mild lower limb or upper limb peripheral athero-thrombosis, aneurysm of the aorta or peripheral arteries, symptomatic carotid artery stenosis and critical lower limb ischaemia, as previously described [15],[16],[18].

Assessment of fasting blood glucose and diabetes

Definitions of other risk factors

Hypertension was defined by a history of high blood pressure or receiving treatment to reduce blood pressure [15],[16],[18]. Smoking status was classified as ever and never smokers [15],[16],[18]. Coronary heart disease (CHD) was defined by a history of myocardial infarction, angina or treatment for coronary artery disease [15],[16],[18]. Estimated glomerular filtration rate (eGFR) was calculated using the Chronic Kidney Disease-Epidemiology Collaboration group (CKD-EPI) formula since we have previously found this to be most accurately associated with complications in PVD patients [16].

Medications

In addition to anti-diabetic drugs, each patient’s medications were recorded including whether the participants were prescribed a statin, aspirin, another anti-platelet agent, a beta-blocker, a calcium channel blocker (CCB), an angiotensin converting enzyme (ACE) inhibitor, an angiotensin receptor blocker (ARB) and frusemide [15],[16],[18].

Follow-up

Patients were followed up through attendance at out-patient clinics and /or as an in-patient as part of their normal medical care as previously described [15],[16],[18].

Recording of outcome data

The primary outcome was mortality. The secondary outcome was requirement for a peripheral artery intervention. Outcome data was recorded during clinical reviews on prospectively defined report forms. Charts and hospital electronic records of all patients were reviewed by a vascular specialist or clinical researcher. For peripheral artery intervention assessments, patients were censored at the time of the first intervention or at the date of last in/out patient review or death if no intervention was required.

Statistical analyses

Nominal data are presented as numbers and percentages. The association of diabetes categories with the clinical presentation and risk factors of the patients was assessed using Kruskal Wallis and chi-squared tests. The associations of diabetes categories with death and requirement for peripheral artery intervention were assessed using Kaplan Meier estimates, log rank test and Cox proportional hazard analyses. Cox proportional hazard analyses were adjusted for varying combinations of risk factors (age, sex, hypertension, ever smoking, CHD, presenting complaint, statin prescription, aspirin prescription, other anti-platelet prescription, beta blocker prescription, CCB prescription, ACE inhibitor prescription, ARB prescription, frusemide prescription and eGFR) across four different models. These covariates were included as they are recognised determinants of outcome for patients with cardiovascular disease.

For Cox proportional hazards analyses, diabetes categories were defined as indicators in the following order: No diabetes; impaired fasting glucose; non-medicated diabetes; diabetes prescribed oral hypoglycaemics only; and diabetes prescribed insulin. Presenting peripheral artery disease complaint was defined as indicators in the following order: asymptomatic carotid stenosis; mild lower limb or upper limb peripheral athero-thrombosis; aneurysm of the aorta or peripheral arteries; symptomatic carotid artery stenosis; and critical lower limb ischemia. Age and eGFR were included in regression models as continuous numbers. Binary variables were defined as present or absent.

Risk factors of the cohort

1637 PAD patients were included in the current study. The presenting complaints of the patients included aortic or peripheral aneurysms (n = 707; 43.2%), mild limb peripheral artery disease (n = 381; 23.3%), symptomatic carotid artery stenosis (n = 226; 13.8%), critical limb ischaemia (n = 184; 11.2%) and asymptomatic carotid artery stenosis (n = 139; 8.5%). The median age of the patients was 71 years and approximately three quarters were male (Table 1). Risk factors for cardiovascular disease included a history of hypertension and smoking in 77% and 82%, respectively. Forty eight percent of patients had a past history of diagnosis or treatment of CHD.

Relationship between risk factors, presentation, impaired fasting glucose, diabetes and its management

At entry 372 (22.7%) patients were receiving treatment for type 2 diabetes by oral hypoglycaemics alone (n = 296; 18.1%) or insulin (n = 76; 4.6%). Fasting blood glucose concentrations (≥7 mM) suggested diabetes in a further 150 (9.2%) patients who were on no medications for diabetes. A further 460 (28.1%) patients had fasting blood glucose concentrations between 5.6 and 6.9 mM which we defined as impaired fasting glucose. The other 655 (40.0%) patients had a normal fasting blood glucose concentration (<5.6 mM). Diabetes category was significantly associated with age, hypertension, CHD, eGFR, and prescription of statins, aspirin, beta blockers, CCBs, ACE inhibitors, ARBs and frusemide (Table 1). Patients with diabetes prescribed insulin were younger, had lower eGFR and more commonly had a history of hypertension and CHD than those with no diabetes (Table 1). Patients with diabetes prescribed insulin were also more likely to be prescribed statins, beta blockers, CCBs, ACE inhibitors and frusemide than those without diabetes (Table 1).

No difference in the prevalence of asymptomatic or symptomatic carotid stenosis was observed between the diabetes categories. Patients with non-medicated diabetes were more likely to present with aortic or peripheral artery aneurysms. Patients with diabetes who were prescribed insulin more frequently presented with mild limb peripheral artery disease and critical limb ischaemia.

Relationship between impaired fasting glucose, diabetes and its management with mortality

Median follow-up time was 1.7 (interquartile range 0.5-4.1) years. Three hundred and 17 (19.4%) patients died during the follow up period. Kaplan Meier analyses suggested that mortality was greatest in patients with diabetes prescribed insulin who had an estimated mortality of 20.9% at 2 years (Figure 1). Estimated mortality at 2 years in these patients was greater than that of patients with no diabetes (10.6%; p = 0.004), impaired fasting glucose (9.6%; p = 0.006) and diabetes prescribed oral hypoglycaemics only (6.3%; p = 0.008). Estimated mortality at 2 years in patients with non-medicated diabetes was similar to that of patients with diabetes prescribed insulin (20.5%; p = 0.611). Cox proportional hazard analyses suggested that, depending on the other risk factors adjusted for, mortality increased 1.6-1.7 fold in patients with non-medicated diabetes compared to patients with no diabetes (Table 2). Similarly patients with diabetes prescribed insulin had an increased mortality of 2.0-2.9 fold (Table 2). Patients with impaired fasting glucose and diabetes prescribed oral hypoglycaemics only had similar mortality to that of patients with no diabetes (Table 2).

Relationship between impaired fasting glucose, diabetes and its management with peripheral artery interventions

During follow up, 963 (58.8%) patients required a peripheral artery intervention. Patients with non-medicated diabetes were more likely to require a peripheral artery intervention during follow-up (Figure 2). Kaplan Meier analyses suggested that by two years after entry 69.3% of these patients had required a peripheral artery intervention. This intervention requirement was significantly greater than for patients with no diabetes (57.4%; p = 0.008) and those with diabetes prescribed oral hypoglycaemics only (57.1%; p = 0.018), and almost significantly higher than for patients prescribed insulin (60.1%; p = 0.063). The intervention rate in patients with impaired fasting glucose (63.9%; p = 0.210) was similar to that of patients with non-medicated diabetes (Figure 2). Cox proportional hazard analyses suggested that the requirement for peripheral artery intervention was approximately 1.3-fold greater in patients with non-medicated diabetes compared to those with no diabetes in different models which included adjustment for other risk factors (Table 3). The requirement for peripheral artery intervention was similar in patients with impaired fasting glucose, diabetes prescribed oral hypoglycaemics only, diabetes prescribed insulin and no diabetes (Table 3).

Study strengths and limitations

The current study has a number of strengths and weakness. Strengths include the large sample size, the adjustment for many potential confounding factors and the large range of different peripheral artery disease complaints studied. Weaknesses include the absence of information on oral glucose tolerance tests, dietary management of diabetes or HbA1C. The overall follow-up of the patients was also relatively short at a median of approximately 2 years. Importantly, we were only able to collect clinical characteristics of the included patients at the time of recruitment. Hall and colleagues recently reported that altering anti-glycaemic medication regime during a 3.5 year follow-up significantly influences cardiovascular outcomes in patients with diabetes [31]. However, the design of the current study limited the ability to accommodate changes in risk factors which may have influenced cardiovascular outcome in our patient cohorts. Finally, the cause of death was not available and it is therefore possible that other factors may have contributed to overall mortality in the current study, complicating interpretation of our findings. Future studies are required to specifically investigate the relationship between diabetes presentation, management and cardiovascular mortality in the northern Australian PAD population.

Study conclusions

The findings of this work have important implications. Firstly dysglycaemia is very common in PAD patients. Screening newly referred patients for diabetes and where appropriate referring them to a diabetes specialist would appear appropriate as previously suggested [38]. Secondly, patients with fasting blood glucose ≥ 7 mM not currently receiving medication for diabetes are at increased risk of major adverse events and may benefit from more intensive management. Patients with established diabetes receiving insulin are at particularly high risk of major adverse events which should be considered in decisions about surgical interventions.