Risk factors for diabetic foot complications among patients with type 2 diabetes in Austria–A registry‐based retrospective cohort study

Abstract Aims Diabetic foot complications, a serious consequence of diabetes mellitus, are associated with a tremendous burden on both individual patients and health care systems. Since prevention strategies may reduce the incidence of this complication, identification of risk factors in large longitudinal studies is essential to optimize early detection and personalized screening of patients at increased risk. Materials and methods We conducted a registry‐based retrospective cohort study using data from 10,688 patients with type 2 diabetes mellitus aged ≥18 years. Cox regression models were used to identify risk factors for foot complications while adjusting for potential confounders. Results We observed 140 diabetic foot complications in our patient cohort. The multivariate Cox regression model revealed neuropathy, peripheral arterial disease and male gender as being positively associated with foot complications. The same effect was detected for nephropathy in the time >10 years after T2DM diagnosis. For higher age at diagnosis and use of insulin, however, a negative association was retrieved. Conclusion Male gender and several diabetes‐related comorbidities were identified as risk factors for subsequent initial foot complications in patients with type 2 diabetes mellitus. These findings suggest that personalized early detection of patients at increased risk might be feasible by using information on demographics, medical history and comorbidities.


| INTRODUC TI ON
Diabetes mellitus (DM) represents one of the major public health concerns worldwide. Global estimates indicate that 463 million people aged 18-99 years are currently affected by DM and this number is projected to increase to over 700 million people by 2045. The rising prevalence of diabetes can be explained by ageing, population growth and lifestyle alterations. [1][2][3] Diabetic patients face a high The purpose of this study was to determine which factors are associated with an increased risk for subsequent diabetic foot complications in patients with type 2 diabetes mellitus (T2DM) using real-world data from a large cohort of patients. To this end, data from the Diabetes Registry of Tyrol (DRT) were used. This is one of the largest diabetes registries in Europe and represents all hospitals and several outpatient practices in one out of nine federal states in Austria.

| Study population
The DRT was established in 2006 aiming to measure and improve the quality of care for diabetic patients in Tyrol. Data are collected in ten participating hospital sites covering all hospitals in Tyrol as well as eight outpatient practices of specialists of internal medicine. All patients with newly diagnosed T1DM, T2DM and gestational diabetes mellitus, but also prevalent diabetes patients who attend an outpatient department, are collected in the DRT. After patients undergo a comprehensive clinical assessment during the first visit at one of the participating sites, they are invited to return for quarterly visits.
Until 2019, data on more than 24,000 diabetic patients were collected within this registry. 22,23 The registration is performed within the hospital information systems, which incorporate demographic data, diabetes-related clinical and biochemical parameters and data on late complications related to diabetes. 23 After pseudonymization, the data are transferred to the DRT. This allows linkage of data for a specific patient registered in different departments and guarantees data confidentiality.

| Case identification and definitions
The following three tests are used for establishing the diagnosis of diabetes mellitus: (i) ≥126 mg/dl (≥7.0 mmol/L) of plasma glucose in a fasting state, (ii) ≥200 mg/dl (≥11.1 mmol/L) either in a nonfasting state or two hours after oral intake of 75 grams of glucose or (iii) a HbA1c value of ≥6.5%. An unequivocal diagnosis requires either two different tests, or one test performed on two separate days. 24 In addition to the year of diabetes diagnosis and the patient's demographic data, height and smoking status are documented at entry in the registry. At each subsequent visit, data on weight, physical activity (defined as at least 2.5 h per week) and current diabetes treatment are updated, and the values for blood pressure and HbA1c are assessed.
Retinopathy is diagnosed according to the guidelines provided by the Austrian ophthalmologist association, which are based on the international retinopathy severity scales. 25  (4) no present or prior foot ulcer at the first visit of a patient in the registry; (5) patients with complete data sets concerning relevant demographic and clinical data that were considered potential risk factors for DF. Thereby, patients with other types of diabetes (e.g. T1DM or gestational diabetes) and patients younger than 18 years at diagnosis were excluded. Data collected between 2006 and 2019 were included in the analysis.
As preceding risk factors for the initial development of DF were assessed, only data collected at visits before the year of DF diagnosis were included. The variables collected at each visit were aggregated to yield data values on patient level. For HbA1c, systolic and diastolic blood pressure and weight, the mean values were used. The patient's smoking status at time of diagnosis was classified as 'active smoker' or 'ex-or never-smoker'. Physical activity was considered applicable if the patient responded positively to being active for ≥2.5 h per week at least at one visit collected in the registry.
Based on the commonly used classification of age groups within the Diabetes Registry Tyrol, the patient's age at diagnosis was categorized into three groups: ≤50 years, >50 and ≤70 years, and >70 years of age. Hypertension was considered evident if the mean of all systolic or diastolic blood pressure readings prior to DF development or until last visit was at or above 140 or 90 mmHg, respectively. The body mass index (BMI) was calculated, and obesity was defined as a BMI of 30 kg/m² or more in accordance with the definition by the World Health Organization (WHO). 28 For each patient, the mean of all HbA1c values was calculated and used to assign the respective patient to one of three groups: <6.5%, 6.5%-9.0%, and >9.0%. The boundaries were chosen due to different approaches to antihyperglycaemic treatment as specified in the guidelines of the Austrian Diabetes Association, 29 The use of insulin or insulin analogues was assessed as potential risk factor for DF. Thereby, a distinction was made between patients for whom this form of treatment was documented at least at one visit, and patients who have never been treated in this way. Additionally, it was assessed, if the patients participated in an educational programme, and if the conduction of foot inspections was documented for at least one visit.
The following pre-existing late complications were considered as potential risk factors for DF and assessed as binary variables: nephropathy, retinopathy, neuropathy, myocardial infarction, stroke, peripheral arterial disease and coronary bypass/percutaneous transluminal coronary angioplasty (PTCA).
Since the need for lower extremity amputation usually follows a preceding DF complication such as a foot ulcer, the year of the first documented amputation was usually after the year of the first DF diagnosis. However, in rare cases in which no DF diagnosis preceded a present year of amputation, the year of amputation was defined as the year of the DF diagnosis.

| Statistical analysis
While categorical variables were reported in proportions, continuous variables were described as means ± standard deviations. For the comparison of variables between the cohort of patients with DF and the cohort of patients without DF, Chi square (χ²) and Mann-Whitney U tests were used for categorical and continuous variables, respectively. p-Values <.05 were considered statistically significant.
Cox regression analysis was used to analyse the association between potential risk factors and subsequent DF complications. To this end, the time from initial diabetes diagnosis to incidence of DF was assessed, while the end of follow-up was considered the cen-

| Patient characteristics
In total, 10,688 out of 23,593 patients fulfilled the predefined inclusion criteria and were included in the analysis. The flow diagram of the patient selection is depicted in Figure 1.
The overall mean (±SD) age at diagnosis was 63.21 ± 12.58 years, and 44.3% were female. Overall, 140 DF events occurred during a mean follow-up period of 9.75 years.
Baseline characteristics of the study population are depicted in

| Risk factors for DF
In the univariate Cox model, the most strongly associated variable with DF was peripheral arterial disease with a hazard ratio of 4.50  To evaluate the robustness of the multivariate Cox model developed by means of a backward elimination approach, an alternative model was built using a forward approach. In this second multivariate model, the same set of independent risk factors for DF development was identified, highlighting the robustness of the analysis.

| DISCUSS ION
Given the tremendous personal and financial burden associated with foot complications in patients with T2DM, the need for better understanding of this late complication is critical for its prevention. We have performed a large registry-based retrospective cohort study in a population of T2DM patients without previous foot complications. In this study, a prevalence of DF of 1.31% was reported, a value that lies below the ranges previously reported for European countries (1.7-4.8%). 31 However, this may derive from the fact that all patients with foot complications present at the first visit recorded in the registry were excluded from the analysis in order to take only information prior to DF diagnosis into account.
In the time-dependent multivariate Cox regression model, neuropathy, peripheral arterial disease, nephropathy after 10 years after  20,32 Peripheral neuropathy was identified as the strongest risk factor for foot complications in this studied patient cohort. These findings are consistent with other studies that were using age and gender adjusted multivariate logistic regression models and in which the association has also been reported. 19,33,34 This strong association is likely to reflect the high level of nerve damage present in the lower extremities of diabetic patients, which highlights the need for increased screening for lower extremity nerve defects in this patient population. Peripheral arterial disease was the second strongest factor associated with DF in this study. Given the fact that ischemia causes tissue damage and leads to poor wound healing, peripheral arterial disease is known as an important risk factor in the pathogenesis of foot complications. 8,35 The crucial role of peripheral arterial disease as a risk factor for DF development has been highlighted by various previous studies. 33,34,36 Other diabetes-related complications, namely nephropathy and previous myocardial infarction or coronary bypass/PTCA were significantly associated with foot complications in the univariate model.
However, when adjusting for other potential risk factors in the timedependent multivariate model, the association remained significant times higher in such patients later than 10 years after the diabetes diagnosis. Nephropathy might be anticipated as a risk factor for DF due to the common physiological origin of microvascular late complications. 4 However, in previous studies, the role of nephropathy as a risk factor remained inconsistent: While several groups identified a positive association, 34,36,37 no association was retrieved in other studies. 19,38,39 To the knowledge of the authors, however, there is no study available in which the association between nephropathy and DF was analysed in separate time intervals.
Male gender was identified as a strong predictor for DF in the multivariate Cox model. The identified risk of DF development is almost doubled in male patients compared to female patients. A similar effect was shown consistently in various other studies. 19,33,37 This effect may be explained to some extent by the higher foot pressure found in male patients, probably due to higher mean height in men compared to women. 40 In addition, women are known to be more active in terms of self-care and preventive care concerning diabetic foot lesions, whereas men show a more passive attitude. 41 Higher age has been identified as having a protective effect on the DF development (i.e. with increasing age at diagnosis, the hazard of foot ulcer was found to decrease). This effect was not only seen when comparing >70 years and ≤50 years of age at diagnosis, but also when age between 51 and 70 years was compared to ≤50 years.
Additionally, other groups reported a negative association between higher age and different endpoints related to foot complications:  43 However, as highlighted in two recent systematic reviews on risk factors for DF development, the results on the potential association between age and DF are highly contradictory. 17,18 While in several studies, a positive association was identified between higher age and various endpoints such as foot ulcers and lower extremity amputations, 19,38,39 other groups did not find any association. 33,[35][36][37] There are different hypotheses that aim to explain the protective effect of age at diagnosis: A possible explanation is that older patients with severe courses of disease, that render them immobile, might be underrepresented in the registry. This selection bias would lead to the possibility that the older patient groups represented in the registry are those who are healthier. 42 Dekker et al. furthermore hypothesized that younger patients are more physically active compared to older patients and are therefore more prone to traumatic situations which increases the risk of foot ulcers. 34 However, further studies are needed to gain more detailed insights on the relationship between age and DF development.
A protective effect was detected concerning the use of insulin or its analogues. This finding is not in line with several studies that identified a positive association between insulin use and subsequent foot complications 19,37,44 or did not find any association. 45 However, in the cohort study on hand, potential risk factors have only been analysed prior to the development of a foot complication, whereas in several other studies, patients with DF were compared to patients without DF. Therefore, a positive association could derive from the fact that patients who are already receiving treatment for foot complications are more likely to be insulin users. 19,44 In addition, insulin treatment prior to DF development might improve glycaemic control, thereby preventing subsequent late complications.
Nonetheless, further prospective studies are required to investigate this association in more detail.
Although the level of statistical significance was not reached in the multivariate analysis, a hazard ratio as high as 3.76 for HbA1c values >9% compared to <6.5% in the time after 15 years after the diabetes diagnosis might be of high clinical relevance. A positive association between HbA1c levels and foot complications was shown by several groups, 19,36,39 while others could not find any association. 37,45 Our data suggests that since glycaemic control can be altered by lifestyle changes and/or treatment modalities, improving glycaemic control might be beneficial to reduce the risk of subsequent DF events. Another modifiable risk factor, that is, the patient's BMI, showed a hazard ratio of 1.38 in the univariate model, and although this factor was not statistically significant, it might still be clinically relevant. Various studies on the association between BMI and subsequent foot complications have come to discordant results ranging from positive associations to no effects at all to a protective effect of higher BMI. 18 When considering the same level of mobility, there is higher pressure on the lower extremities in obese patients compared to patients with a lower BMI. This is proposed to be linked to more frequent DF events. 40 However, there is still a lack of consensus regarding the relationship between BMI and foot complications. 46 Additional studies are thus required to clarify to this relationship. From a public health perspective, a reduction in overweight is anticipated to be beneficial for the patients' overall health independent of potential foot complications.
Our study is characterized by several strengths and weaknesses.
The main strength of this multicentre study lies in its large sample size with a wide range of demographic, clinical and behavioural data.
The DRT is a region-wide registry covering all ten hospitals and eight outpatient practices in Tyrol. Further strengths include data collection by specialized personnel and the fact that the cohort is derived from a large web-based electronic registry focussing on diabetes and its complications with regular follow-ups and data validation.

| CON CLUS ION
We have conducted a large retrospective cohort study to investigate the association between various potential risk factors and subsequent initial development of DF complications in patients with T2DM. Our study revealed statistically significant associations of DF with neuropathy, peripheral arterial disease, nephropathy and use of insulin or insulin analogues. Moreover, demographic characteristics such as age at diagnosis and gender were shown to play an important role in the risk for DF. We therefore suggest that readily available information on the patients' demographic data, medical history and comorbidities may facilitate personalized screening. Large longitudinal studies are needed to investigate whether the reduction in existing risk factors leads to a decrease in the number of subsequent foot complications in patients with T2DM.

ACK N OWLED G EM ENTS
We thank the clinical staff of all hospitals and outpatient practices involved in the DRT and the employees of the Department of Clinical Epidemiology (Tirol Kliniken, Innsbruck, Austria) for collecting the data and their help in the conduct of this study.

CO N FLI C T S O F I NTE R E S T
All authors declare that there is no conflict of interest.

AUTH O R CO NTR I B UTI O N S
SR developed the study protocol, performed the statistical analysis, and wrote the manuscript. BR was involved in the conduction of the statistical analysis and the design of the tables and figures. HS and ML contributed their clinical expertise to writing the manuscript. WO was involved in the study design, the conduction of the analysis and writing the article. All authors have read and approved the final manuscript.

E TH I C A L A PPROVA L
The present study was approved by the Ethics Committee of the Medical

DATA AVA I L A B I L I T Y S TAT E M E N T
The data that support the findings of this study are available from the