Risk factors for diabetic foot complications in type 2 diabetes—A systematic review

Abstract Aims With increasing numbers of patients with type 2 diabetes mellitus (T2DM) worldwide, the number of associated diabetic foot complications might also increase. This systematic review was performed to summarize published data about risk factors for the diabetic foot (DF) syndrome in order to improve the identification of high‐risk patients. Materials and methods Six electronic databases were searched for publications up to August 2019 using predefined stringent inclusion and exclusion criteria. Results Of 9,476 identified articles, 31 articles from 28 different study populations fulfilled the criteria for our evaluation. The overall quality of the studies was good, and the risk of bias was low. There was large heterogeneity among the studies concerning study protocols and patient populations analysed. A total of 79 risk factors were analysed within this review. The majority of studies described a consistently positive association with different outcomes of interest related to DF for gender, peripheral neuropathy, retinopathy, nephropathy, poor glycaemic control, insulin use, duration of diabetes, smoking and height. For age, hypertension, dyslipidaemia and body mass index, the results remain inconsistent. Conclusion A most up‐to‐date literature review resulted in glycaemic control and smoking as the only amenable risk factors with a consistently positive association for DF. Due to the high personal and financial burden associated with DF and the large heterogeneity among included studies, additional longitudinal studies in large patient populations are necessary to identify more modifiable risk factors that can be used in the prediction and prevention of DF complications.

healthcare systems: in 2019, the global health care costs for diabetes totalled 760 billion US dollars for patients in the age group between 18 and 99 years. 1,2 Patients with diabetes face a high risk of developing serious adverse health conditions that shorten the life expectancy, lower the quality of life and increase medical care costs. 1,3 The diabetic foot (DF) syndrome is a serious diabetic late complication strongly related to diabetic neuropathy and peripheral artery disease. Tissue necrosis can result in a need for lower extremity amputation (LEA). 1 According to the International Working Group on the Diabetic Foot (IWGDF), DF is defined as: 'Infection, ulceration, or destruction of tissues of the foot of a person with currently or previously diagnosed diabetes mellitus, usually accompanied by neuropathy and/or peripheral arterial disease in the lower extremity'. 4 Around 25% of all patients with diabetes develop foot complications during their course of disease. 5 The condition constitutes a major cause for hospital admissions in people with diabetes, accounting for nearly 70% of all amputations conducted in the United States in 1997. 1,6,7 Moreover, diabetic foot ulcers (FU) and amputations make up the most expensive diabetic late complication in terms of hospital costs. 8 In the year after the first FU, the health expenditures for patients with diabetes with FUs are five times higher than for those without FUs and almost three times higher in the subsequent years. In 2007, one-third of all costs for diabetes were linked solely to foot complications. 9 Patients with diabetes suffering from FUs reveal a 10-20 times higher risk for amputation than subjects without diabetes, 10 and FUs are further associated with a higher mortality risk compared to those patients without foot complications. 11 Approximately 1% of all patients with diabetes have to undergo lower limb amputation in high-income countries, with the percentage being higher in low-and middle-income countries. 1 In addition, patients with a history of DF complications carry a higher risk of subsequent re-ulcerations. 12 DF conditions, especially with severe complications and the need for amputations, are one of the most serious and preventable diabetic late complications. Besides the efforts made on conducting regular foot examinations and the progress on risk classification systems, both prevention and early detection methods must be improved. 13,14 A further necessary aspect in the prevention would be the identification of risk factor profiles allowing to identify patients at high risk for foot disease.
A large number of articles have been published on this matter, however, with a large heterogeneity in the conducted studies and large differences in their quality. In contrast to more recent reviews on other aspects of the diabetic foot such as management and costs of this late complication, 15,16 only few reviews have been published on the associated risk factors, with the last publication in 2012. 17 Both the presentation of results and the number of published articles since the last published review on risk factors for diabetic foot complications justify a most up-to-date systematic review, which was designed to identify and characterize the published risk factors associated with the DF in type 2 diabetes mellitus (T2DM), which comprises approximately 90%-95% of all patients with diabetes. 18 The results of the review should on the one hand guide physicians, researchers, patients and other interested parties in the identification of patients at high risk of developing DF complications and on the other hand identify risk factors that can serve as starting points to be tackled in order to reduce this risk. | 3 of 32 ROSSBOTH eT al. and decided with the corresponding author. In these steps, studies that did not fit the aforementioned inclusion and exclusion criteria were removed from further analysis (see Figure 1). The reference sections of included studies were checked in order to identify potential studies, which had been missed earlier and are relevant. Furthermore, if more than one publication analysed data from the same study or database, it was checked whether the subpopulations and/or risk factors differed between the publications, and only if this was the case, more than one publication was included from the same source of data. Otherwise, the most recent publication would have been included. After the final number of eligible studies has been identified, the publications were summarized in line with the approach published by Drinkwater et al, who performed a well-structured, comprehensive, and easily understandable systematic review on risk factors for cataract in patients with T2DM. 20 Due to the large clinical and methodological diversity of the included studies (concerning, eg patient populations, outcomes and study designs), the conduction of a systematic review was more reasonable than the performance of a meta-analysis. 21 Important characteristics and data from the eligible studies were brought together in tabular forms. The information entered included author and year of study, country, study design, study name, patient characteristics (sample size, number of events, baseline age at study entry, proportion of T2DM, proportion of female patients, diabetes duration at time of development of outcome, follow-up time), potential conflicts of interest, methods and limitations, results from multivariate analyses as well as the covariates included in the models. The quality of included studies was assessed using the Newcastle-Ottawa

Quality Assessment Forms for Cohort Studies and Case-Control
Studies, 22 with a median follow-up time of 3 years chosen to be sufficient for outcome question 2 in case of cohort studies. The risk of bias was assessed for each included publication using the Cochrane handbook guidelines. 23 In the following sections, for reporting effects for a specific potential risk factor we use the wording positive or negative association or relationship synonymously for statistically significant effects only. In addition, we use the notation consistent association if only positive effects and null effects or only negative effects and null effects have been reported and inconsistent association if both positive and negative effects have been reported.  were assessed for eligibility via screening of title, abstract and/or full text. A final number of 31 articles were included in the analysis (see Figure 1). 24 The mean duration of diabetes ranged from 3.7 years 38 to more than 13 years 25,28 in the different patient populations; however, this value was not stated in nine of the 31 articles. 24,26,34,41,45,47,48,50,52 The mean follow-up time in longitudinal studies varied between one year 24,39 and 13 years. 44 The methodological aspects and the corresponding limitations of the particular studies are summarized in Table 2  The findings of the single publications were brought together in Table 4 to build an overview of the associations that have been shown for the single risk factors across all included publications. In total, the relationship between 79 different risk factors and the five previously defined outcomes has been studied. Apart from male gender, peripheral neuropathy (PN), retinopathy, nephropathy, poor glycaemic control, insulin use, duration of diabetes, smoking and height, for all of which a positive association with the outcome of interest was shown, the results for the other risk factors showed higher discordances. A total of 41 risk factors were each analysed in one study only.

| RE SULTS
The assessment of the quality of the included studies using the

Newcastle-Ottawa Quality Assessment Forms for Cohort Studies
and Case-Control Studies yielded results ranging from six to nine out of nine possible stars. Table 5 depicts the risk of bias in the included studies as assessed using the Cochrane handbook guidelines.
Although, in a number of cases, some aspects could not be assessed, none of the included studies showed a risk of bias in more than one category.

| D ISCUSS I ON
This systematic review was performed to create a list of the as-

| Gender
One of the risk factors for which the highest consistency was retrieved was male gender. Although the prevalence of diabetes in general and especially the one of DF complications is slightly higher for men compared to women, 55 34 8 studies showed a positive association, 25,[41][42][43]45,46,48,53 while 3 studies found a negative association 24,34,50 and 10 studies showed no association with age. 27,28,30,[36][37][38][39][40]47,54  In one study, non-Saudi nationality was associated with a higher risk for foot complications compared to Saudi nationality. 37 Other studies found higher risk in American vs European, 31 Maori vs European, 42 Malay vs Chinese, 50 and Africa American, Location (urban vs rural) One study showed a positive association between low income and the outcome 45 ; another study showed a positive association with low socio-economic position, while not detecting an association with low household income. 48 Two other studies did not find an association. 24,39 Glycaemic control Poor glycaemic control/ HbA1c In 6 studies, a positive association of the outcome with poor glycaemic control/high HbA1c values was found, 25,27,[42][43][44]53 while 4 studies showed no association. 30,37,39,40 Fasting blood/plasma glucose One study showed a negative association and therefore protective effect with metformin use. 35 Furthermore, no association was detected for use of OHAs, 37,40,44 clopidogrel, 37 statins, fibrate/ niacin, 30 glitazones, gliptins and sulphonylureas. 35 One study showed no effect of treatment modality in general (insulin vs OHA vs diet). 27 Characteristics of diabetes  25,27,30,37,42,46,48,53 while 6 studies showed no association with the duration of diabetes. 24,36,39,40,43,44 Type of diabetes While only one study showed a positive association with T2DM compared to T1DM, 43 two other studies showed a negative association 46,48 and two studies found no association. 24 one study found a negative association. 30 In addition, two studies detected no association. 36,54 LDL-cholesterol One study showed a positive association with the outcome, 30 while two studies detected no association. 36 In one study, a positive association was found between dyslipidaemia and the outcome 47; in one study, a negative association was shown with hyperlipidaemia. 41 Two studies could not detect an association between dyslipidaemia and the outcome of interest. 43,46 Lifestyle habits While 3 studies showed a positive association with smoking, 39,42,46 no association was detected in further 3 studies. 30 1 34 While 8 studies found a positive association between hypertension/blood pressure and the outcome of interest, 27,30,34,43,44,46,47,54 a negative association and therefore a protective effect were shown in two other studies. 38,41 In one study, a U-shaped association was detected: both high and low values of systolic blood pressure were associated with higher risk of the outcome. 42 In 3 studies, no association was found. 25,39,40 Pulse pressure While two studies showed a positive association, 43,47 one study found a negative association and therefore a protective effect of higher values of BMI/weight. 39 In addition, one study showed a negative association of obese vs normal weight, while no association was found for over-and underweight vs normal weight. 30 No association was shown in 6 other studies. 34,38,40,42,44,53 Height While a consistently positive association was found in 12 studies, 24,25,27,28,30,34,40,41,43,44,47,53 two studies were not able to find an association. 36 While a positive association with the outcome of interest was found in 7 studies, 27,28,36,37,39,41,43 one study found a negative association and therefore a protective effect of PVD. 53 In one study, while a positive association was found between PVD and the outcome FU, no association was shown with the outcome CA. 34 In two other studies, no association with the respective outcome was shown. 25 While one study showed a positive association, 27 one study showed a negative association and therefore a protective effect. 41 In one study, no association was shown. 25 While 2 studies showed a positive association, 28,30 no association was found in one study. 44 Heart failure In one study, a positive association was shown for the outcome LEA. 41 In another study, a positive association was shown for the outcome FU, but not for the outcome LEA. While a positive association was found in two studies, 30,44 no association was shown in two other studies. 28,41 Angina/ischaemic heart While 7 studies showed a consistently positive association between retinopathy and the outcome, 25,27,28,30,39,41,44 there was one study that showed a positive association with the outcome 0 While a positive association was shown in 5 studies, 27,30,39,50,52 no association was found in 3 studies. 25,41,43 In one further study, While, for total bilirubin, one study showed a positive association, 36 another study showed no association. 39 In a third study, a positive association was found only for indirect bilirubin, but not for direct and total bilirubin. 32 C-reactive protein Year of amputation

TA B L E 4 (Continued)
studies that analysed this association, with risk ratios ranging from 1.05 to 25.4. 24,25,27,28,30,34,40,41,43,44,47,53 In only two studies, no association was shown. 36,39 However, while one of those two studies was a cross-sectional study that did not detect a relationship between elevated vibration perception threshold, an indication of PN, and FU, 36 the other one observed the patients for a follow-up time of only one year in order to assess the development of FU, a time period that might probably be too short to detect long-term complications in a comprehensive manner. 39 In eight studies that assessed the potential association of retinopathy with DF, a consistently positive relationship was shown. 25,27,28,30,34,39,41,44 The only limitation in this agreement is that Dekker et al could show this positive association only when analysing the outcome FU but did not detect an association between retinopathy and the outcome CA. 34 For nephropathy, a positive relationship was shown in six out of nine studies, 27,30,34,39,50,52 while the other three did not detect an association. 25,41,43

| Glycaemic control
Although a strong positive relationship with poor glycaemic control would be logical for all late complications of diabetes, discrepancies were shown in the results regarding HbA1c values, fasting or postprandial blood and plasma glucose concentrations: for HbA1c, a positive association was shown in six studies, 25,27,[42][43][44]53 while, in four studies, no association could be detected. 30,37,39,40 In those that detected a positive association, the risk ratios ranged from values close to one (eg Sarfo et al showed a hazard ratio of 1.11 per one unit (%) increase of HbA1c 43 ) to odds ratios larger than six. 53 In addition, of the four studies that analysed fasting blood glucose, only two showed a positive relationship, 38,46 while two other studies did not find any association. 27,39 Postprandial glucose was only assessed as a potential risk factors in one study, in which a positive association with the outcome FU was identified. Notably, the study group that described this association between postprandial glucose and FU could not find any association of HbA1c and fasting blood glucose with FU. 39 When comparing the study characteristics of the articles that showed varying results concerning the relationship between glycaemic control and DF, there is no notable heterogeneity concerning study design, population sizes or other characteristics that could explain the differences in the results.

| Age and duration of disease
With being examined in 21 studies, age was the risk factor for which a potential relationship with DF was analysed the most. However, the results are highly inconsistent: while eight studies showed a positive relationship with the respective outcomes, 25,[41][42][43]45,46,48,53 a negative relationship and therefore a protective effect of patients' age were shown in three studies. 24,34,50 In addition to that, ten studies could not detect an association between the patients' age and the presence of foot complications. 27,28,30,[36][37][38][39][40]47,54

| Diabetes treatment
When looking at the studies that analysed diabetes treatment and its potential association with foot complications, the picture on a possible influence of insulin use is rather consistent: five out of nine studies detected a positive relationship between insulin and foot complications, 25,30,35,39,54 and no study showed a negative association. For the use of oral hypoglycaemic agents (OHA), the picture is less consistent: while, in one study, a protective effect was shown with metformin use, 35 no association was detected with other OHA in several studies. 30,35,37,40,44 However, these results have to be interpreted with caution since insulin use is associated with patients showing more severe courses of disease and whose blood glucose levels could not be controlled by lifestyle changes or the use of OHA such as metformin. [58][59][60] Besides that, it might be hypothesized that patient groups from earlier years have not been treated according to current treatment guidelines and might have received insulin treatment at earlier time points during their course of their disease.

| Hypertension and dyslipidaemia
Since physiological anomalies such as hypertension and dyslipidaemia are quite common in T2DM, 18

TA B L E 5 (Continued)
eight out of 14 that analysed this association, showed a positive relationship. 27,30,34,43,44,46,47,54 However, in two studies, a protective effect of high levels of blood pressure was described. 38,41 While one of those studies was a rather small retrospective cohort study with 375 patients, in which neither the mean duration of diabetes nor the follow-up time was given, 38 the other study was a large prospective cohort study analysing more than 45,000 subjects. However, also for the latter study, the patients' duration of disease and the followup time were not stated, and the validity of the results can therefore not be fully assessed. 41 Dyslipidaemia is often associated with T2DM: when glucose cannot be metabolized by the cells, fats are mobilized, leading to high levels of fatty acids in the bloodstream. 61 However, it seems that dyslipidaemia is not associated with DF conditions: of four studies that analysed this potential risk factor, a positive association of dyslipidaemia with FU was only found in one cross-sectional case-control study, 47 while, in another study, a protective effect for LEA was shown with hyperlipidaemia. 41 Two further studies identified no association with the outcome of interest. 43,46 In addition, the three studies that analysed the effect of increased cholesterol levels at study entry consistently showed no effect. 38 and late complications such as macrovascular damage that can result in myocardial infarction, PVD or stroke, 3,56,62 it has to be considered that in some articles, it was not possible to distinguish between the diagnosis of dyslipidaemia and/or hypertension and current blood values which can reach normal levels after proper therapy. Therefore, results on dyslipidaemia and/or hypertension as potential risk factors, especially protective results, must be interpreted with caution.

| Obesity, physical activity and height
Although obesity and lack of physical activity are two of the major risk factors for the development of T2DM 18,63 and the biggest part of T2DM might even be attributed to obesity, 64 those factors do not seem to play a crucial role in the development of DF complications: out of 10 studies that evaluated the association of BMI or weight, 30,34,[38][39][40][42][43][44]47,53 only two identified a positive relationship with the outcome, 43,47 while one study showed a negative association. 39 Another study showed a negative association of obese versus normal weight, while no association was found for over-and underweight versus normal weight. 30 In six studies, no association was shown. 34,38,40,42,44,53 Exercise was only analysed as a risk factor in one study, in which no association was shown with the outcome FU. 27 Notably, the analysis of a possible association between height and DF complications led to consistent results over three studies, in all of which a positive association was shown with the outcome LEA. 30,42,46 This might be due to the fact that a taller body implies larger levels of pressure on the limbs or due to neuropathy depending on the length of nerve fibres with longer fibres being more affected than shorter ones. 65

| Peripheral vascular disease and cardiovascular disease
Since T2DM is a metabolic syndrome that increases the risk of heart disease and stroke, 63 showed a positive relationship for this potential risk factor. 54 Besides that, a prospective cohort study on more than 45,000 patients in Taiwan showed a protective effect which might be explained by the fact that patients diagnosed with CVD usually receive medical treatment such as drugs against hypertension, antiplatelet therapy or lipidlowering therapy, thus preventing peripheral arterial insufficiency. 41 In addition, no effect between CVD and any DF was again stated by Al-Rubeaan et al. 25 From a physiological point of view, the protective effect is not expected, since not only PN, but also the damage of blood vessels, which should be advanced in patients with history of PVD and CVD, enhances DF damage, leading to potential necrosis of tissue and the need for amputation. 1

| CON CLUS ION
An important distinction can be made between amenable and nona-

ACK N OWLED G EM ENTS
We appreciate that the Tyrolean Diabetes Working Group sponsored the costs for access to EMBASE.

CO N FLI C T S O F I NTE R E S T
None declared.

E TH I C A L A PPROVA L
The study was conducted in accordance with the Declaration of

Helsinki and approved by the Research Committee for Scientific
Ethical Questions at UMIT University.

DATA AVA I L A B I L I T Y S TAT E M E N T
All relevant data are included in the manuscript.