Factors associated with diabetic foot ulcers and lower limb amputations in type 1 and type 2 diabetes supported by real-world data from the German/Austrian DPV registry

2.1 Data collection

The current analysis is based on the prospective, multicenter diabetes patient registry (DPV [Diabetes-Patienten-Verlaufsdokumentation]), which is a standardized electronic health record developed at the Institute of Epidemiology and Medical Biometry, Ulm University, Germany. The DPV registry provides long-term real-world data on diabetes treatment and outcome from more than 500 centers in Germany, Austria, Switzerland, and Luxembourg. The DPV initiative and the analysis of pseudonymized data were approved by the Ethics Committee of Ulm University (approval number: 314/21) as well as by local review boards. The biannually transmitted data are checked for inconsistency or implausibility and reported back to the respective centers for correction, if necessary.

2.2 Design and participants

The documentation of patient data is prospective in the DPV registry and these data were retrospectively analyzed. All adults (age ≥ 18 years) with a diagnosis of T1D or T2D in the DPV registry were included if they had documented data between 2000 and 2021 and a disease duration of ≥1 year. Individuals with a documented diagnosis of DFU or lower limb amputation before the year 2000, before the age of 18 years, within the first year after their diabetes manifestation, or at an unknown date were excluded from this analysis. Additionally, individuals with a Wagner grade of 0, indicating pre- or postulcerative lesions (N = 17 165, 4.6%), were excluded as they could neither be appropriately allocated to the DFU group nor to the control group (Figure S1). All remaining individuals were allocated to one of four groups (separately for T1D and T2D): (a) no DFU and no lower limb amputation, (b) DFU without any lower limb amputation, (c) DFU and a minor amputation, and (d) DFU and a major amputation. For this stratification, lower limb amputations comprised all amputations at the visit of the first documentation of DFU and thereafter. Amputations were classified as minor if the amputation was performed below the ankle and as major if the amputation was conducted above the ankle.

For group (a) without DFU, aggregated data from the most recent treatment year of each individual were analyzed. For individuals in the other three groups, the year before the first diagnosis of DFU was analyzed for cross-sectional comparisons to people without DFU. For these individuals, the visits following their first documentation of DFU were additionally analyzed with respect to amputations (longitudinal analysis).

2.3 Propensity score matching

A total of 359 460 individuals fulfilled the inclusion criteria (31 801 with DFU). To control for age, sex, diabetes duration, and treatment year, the individuals with DFU were matched one to one to the control group without DFU regarding these variables, separately for T1D and T2D. The variables sex and treatment year were matched exactly. Age and diabetes duration were included continuously as well as categorized (T1D: age 18 < 25, 25– <50, 50+ years, diabetes duration 1–10, >10–30, >30 years; T2D: age 18– <60, 60– <80, 80+ years, diabetes duration 1–5, >5–10, >10–20, >20 years) into the nonparsimonious propensity score. One-to-one matching was conducted using a greedy-matching algorithm with a caliper width of 0.2. Standardized differences were assessed before and after matching to evaluate balancing of covariates between the matched cohorts (Table S1). A standardized difference of <10% for a baseline covariate reveals a negligible imbalance.¹³ For 31 732 individuals with DFU a matching partner could be allocated leading to a final cohort of 63 464 adults with type 1 diabetes (N = 5352) or type 2 diabetes (N = 58 112).

2.4 Data management

The diagnosis and classification of diabetes types were conducted by the diabetologists at the participating centers according to the clinical presentation of the people with a diabetes manifestation and lab results, based on German guidelines.¹⁴ Anthropometric measurements were performed in the local centers according to in-house protocols. The BMI was calculated as weight in kilograms divided by height in meters squared. Lipid values were converted to mg/dL. Total cholesterol (TC), low-density lipoprotein (LDL), high-density lipoprotein (HDL), and triglycerides (TG) were analyzed. Non-HDL was calculated through the following formula: TC–HDL. Additionally, the ratio of TC/HDL and LDL/HDL was calculated. HbA1c values were standardized to the Diabetes Control and Complications Trial reference range of 4.05%–6.05% (20.7–42.6 mmol/mol) using the multiple of the mean transformation method to account for different laboratory methods.^{15, 16} Hypertension was defined as systolic blood pressure (BP) ≥140 mm Hg or diastolic blood pressure ≥90 mm Hg. Metabolic syndrome was defined per the International Diabetes Federation Consensus Statement¹⁷ as obesity (BMI ≥30 kg/m², because waist circumference was not available in most individuals) and one of the following: TG >150 mg/dL, HDL <40 mg/dL (males) or <50 mg/dL (females), lipid-lowering medication, systolic BP ≥130 mm Hg or diastolic BP ≥85 mm Hg, or antihypertensive medication. The third criterion of elevated fasting plasma glucose was present in all participants through the documentation of diabetes. Smoking was defined as smoking at least one cigarette per day (yes/no). Alcohol abuse was considered as a daily alcohol consumption of >24 g in men and >12 g in women.¹⁸ Missing data on smoking behavior were not used for analysis, whereas no documentation of critical alcohol intake at any visit was classified as no alcohol abuse. Smoking behavior, alcohol abuse as well as documented diagnoses of neuropathy, PAD, and retinopathy¹⁹ according to guideline-based in-house evaluations of the clinicians were aggregated over the whole documentation period (for adults with DFU only before the first documentation of DFU). Nephropathy was defined as estimated glomerular filtration rate <60 (mL/min/1.73²) according to the Chronic Kidney Disease Epidemiology Collaboration formula²⁰ or as documented microalbuminuria or if renal transplantation or a history of dialysis was documented. For identification and classification of the severity of DFU Wagner grades were used as this classification system is still recommended in the most recent guideline on diabetic foot from the German Diabetes Association and therefore was used by the centers participating in this study during foot examinations and is included in the DPV documentation software.²¹ Wagner grades range from 0 to 5 indicating no current ulcer, but pre- or postulcerative lesions (Wagner 0), superficial ulcer (Wagner 1), deep ulcer without abscess or bone involvement (Wagner 2), ulcer with abscess and bone involvement (Wagner 3), localized ulcer with gangrene (Wagner 4), and extensive ulcer with gangrene (Wagner 5). Wagner grades of 1 or higher were defined as DFU. Individuals with Wagner 0 were excluded. For each individual the highest documented classification at time of first occurrence of DFU was considered in individuals with DFU. In individuals without DFU the documented foot examination within the most recent treatment year was used for identification of those with Wagner 0. If one of the two matching partners had missing values on an outcome variable, data from both matching partners were excluded from the analysis on this variable to avoid bias through asymmetry of missing data between people with and without DFU.

2.5 Statistical analysis

All statistical analyses were generated using SAS (Statistical Analysis Software, SAS Institute Inc., Cary, NC, USA) Version 9.4, Built M7, on a Windows Server 2016 mainframe. For all analyses a two-sided p value of <.05 was considered as significant. Descriptive statistics were performed for all individuals before matching and stratified by DFU and amputation separately for T1D and T2D (Table 1). The results are shown as median with quartiles for continuous variables and as proportions for binary variables. p values were calculated using the chi-square test for categorial variables and the Wilcoxon's rank sum test for continuous variables and adjusted for multiple comparisons, separately for T1D and T2D. Besides the unadjusted proportions of people with DFU or amputations by diabetes type among the unmatched cohort, odds ratios (OR) were calculated using logistic regression models, adjusted for age categories (18– <40, 40– <60, 60+ years), sex, diabetes duration categories (1–10, >10–25, >25 years), calendar year categorized (2000–2005, 2006–2010, 2011–2021), smoking history, and alcohol abuse to compare the risk for DFU and amputation of individuals with T1D and T2D. Cross-sectional comparisons of the matched cohort with DFU vs the control group were conducted irrespective of DFU with or without amputations, both for the whole group and stratified by sex. p values were calculated in the same way as within the descriptive analysis and adjustment for multiple comparisons was implemented within each subgroup. The risk for amputation following the first documentation of DFU was analyzed with Cox regression to calculate hazard ratios (HRs). Individuals with DFU and at least 100 days of follow-up (first documentation of DFU was considered as baseline), but without an amputation within these first 100 days were included. Therefore, 1107 adults with T1D and 8586 adults with T2D were selected for longitudinal analysis (Figure S1). Due to the reduced number of individuals included in this analysis, minor and major amputations were pooled as outcome variable. HRs were calculated unadjusted and adjusted for sex, age categories, diabetes duration, and calendar year at baseline. Age groups for this analysis differed from age groups used for matching in T1D, as this follow-up cohort was considerably older (age groups for follow-up analysis in T1D: 18– <40, 40– <60, 60+ years). We additionally conducted a “complete model” analysis with all relevant variables included in one model. Individuals were censored at the time of last contact if an amputation was never conducted.

3.1 Description of unmatched cohort

Among all individuals in the unmatched cohort (N = 359 460), the proportion of males was 53% in T1D and 52% in T2D. Median age with interquartile range was considerably lower in T1D (32.8 [19.6–53.6] years) than in T2D (71.0 [61.4–78.8] years), but diabetes duration was longer (T1D: 13.2 [7.3–23.2] years; T2D: 10.2 [5.4–16.7] years). For both diabetes types, individuals with DFU were more often male, were older, had longer diabetes duration, and the analyzed calendar year been earlier. Further characteristics of the unmatched cohort are presented in Table 1 and standardized differences of matching variables before and after matching are shown in Table S1. All matching variables could be equalized sufficiently as all standardized differences were ≤1% after the matching procedure.

Besides individuals with Wagner grade 0, who were excluded from the analysis, about one half of the remaining individuals had a documented Wagner grade. Of those with a documented foot examination, 88.5% were classified as healthy (no signs of DFU) and 5.0%, 3.0%, 2.2%, 1.2%, and 0.1% had the Wagner grades 1–5.

3.2 Proportion with DFU

The unadjusted proportion of people with DFU was 5.0% in T1D and 9.5% in T2D. The adjusted OR with 95% confidence limits for DFU for T2D vs T1D was 1.22 [1.16–1.29], p < .001. Among individuals with DFU, the unadjusted proportion with amputations in the subsequent course of their disease was 36.5% in T1D and 38.1% in T2D representing an adjusted OR for amputations among adults with T2D of 1.17 [1.06–1.30] (p = .003) compared to T1D.

3.3 Cross-sectional comparisons

Within the matched cohort with T1D, the factors associated with DFU were higher body height (p = .013), increased HbA1c (p < .001), triglycerides as well as the ratio of TC/HDL or LDL/HDL (all p < .01), smoking history (p = .007), and multiple subcutaneous daily injections with insulin pens (p < .001). The results regarding triglycerides levels were similar even with additional adjustment for HbA1c, type of insulin administration (injection vs pump use) and prescription of lipid-lowering medication. No clear evidence was observable for blood pressure, but the proportion treated with antihypertensive medication was higher in individuals with DFU (p < .001, Table 2). In women, BMI, metabolic syndrome, and lipid levels, especially TG and the ratios of TC/HDL or LDL/HDL, were related to DFU, whereas in men HbA1c and a smoking history were more prominent (Table 2). In adults with T2D, alcohol abuse, a smoking history, and insulin therapy (all p < .001) were related to DFU, whereas HbA1c did not differ and lipid levels as well as the proportion with hypertension were lower in individuals with DFU. Similarly to T1D, the proportion treated with antihypertensive medication was higher (p < .001) in individuals with DFU (Table 3). For both diabetes types the proportion with documented neuropathy, PAD and nephropathy was significantly increased (all p < .001) whereas the increase in retinopathy was significant only for T2D individuals with DFU (Tables 2 and 3).

3.4 Longitudinal analysis

Cox regression analyses of the 9693 individuals with DFU and sufficient follow-up data revealed that in T1D, male sex, long diabetes duration, high Wagner grades, and high HbA1c were the strongest predictors for future amputations (Table 4). These results remained significant even in the “complete model” analysis with the exception of diabetes duration (Figure 1). In T2D, the same predictors could be observed, but high BMI was associated with lower HRs for future amputation (Table 4). Again, these results remained more or less stable in the “complete model” except for diabetes duration and HbA1c to some extent (Figure 1). Individuals on insulin had higher HRs whereas individuals on metformin or GLP-1 receptor agonists (GLP-1RA) had lower HRs (Table 4), but this effect vanished partly combining all medications in one model (Figure S2). The crude HRs for those using sodium glucose transporter 2 inhibitors (SGLT2i) were elevated, but the statistical significance vanished with adjustment for age, sex, diabetes duration, and treatment year in both groups (with additional insulin therapy and those without insulin treatment). The crude HRs were higher in the group without insulin therapy (Table 4). Smoking and alcohol abuse did not reveal significantly higher HRs for amputations in both T1D and T2D. Data from the “complete model” analysis on most important predictors are shown in Figure 1 and all analyzed HRs, crude and adjusted, are presented in Table 4.