Introduction

Diabetes mellitus (DM) is a metabolic disorder whose incidence has shown a rapid rise since the 21st century, becoming one of the most severe challenges facing the global public health field1. In 2019, the global prevalence of diabetes was estimated to be 10.5% (536.6 million people) and is predicted to rise to 12.2% (783.2 million) by 20452. DM not only has a powerful impact on the quality of life of patients, but also leads to complications such as cardiovascular disease, kidney disease, and neuropathy, resulting in a heavy burden on society and the medical system. Diabetic foot ulcers (DFUs), characterized by infection or destruction of the soft tissues of the foot and classified based on the affected area and extent of the lesions3, is a critical and severe complication of diabetes, notorious for its lower extremity amputation (LEA) and mortality rates4,5.

Worldwide, approximately 1.6 million people undergo amputations each year, of which approximately 33% are severe amputations. DFUs are the leading cause of non-traumatic amputations, with more than one million diabetic patients undergoing non-traumatic LEA each year, nearly 85% of which are due to DFUs. Crude estimates of 5-year mortality after amputation range from 39 to 68%, which are higher than the mortality rates of some common tumors6. In additional, DFU and their worst adverse consequences, especially amputation, have a catastrophic impact on the mental and physical health of patients, including prolonged hospitalization, heavy economic burden, difficult treatment, and significantly impaired quality of life7. Therefore, implementing effective strategies to prevent LEA in patients with DFUs is crucial, which could be achieved through risk factor identification. However, many related studies have focused on risk factor analysis8,9, and few have established predictive models to estimate the risk of LEA in patients with DFUs7,10,11. Regrettably, some potential risk factors, such as the classification of CKD, Barther index, and osteomyelitis, are omitted from these predictive models, which can reduce the clinical utility.

Therefore, considering these challenges, establishing a nomograms model that includes as many potential risk factors as possible is essential for quantifying the risk of LEA in patients with DFUs and proposing precautionary protocols. The study showed that the nomogram model has an excellent visual prediction effect in predicting adverse events and making clinical decisions. The numerical relationship between specific diseases and risk factors can be graphically represented through a scoring system without any complicated calculation formula, which facilitates the effective screening of high-risk patients and timely intervention measures7. Considering these advantages, we hope to provide a more effective and utilitarian tool for clinicians to identify optimal treatment options for patients with DFUs and avoid the occurrence of LEA.

Methods and materials

Study design and participants

From January 2020 to December 2023, a total of 144 patients with DFUs were enrolled in this study at the First Affiliated Hospital of Nanchang University; all patients were from the Department of Orthopedics, including 107 men, 37 females, aged 30–89 (60.43 ± 12.01); duration of diabetes ranging from 0.5 to 30 (10.01 ± 7.14) years; 3–3650 (120.41 ± 362.13) days. The inclusion criteria were as follows: (a)all participating patients met the DFU diagnostic criteria issued by the International Working Group on the Diabetic Foot(IWGDF) in 2015 and (b) patients with complete case data. The exclusion criteria were as follows: (a) age < 18 years; (b) infections other than DFU; (c) patients with a malignant tumor; and (d) history of lower limb amputation. This retrospective study was approved by the Ethics Committee of the First Affiliated Hospital of Nanchang University and the need for informed consent was waived. All methods in this study were performed in accordance with the relevant guidelines and regulations and the Declaration of Helsinki.

Data collection

We designed a clinical investigation case report form to collect the clinical data from the information system, including gender, age, BMI, course of DM, course of DFU, diabetic complications, Wagner, clinical symptoms, peripheral angiopathy of diabetic (PAD), white blood cells (WBC), neutrophil ratio(N%), hemoglobin, platelet count, C-reactive protein(CRP), erythrocyte sedimentation rate(ESR), albumin (ALB), blood uric acid (BUA), kidney function, pathogen, and the number of blood sugar abnormalities(BSA). We established a database to collect survey data, and all data were input and checked by two people. LEA includes major and foot amputation. A major amputation (lower knee or higher amputation) is an amputation above the plane of the ankle joint, while, foot amputations were performed at or below the ankle joint. The study population was divided into two groups, including amputation and no amputation groups.

Statistical analysis

All statistical analyses were carried out using the SPSS26.0 and R4.3.3 software. Continuous variables were expressed as mean ± standard deviation or median and were compared using the independent sample t-test. Means were used for normally distributed data, whereas medians and interquartile range(IQR) were used for no-normally distributed data. Means and medians were compared using the Student’s t-test or Wilcoxon test as appropriate. The 95% confidence intervals (CI) were estimated using binomial distribution. Statistical significance was set at P < 0.05. Single-factor logistic regression analysis and multivariate logistic regression analyses were conducted to identify the predictive factors for distinguishing correlations with the amputation. We used R4.3.3 software to establish a nomogram prediction model to predict LEA, and further applied the calibration curve and the area under the curve (AUC) of the receiver operating characteristic(ROC) curve to estimate the performance of the nomogram prediction model. Finally, a decision analysis curve (DCA) was used to evaluate the clinical efficacy of the nomogram by analyzing the net benefit under different risk thresholds in patients with DFUs.

Results

Baseline clinical characteristics of participants

Among the 144 DFU patients with DFUs in our study, 107 were males and 37 were females. Of these 72(50%) patients had undergone amputation(amputation group), with an average age of 62.75 ± 11.338 years, and 72(50%) had not amputation(no amputation group), with an average age of 58.11 ± 12.276 years. In the univariate analysis, the differences of the clinical data for 12(age, Wagner grades, PAD, CKD, WBC, N%, Hemoglobin, Platelet, CRP, ESR, ALB, number of BSA) of the 20 variables between the amputation group and no amputation groups were statistically significant(P < 0.005), as shown in Table 1. The average period from hospitalization to amputation was (6.29 ± 6.59)days. On admission, patients had a systolic BP between 76 and 175 mmHg, diastolic BP between 50 and 110 mmHg, the cardiac ejection fraction ranged from 50 to 92%. Pathogens causing DFUs were isolated from 78 foot ulcer specimens of 144 patients, resulting in an infection rate of 54.16%. Of these, 44 were single infections, and 26 were mixed infections (≥ 2 bacteria), leading to a mixed infection rate of 33.33%. A total of 105 pathogenic bacterial strains were isolated, as shown in Fig. 1, yielding an average infection rate of 1.34 strains per case. The three most common pathogenic genera were Staphylococcus 22 (20.95%), Enterococcus 17 (16.19%), and Escherichia 13 (12.38%). The major pathogenic species in the above genera were Staphylococcus aureus 16 (15.23%), Enterococcus 17 (16.19%), and Escherichia coli 13 (12.38%).

Table 1 Differences in characteristics between the without amputation group and the amputation group.
Full size table
Figure 1
figure 1

Pathology and medication of diabetic foot ulcer pathogens.

Full size image

Multiple factor logistic regression analysis

The variables ( age, Wagner grade, PAD, CKD, WBC, N%, Hemoglobin, platelet, CRP, ESR, ALB, and number of BSA) in the univariate analysis (P < 0.05) were considered primary predictive factors. Taking amputation (assignment: without amputation = 0, amputation = 1) as the dependent variable, the primary predictive factors (assignment: without PAD = 0, peripheral PAD = 1; CRP(1–10 = 0, 11–50 = 1, 51–100 = 2, 101–200 = 3,≥200 = 4; without CKD = 0, CKD = 1; Number of BSA(0–10 = 0, 11–30 = 1, 11–30 = 1, 30–50 = 2, ≥ 50 = 3) as the independent variables) were analyzed using single factor logistic regression. The results showed that the Wagner grade, PAD, CRP, CKD, and number of BSA were the risk factors for amputation in patients with diabetic foot, as shown in Table 2.

Table 2 Risk and protective factors of amputation in patients with diabetic foot ulcer.
Full size table

Development of nomogram prediction model

The nomogram prediction model was established by including the influencing factors of poor prognosis listed in Table 2. The five independent influencing factors(Wagner grade, PAD, CRP, CKD and number of BSA) were introduced to the R4.3.3 software. The contribution of each independent factor was presented visually, and the sum of each factor was the risk predictive value of LEA. The nomogram showed that the risk of LEA increased with Wagner grade, PAD, CRP, CKD, and increased number of BSA; results are reported in Fig. 2. A 73-year-old male with 20-years history of type 2 diabetes had an ulcer on his left foot. The patient underwent debridement and daily dressing changes at a local hospital. However, the ulcer did not heal. Laboratory examination revealed a CRP level of 76.3 mg/L, an ESR of 100 mm/h and a GFR of 35.91mL/min. Radiography revealed abnormal bone changes in the left foot. CT Angiography indicated atherosclerotic changes in the lower limb vessels, and the left anterior and posterior tibial arteries showed local stenosis and occlusion of the fibular artery, as well as parts of the right anterior tibial, posterior, tibial and peroneal arteries. Blood glucose abnormalities occurred 28 times during hospitalization, according to the nomogram that we have constructed, the total score was 244 points, and the predicted value of the LEA was 0.996 (0.97,0.999). Truncation of the left calf was performed eight days after admission, and photographs of the patient before and after the amputation are shown in Fig. 3.

Figure 2
figure 2

Development of the risk nomogram.

Full size image
Figure 3
figure 3

Photos of the patient before and after amputation.

Full size image

Validation of nomogram prediction model

The AUC of the ROC predicting the poor prognosis of patients with diabetic foot ulcer patients was 0.967 (95% CI 0.944 to 0.991), indicating that the nomogram predicts poor prognosis with good discrimination, as shown in Fig. 4. The bootstrap method repeated sampling 400 times for internal verification, and the calibration curve was a line with a slope of approximately 1, indicating that the nomogram had good consistency in predicting LEA. The DCA showed that using this nomogram to predict LEA had a net benefit across a wide range of threshold probabilities. The results are reported in Figs. 5 and 6.

Figure 4
figure 4

The accuracy of the nomogram for predicting LEA using ROC curve.

Full size image
Figure 5
figure 5

Calibration curves of the LEA prediction in the cohort.

Full size image
Figure 6
figure 6

Decision curve analyses (DCA) for the LEA nomogram.

Full size image

Discussion

DFUs are a common complications of diabetes. Annually, about 18.6 million are afflicted with DFUs12. Ulcers are associated with compromised physical ability, diminished quality of life, and increased use of healthcare resources. Without proper treatment, DFUs have the potential to develop into infections affecting the soft tissue, which can further deteriorate into gangrene and may even require LEA. Every year, approximately 2 million Americans with diabetes suffer from DFUs. Within five years post-ulceration, 5% of these patients will have to undergo a major amputation, while an even more alarming 50–70% will die13. The direct expenses for treating DFU and LEA in the United States range from $9 billion to $13 billion per year14. The incidence of new DFUs in China was 8.1%, and the incidence of new ulcers in patients with healed DFUs within one year was 31.6%15. This imposes significant pressure on families of patients and the national healthcare system16,17. However, tools that can effectively predict the risk of amputation in patients with DFUs are scarce. Therefore, predicting the risk of amputation is important. This study aimed to establish a risk-scoring model that can be used in clinical practice to predict the risk of amputation in patients with DFUs. These models can be used to help clinicians identify high-risk factors for amputation early, thus establishing appropriate treatment options and targeted measures to prevent morbidity and mortality.

In previous studies, the amputation rates associated with diabetic foot disease ranged from 4.7 to 47.7%18.In the present study, we enrolled 144 patients with type 2 DM and DFUs; LEA was present in 72 patients. The probability of LEA was as high as 50%, similar to that reported in the literature. We reviewed and analyzed the data and identified several independent risk factors, such as Wagner, PAD, CRP, renal dysfunction. To reflect the role of these risk factors in predicting the risk of amputation more intuitively, they were integrated into a nomogram model. The model has been proven to have good predictive performance and can effectively identify patients at high risk of LEA. These results provide a strong basis for clinicians to prevent and control LEA in the management of diabetic foot disease.

The Wagner classification system19 is frequently used in clinical settings to assess the depth of wounds and determine the presence of complications such as osteomyelitis, necrosis, or gangrene20. Our finding that higher Wagner scores correlated strongly with the risk of amputation (OR 4.359) is reasonable because a higher Wagner grade is diagnostic of severe gangrene, which typically results from ischemic and infectious processes. A recent meta-analysis showed that Wagner Grades 4 (OR 4.3) and 5 (OR 6.4) were significant predictors of amputation risk in patients with DFUs21. Our results are similar to this study. These findings are similar to those yielded in research from Nigeria (OR 5.953) and Ethiopia (OR 4.7)22,23. Hence, timely identification and management of DFUs of earlier Wagner grades are crucial to avoid the need for amputation.

The prevalence of PAD in our study was 52.08%, and Peng et al.reported that the prevalence of PAD was 49.6%7; our results are largely consistent with the published literature. As an essential risk factor for DFU occurrence and adverse clinical outcomes, PAD has harmful effects on wound healing; if not treated promptly, it can lead to amputation and ultimately increase mortality. PAD occurs as a result of atherosclerosis, leading to narrowing or blocking of the arteries and a subsequent decrease in blood flow to the limbs; if not treated promptly, it leads to amputation and ultimately increases mortality24. Additionally, attempts to resolve any infection are impaired because of a lack of oxygenation and difficulty in delivering antibiotics to the infection site25.

Infection is one of the most common causes of amputation, and inflammatory markers play a crucial role in the decision-making process of lower limb infections. A correlation exists between acute-phase reactants and the risk of amputation in individuals with diabetic foot infections. CRP is an acute-phase response protein with a homopentameric structure and acts as a positive acute-phase reactant26. Zhang et al27. systematically reviewed published epidemiological studies to confirm the strong association between DFU infection and CRP. Sharma et al. reported that the sensitivity and specificity of CRP for diagnosing grade 2 ulcers were 77.4% and 84.3%, respectively28. The present study further confirmed that CRP is an independent risk factor for LEA, thus requiring measures to control inflammation in patients with DFUs, such as adequate debridement and effective anti-infective therapy.

Up to 40% of patients with diabetes patients are expected to develop CKD, and 19–34% will suffer from DFUs during their lifetimes29. CKD is an indicator of the generalized vascular status in patients with diabetes. It has also emerged as an independent risk factor for the development of foot lesions in diabetic patients30. In a case-control study involving 351 patients with DFUs, Lee et al. established a significant correlation between a low estimated glomerular filtration rate (eGFR) and an elevated risk of LEA31. Jupiter et al. emphasized that CKD is a standalone factor associated with higher mortality among patients with DFUs32. The prevalence of CKD in our study was 20.83%, and we demonstrated an association between CKD and the risk of severe LEA events. One possible reason is that CKD can affect blood lipids, blood pressure, and other metabolic indicators, further aggravating the risk of vascular disease and emphasizing the need for comprehensive risk assessment and management strategies for patients with CKD.

Steady glycemic control has important implications in preventing amputation. Despite variations in the identified risk factors among studies, blood sugar control consistently appears to be significantly associated with an elevated risk of amputation, as reported in various studies33,34. We indirectly measured the patients’ sugar control levels by counting the number of BSA readings from hospital monitoring. Our research confirmed a correlation between the number of BSA and an increased risk of more severe LEA events. Therefore, reasonable blood glucose control is of great significance in preventing amputation.

In the validation of the nomogram model, the ROC curve is 0.967 (95% CI 0.944 − 0.991), which shows the model had good performance. Our analysis demonstrated that the nomogram was well developed and accurately predicted LEA prediction through discrimination capability evaluation. Then, the Hosmer-Lemeshow test was performed, and the calibration curve was done by producing 400 bootstrap samples to displace the primary samples and repeating the entire modeling process, illustrating an excellent degree of fit. In addition, the DCA curve indicated that the nomogram model was accurate.

Overall, our nomogram model exhibited a more accurate value for LEA prediction, and its construction considerably increased the clinical utility of dramatically improving the precision of the therapeutic options in clinical practice.

However, this study has some shortcomings: (1) the sample size was small, and only orthopedic patients were included; the representativeness may have been insufficient, which might have affected the comprehensiveness and representativeness of the results; (2) this analysis was a retrospective design, which requires further prospective studies in the later stage; (3) the nomogram model mainly adopts internal verification, based on the sample size limit; (4) this study was conducted on amputated patients, and the target cases were limited, which could have led to a bias in our results; (5) hemodynamic information included only cardiac function results, which is a limitation.