2.1 Participants

The institutional ethical review committee of Huashan Hospital of Fudan University approved the prospective study (HIRB No. 2021-01030). Informed consent was obtained from the participants. From May 2022 to April 2023, 84 inpatients with diabetes were enrolled in this study, and 32 healthy individuals with no prior history of diabetes were recruited from the community to serve as the control group. Inclusion criteria for participants: age between 18 and 85 years, ability to walk independently, and no other disease-causing peripheral neuropathy like Guillain–Barre syndrome, disc herniation pressing on nerves, tumors, etc. Exclusion criteria for the healthy control group were as follows: (1) previous diagnosis of diabetes mellitus; and (2) presence of foot ulcers, infections, and amputations. The exclusion criteria for the diabetic group were as follows: (1) a history of foot amputation; (2) the presence of active ulcers on both feet, with tissue edema triggered by acute or subacute inflammation severely affecting the thickness and stiffness of plantar tissue. Patients with diabetes underwent a thorough peripheral nervous system examination by an endocrinologist. The Neuropathy Disability Score (NDS) is utilized to detect peripheral neuropathy. The distribution of NDS scores was shown in Table 1, with a score exceeding 2 signifying neuropathy presence [3]. According to the NDS score, 84 inpatients with diabetes were divided into the DPN group (n = 41) and non-DPN group (n = 43).

2.2 Thickness and Stiffness Measurement of Plantar Tissue

All images were acquired using the Cannon Aplio i900 ultrasound (Cannon Medical Systems, Japan), fitted with a broadband linear array probe of 5–18 MHz frequency, designed for musculoskeletal applications. To avoid the influence of extension or flexion on plantar tissue properties, the examination was performed with the subject in a lateral position with the knee flexed and the ankle relaxed, and the subject's plantar tissue facing or facing back to the examiner. All measurements were conducted by Dr. Li, a highly experienced sonographer with a decade of expertise in clinical ultrasound and specialized knowledge in advanced musculoskeletal ultrasound, who holds an accreditation in the field. Real-time imaging was used to observe the plantar structure and determine the measurement locations, which included the heel, hallux, first metatarsal head (1st MTH), and fifth metatarsal head (5th MTH) on the longitudinal imaging plane across four plantar foot locations. These sites, more prone to higher plantar pressure and diabetic foot ulcer development, were where plantar tissue thickness (mm) and stiffness (in terms of shear wave speed; m/s) were measured. The transducer was gently positioned on the sole in a non-loaded state to prevent tissue compression or deformation resulting from transducer pressure.

The thickness measurement referred to the distance between the skin surface and the corresponding bone cortical convexity in the image. Stiffness measurements were conducted in shear wave elastography (SWE) mode, with color elastography overlaid on the corresponding B-mode images. The elastography sampling frame included the skin and bone cortex at corresponding locations. Circular regions of interest (ROI) with size 5 were traced at the center of the elastography sampling frame, avoiding areas near the skin and bone cortex that are more susceptible to artifacts from shear wave boundary effects or multiple reflection-induced wave superimposition, as shown in Figure 1. The stiffness of plantar tissue at each location was defined as the mean shear wave speed within the corresponding ROI. In the case of subjects without foot ulcers, data was gathered from both feet and then averaged. However, for those with active ulcers on one foot, considering that inflammation can affect the plantar nature and the characteristics of bilateral involvement of DPN, measurements were taken solely from the unaffected foot.

2.3 Statistical Analysis

All statistical analyses were undertaken using SPSS (Version 25, IBM, Armonk, NY, USA). Statistical significance was set at p < 0.05. The Shapiro–Wilk test checked all continuous parameters' normal distribution. Data were expressed as mean ± SD if normally distributed, and median (interquartile ranges) if otherwise.

Differences in demographic data, plantar thickness, and stiffness between groups (HC/non-DPN/DPN) were analyzed using ANOVA or the Kruskal-Wallis rank sum test. Sex ratios were evaluated with Pearson's Chi-squared test. Either the Mann–Whitney U-test or t-test was used for comparing clinical parameters between the DPN and non-DPN groups. The Kruskal-Wallis rank sum test examined variations in thickness or stiffness among the four different locations.

We used univariate logistic regression models for clinical and ultrasonic parameters to identify independent variables that could aid in DPN risk. Multivariate logistic regression then incorporated parameters with p-values less than 0.1. Variables significant at p < 0.05 were included in the model. The model's accuracy in DPN was evaluated using the ROC curve, with the Youden index identifying the best cutoff point.

2.4 Results

A comparison of demographics in the different groups is shown in Table 1. The DPN group was significantly older compared to the non-DPN (p = 0.006) and HC groups (p < 0.001). Concurrently, patients with DPN tended to include a higher percentage of males compared to those without DPN (p = 0.006). The DPN group exhibited a significantly lower BMI compared to the non-DPN group; however, there were no significant differences in height and weight between these groups(p > 0.05).

2.4.2 Differences in Plantar Tissue Thickness and Stiffness at Multiple Locations

There were significant differences in plantar thickness and stiffness in different locations, as shown in Figure 2. The thickness and stiffness of the plantar tissue were more heterogeneous in people with diabetes compared to healthy individuals.

2.4.4 Logistic Regression Analysis for Risk Factors Associated With DPN

Results of the univariate and multivariate logistic regression analysis for DPN are shown in Table 3. The univariate analysis showed that having an older age, a lower BMI, a longer duration of DM, increased cholesterol, male gender, a lower integrated thickness index (Z2TH), and a higher integrated stiffness index (Z1SWE) significantly increased the odds for DPN. Furthermore, in the multivariate analysis, a higher integrated stiffness index (odds ratio [OR] 2.79 [95% CI 1.27–6.12]), longer duration of diabetes (1.68 [1.31–1.91]), and gender(women contrast men) (0.09 [0.01–0.75]) remained independently significantly related to DPN. Based on these results, the estimated logistic regression equation was defined as: logit(p) = −6.451 to 2.437 (men) + 0.526 (duration of diabetes) + 1.027 (Z1SWE). The ROC curve was plotted based on the explanatory probability of the multivariate logistic regression model, and the area under the ROC curve was 97.7%; in addition, the optimal probability cutoff value of 0.56 yielded a 92.7% sensitivity (correctly classifying the group with DPN) and 95.3% specificity (correctly classifying the group without DPN) for this model (Figure 3).

TABLE 3. Univariate and multivariate logistic regression model of DPN.

Variables	non-DPN (n = 43)	DPN (n = 41)	p-value univariate analyses	p- value multivariate analyses	OR value (95 CI%)
Age (y)	59 (51–66)	64 (59–70)	0.004*
BMI (kg/m2)	26.39 ± 3.81	24.24 ± 3.58	0.014*
Duration (y)	2 (0.58–6)	15 (12–22)	< 0.001*	< 0.001	1.676 (1.313–1.909)
HbA1C (%)	7.6 (6.5–10)	8.2 (7.3–10.1)	0.138
FBG (mg/dL)	7.7 (6.1–9.1)	7.7 (6.1–9.1)	0.470
PBG (mg/dL)	11.6 5 ± 4.48	11.18 ± 3.94	0.610
Cholesterol (mg/dL)	4.47 ± 1.07	3.85 ± 1.41	0.032*
Triglyceride (mg/dL)	1.73 (1.02–2.68)	1.28 (0.92–1.88)	0.317
Male (%)	21 (48.8)	32 (78)	0.007*	0.026	0.087 (0.01–0.747)
Z1TH	−0.0883 (−1.0102 ~ 0.9708)	0.2793 (−0.6696 ~ 0.7319)	0.517
Z2TH	−0.1165 (−0.7746 ~ 0.4962)	−0.1646 (−0.4216 ~ 0.4372)	0.079*
Z1SWE	−0.4599 (−1.1839 ~ 0.3235)	0.0714 (−0.6497 ~ 1.3959)	0.020*	0.010	2.792 (1.274–6.119)
Z2SWE	−0.0470 (−0.5323 ~ 0.3885)	−0.0414 (−0.6311 ~ 0.8165)	0.913

• Abbreviations: DPN, patient with diabetic peripheral neuropathy; non-DPN, patient without diabetic peripheral neuropathy; Z1TH/Z2TH, integrated thickness index generated by principal component analysis (PCA); Z1WSE/Z2WSE, integrated stiffness index generated by PCA.
• *p < 0.1.

2.5 Discussion

Our study investigated differences in plantar tissue thickness and stiffness among individuals with diabetes (with and without DPN) and healthy controls, highlighting the potential of these measures for DPN. We constructed integrated thickness and stiffness indexes that comprehensively capture the characteristics of plantar tissues by using PCA. Furthermore, the model developed in this study, which combines the plantar stiffness index with gender and diabetes duration, provides a robust and noninvasive tool for identifying DPN in individuals with diabetes.

Several studies report conflicting findings on plantar tissue thickness changes in people with diabetes. Our study found an increased overall plantar tissue thickness in people with diabetes, more so in those with neuropathic complications. Notably, significant differences in plantar tissue thickness at the 1st MTH and heel were observed between some groups. This corresponds with Chao's findings, revealing thicker plantar tissue in people with diabetes and increased thickening at more locations, including the 3rd and 5th MTH and the hallux [25]. It is worth mentioning that they used an ultrasonic palpation system to measure the thickness of the plantar tissue, a method that does not allow for visualization and therefore may not be as precise as ultrasound when choosing where to measure. This approach was also used in a study by Sun [21], which showed no significant difference in plantar thickness between healthy individuals and those with diabetic neuropathy. Plantar soft tissues are rich in collagen. Therefore, collagen components are susceptible to non-enzymatic glycation. Increased tissue thickness in people with diabetes is thought to be associated with a hyperglycemic state, which promotes non-enzymatic glycation of structural proteins in collagen, elastin, and keratin, which promotes intermolecular cross-linking of soft tissues, which in turn leads to tissue thickening [29-32]. Meanwhile, other studies have found the exact opposite results. The thickness of the heel and the first and second metatarsal heads was greater in healthy controls than in diabetic patients, who in turn had greater foot thickness than those with ulcers. Based on this result, the study hypothesized that the loss of tissue mass in the feet of people with diabetes contributed to their susceptibility to foot ulcers [33]. This discrepancy could potentially be attributed to the fact that plantar tissue thickness is impacted by additional factors, such as BMI and age [29, 34].

Stiffness, indicating the elasticity of a material when a compressive force is applied, is reflective of the biomechanical properties of soft tissues. Alterations in the stiffness of plantar tissues can interfere with their natural function as shock reducers and absorbers [20, 21, 35]. Our research found that the DPN group exhibited significantly stiffer plantar tissues at all measured locations compared to the HC group. These findings align with previous studies assessing plantar stiffness using ultrasound palpation systems and Shore durometers [7, 20, 21, 25, 34]. Sun's [21] results, closely mirroring ours. Meanwhile, Chao's [25] research highlighted that plantar stiffness was more pronounced in patients with advanced progression of diabetic ulcers. Duan's [20] study revealed that the tissue stiffness in the hallux, forefoot, midfoot, and rearfoot regions was notably higher in DPN patients compared to non-DPN patients when the patient's activity load was elevated (p < 0.05). The findings from all the aforementioned studies collectively indicate that the severity of plantar stiffness intensifies with the progression of diabetes and peripheral neuropathy. In contrast, Naemi et al. [24] utilized real-time strain elastography, revealing a lower relative heel stiffness in ulcerated patients compared to nonulcerated. They measured heel stiffness at the non-ulcer site in patients with active ulcers, and it is worth discussing that the tissue edema from acute inflammation during ulcer presentation could distort the true heel stiffness. Also, strain elastography's repeatability and stability are inferior to SWE.

Numerous studies highlighted the link between plantar stiffness changes and the glycosylation response in those at risk for diabetic foot. Glycosylation, a glucose-mediated intermolecular cross-linking process, contributes significantly to tissue aging [32]. These crosslinks typically result in tissue dysfunction, reducing critical flexibility and permeability, and decreasing metabolism [30]. It is the alteration of tissue properties by glycation (e.g., causing the tissue to stiffen) that triggers many devastating late complications [10, 23, 30, 31].

Another category of research emphasizes that changes in plantar stiffness are closely related to plantar loading [20, 35, 36]. If plantar stress exceeds the tissue's carrying capacity, injury can result. From an engineering perspective, tissue damage alters its mechanical properties, which means that changes in tissue biomechanics as a result of damage can be used to detect injury. An in vivo model devised by Chatzistergos found that repeated loading to the pain threshold resulted in statistically significant persistent stiffness in the overloaded area. Whereas the pain threshold is raised when neuropathy occurs, prolonged overloading of the plantar foot is destined to stiffen the tissue [35]. In healthy populations, runners with greater plantar loading have softer heels compared to cyclists [36], but studies in people with diabetes have shown the opposite [20]. This disparity suggests that plantar tissue characteristics may be altered in people with diabetes, who have stiffer plantar tissue, which reduces the ability to distribute pressure over a wider area to reduce high stress, leading to reduced adaptive capacity.

Similar to other studies, patients with DPN had significantly higher plantar tissue stiffness at multiple sites compared to healthy controls [21, 25]. However, the locations of significant differences in plantar stiffness between people with or without neuropathy in diabetes were inconsistent, with only the 5th MTH showing higher stiffness in the DPN group in our study. Differences in plantar nerve distribution may explain this phenomenon to some degree [11-13]. Strzalkowski summarized the estimates of plantar innervation density with a synthesis of the relevant literature, concluding that the highest density of plantar innervation was in the toes (23.3 units/cm²), followed by the lateral arch (15.4 units/cm²) and the lateral metatarsals (11.2 units/cm²) [12]. In this study, the 5th metatarsal head belongs to one of the regions with the highest density of nerve distribution, and its changes may be most pronounced when neuropathy occurs, as we found, leading to the plantar tissues stiffening first. Some studies have been designed to sum the plantar stiffness values from multiple locations of the plantar to consider the plantar as a whole for research purposes [20]. Based on the theory that the differences in nerve distribution [12, 13] and the unbalanced degree of denervation produced during the development of diabetic neuropathy [37], it can be inferred that the changes in the biomechanical properties of the plantar foot are also uneven. At the same time, our findings indicated significant differences in thickness and stiffness measurements across four different plantar tissue sites. Variability is more pronounced in people with diabetes compared to healthy individuals. Therefore, when studying the biomechanical properties of the plantar tissues, it may be more meaningful to consider the plantar tissues as a whole and take into account the weighting of the different locations.

Our study is the first to employ PCA to integrate thickness and stiffness data from multiple plantar locations by projecting the data onto principal feature vectors, thereby reducing the dimensionality of the original dataset while preserving useful information and providing a more comprehensive assessment of tissue characteristics. The PCA technique combines subsets of highly correlated but distinct variables to form factors that summarize the information contained in a continuous multivariate data set [38, 39]. The integrated stiffness index Z1SWE, which explained 49.07% of the total variance of the stiffness variables of plantar tissue, differed significantly between the DPN and DM groups (p = 0.02). Z1SWE integrated the stiffness of four tested locations, which, to some extent, reflects the stiffness parameters of the entire foot and is more valuable than a single site.

Clinical parameters are of undeniable importance in screening populations at high risk for diabetes complications [4, 40, 41]. In our univariate logistic regression model, age, gender, BMI, duration of diabetes, and cholesterol level were significant risk factors for DPN. Research on diabetic foot ulcers summarized that age was the first-degree risk factor, and male and duration were third-degree risk factors [3, 4, 41]. A population-based cross-sectional study in China examining the prevalence and risk factors of DPN found that patients with DPN were significantly older compared to those without DPN [median age (Q1, Q3) was 63 (55, 71) years versus 59 (50, 67) years, p < 0.001]. They also had a longer duration of diabetes [median duration (Q1, Q3) was 8 (5, 14) years versus 6 (3, 11) years, p < 0.001], and a higher proportion of males was observed (38.4% vs. 19.2%, p < 0.002) [42]. The age-standardized male-to-female ratio of years lived with disability due to neuropathy, amputation, and foot ulcers increased progressively, rising from 0.96 to 1.93, further indicating a higher risk among men compared to women [43]. Oguejiofor et al. showed that 100% of diabetic patients with a disease duration of more than 15 years had peripheral neuropathy, regardless of whether they had symptoms of neuropathy [40]. Skin biopsies directly showed that intraepidermal nerve fiber density was negatively correlated with the duration of diabetes [10]. Age, hyperglycemia status, and duration of diabetes could accelerate tissue aging, probably all based on the process of glycation, which is a major cause of tissue dysfunction [29, 30].

The integration of plantar stiffness index with gender and diabetes duration offers a novel and comprehensive approach for explaining DPN. The model shows that a one-unit increase in Z1SWE in plantar tissue raises the chances of developing DPN by 179% (i.e., EP(b) = 2.792), and a one-year increase in diabetes duration ups the risk by 68% (i.e., EP(b) = 1.676). Men with diabetes are 11.5 times more susceptible to DPN compared to women (i.e., EP(b) = 0.087). This noninvasive method could serve as a valuable tool for DPN risk assessment in clinical practice, particularly in resource-limited settings where advanced diagnostic tools may not be readily available. By identifying individuals at high risk of DPN, this model could facilitate timely interventions, potentially reducing the incidence of DPN-related complications such as foot ulcers and amputations.

This study has several limitations. First, although it is the first to integrate stiffness information from four plantar sites to and explain DPN, the principal components accounted for only 49.07% of the total stiffness information, likely due to the limited number of sites measured. Second, this study has a cross-sectional design, limiting its findings to explaining the presence of DPN and its associated factors at the time of data collection, without establishing causality or predictive relationships. Additionally, the small sample size may affect the generalizability of the results. Future longitudinal and larger multi-center studies are needed to validate the model, establish temporal relationships, and further explore the link between plantar biomechanical properties and DPN.

In conclusion, this study revealed that individuals with diabetes, particularly those with DPN, exhibit thicker and stiffer plantar tissues at specific locations compared to healthy controls. The integrated stiffness index, derived from multiple plantar locations through PCA, effectively distinguished between individuals with and without DPN. By incorporating clinical parameters such as diabetes duration and gender, we identified that longer diabetes duration, male sex, and a higher integrated stiffness index are associated with an increased likelihood of DPN. These findings suggest that plantar tissue characteristics, combined with clinical data, provide a comprehensive approach for describing and explaining DPN. From a broader perspective, the integration of readily obtainable clinical parameters and plantar stiffness measurements offers a potential screening tool for DPN, which could enhance the precision of prevention and intervention efforts, particularly for high-risk populations.