This prospective, observational cohort study was conducted at 2 tertiary centers in Taiwan: National Taiwan University Hospital, Hsinchu Branch, and Shin Kong Wu Ho-Su Memorial Hospital. Patients with ESRD undergoing maintenance HD were screened for inclusion. Exclusion criteria included age under 20 years, prior lower-limb amputations (minor or major), and refusal to provide informed consent. Each lower limb was evaluated as an independent unit. Following enrollment, data on participant characteristics, biochemical profiles, and ABI measurements were collected. The presence of severe PAD was confirmed using DUS as the gold standard, defined by monophasic or absent flow in the distal posterior tibial artery (PTA) and distal anterior tibial artery (ATA) at 5 cm above the ankle.

This study was approved by the Institutional Review Board (IRB) of both the National Taiwan University Hospital, Hsinchu Branch (IRB number 111‐105-E), and Shin Kong Wu Ho-Su Memorial Hospital (IRB number 20230702D); written informed consent was obtained from all participants.

Following study enrollment, a systematic review of electronic medical records and baseline clinical assessments of the participants was conducted. Demographic information, including age and sex, and a detailed history of comorbid conditions such as hypertension, diabetes mellitus (DM), hyperlipidemia, CAD, congestive heart failure, and prior cerebrovascular or myocardial events were recorded. Current medication profiles were also documented, specifically the use of erythropoietin, vitamin D analogs, antihypertensive agents, and iron supplements. Comprehensive laboratory data were gathered, including complete blood counts, serum albumin, and markers of dialysis adequacy such as predialysis and postdialysis blood urea nitrogen and Kt/V. Additionally, we recorded metabolic indicators including fasting blood glucose, lipid panels, and bone mineral markers such as calcium, phosphorus, and intact parathyroid hormone levels. As part of the initial vascular screening, the ABI was measured for each limb.

To record skin temperatures, contact thermometers were placed on the instep and sole of both feet, as well as the wrist of the nonaccess arm (Figures 1 and 2). The instep (dorsal midfoot) overlies the anterior tibial and dorsalis pedis angiosomes, whereas the sole reflects the posterior tibial angiosome; monitoring both sites therefore samples the 2 principal tibial-artery territories assessed by DUS, and these acral sites are among the earliest to display perfusion-related thermal changes. We used Temp Pal (iWEECARE Co Ltd), a wireless, mobile health (mHealth)–based thermometer system that allows for simultaneous, multisite continuous monitoring. Five devices were paired via Bluetooth to a single iOS mobile device running the Temp Pal app, which captured and uploaded continuous temperature data to a cloud-based server for export. This system uses an autocalibration algorithm to ensure a high degree of measurement reliability, maintaining accuracy and precision within ±0.05 °C. Baseline body temperature was measured at the start and end of each HD session using an infrared ear thermometer, with simultaneous readings obtained from the skin sensors. Throughout the entire procedure, skin temperatures at the wrist and both feet were monitored continuously to capture fluctuations related to intradialytic hemodynamic shifts. Furthermore, clinical parameters, including dialysate temperature, blood pressure, and heart rate, were systematically recorded during the session.

To evaluate the presence and severity of PAD, all participants underwent DUS performed by experienced technicians. DUS served as the gold standard for this study, specifically focusing on flow patterns in the tibial arteries. Measurements were obtained from the distal PTA and ATA at a standardized location 5 cm above the ankle. Severe PAD was defined by the identification of monophasic or absent flow in either the distal PTA or ATA. These flow patterns were contrasted with normal (triphasic or biphasic) flow to categorize the limbs for comparative analysis of intradialytic temperature shifts.

Inferential analyses were conducted at the limb level; 109 patients contributed 218 limbs eligible for analysis. Temperature parameters were compared between limbs with normal vs abnormal distal PTA/ATA flow using generalized estimating equation (GEE) with a normal distribution and an identity link, accounting for the intrapatient correlation between the 2 limbs of the same individual by using robust (empirical) standard errors. We further performed multivariable GEE analyses to investigate the independent associations of each temperature parameter and ABI value with abnormal distal PTA/ATA flow. The models were adjusted for age, DM, CAD, congestive heart failure, Kt/V (Daugirdas formula), and calcium × phosphorus product. We intentionally limited the adjustment to 6 confounding variables to avoid model overfitting, given that the number of events was 51 for abnormal distal PTA flow and 48 for abnormal distal ATA flow.

Discriminative performance of abnormal distal PTA/ATA flow was evaluated using receiver operating characteristic (ROC) curve analysis, with area under the curve (AUC) calculated as the primary metric. The 95% CI for the AUC was derived using DeLong nonparametric method. When the AUC was statistically significant (ie, the 95% CI did not include 0.5), the optimal cutoff was determined using the Youden index to balance sensitivity and specificity. To determine if temperature monitoring offers incremental diagnostic value beyond the ABI, the statistical significance of differences between the respective AUC values was assessed via DeLong test. Statistical significance was defined as a 2-sided P value of less than .05. All analyses were conducted with MedCalc Statistical Software (version 20.019; MedCalc Software).

The study included 109 patients with ESRD undergoing maintenance HD. The mean age of the participants was 65.2 (SD 11.9) years, and 56.9% (n=62) were male participants. Comorbid conditions were highly prevalent in this cohort, with hypertension (n=89, 81.7%) and DM (n=61, 55.9%) being the most frequent, followed by gout (n=29, 26.6%), CAD (n=26, 23.9%), and congestive heart failure (n=25, 22.9%). A smaller proportion of patients had a history of malignancy (n=6, 5.5%), cerebrovascular accidents (n=4, 3.7%), and myocardial infarctions (n=2, 1.8%). Regarding pharmacological management, the majority of patients were receiving erythropoietin (n=97, 88.9%) and iron supplements (n=84, 77.1%), while 58.7% (n=64) were on antihypertensive medications. Laboratory assessments revealed a mean hemoglobin of 11.6 (SD 1.2) g/dL and mean albumin of 3.99 (SD 0.37) g/dL. Dialysis adequacy was reflected by a mean Daugirdas Kt/V of 1.62 (SD 0.21). Biochemical markers related to bone and mineral metabolism showed a mean calcium of 8.92 (SD 0.98) mg/dL, mean phosphorus of 5.03 (SD 1.35) mg/dL, and median intact parathyroid hormone level of 290.7 (IQR 106.1-570.8) pg/mL. Iron status was characterized by a mean transferrin saturation of 27.9% (SD 11.8%) and median ferritin level of 567 (IQR 333-738) ng/mL. Detailed demographic, clinical, and biochemical data are summarized in Table 1.

A total of 218 limbs were evaluated to analyze the relationship between distal arterial flow and various physiological parameters (Table 2). The parameters analyzed comprised the ABI and the temperature parameters, namely, wrist-to-foot and ear-to-foot temperature differences at the end of HD and absolute foot temperature before and at the end of HD. Limbs with abnormal flow in either the distal PTA or ATA exhibited significantly lower ABI values compared to those with normal flow (mean 0.90, SD 0.26 vs mean 1.13, SD 0.20 for PTA; mean 0.89, SD 0.27 vs mean 1.13, SD 0.200 for ATA; both P<.001). However, it is noteworthy that despite these significant differences, limbs with confirmed severe PAD still exhibited mean ABI values as high as 0.89 (SD 0.27) to 0.90 (SD 0.26), which border the clinically normal range and highlight the limitations of ABI in representing disease severity in this population.

Regarding intradialytic temperature dynamics, the temperature difference between the wrist and the foot at the end of the HD session was significantly smaller in limbs with abnormal flow (Table 2). Specifically, at the sole, the wrist-foot temperature difference was mean 1.05 (SD 3.05) °C in the abnormal PTA group compared to mean 2.23 (SD 2.57) °C in the normal group (P=.02). A similar pattern was observed at the instep, reaching statistical significance for ATA flow (P=.046) but showing only a borderline association for PTA flow (P=.051). While foot temperatures before the HD session did not differ significantly between groups, the foot temperature at the end of the session tended to be higher in limbs with abnormal ATA flow compared to those with normal flow (mean 31.41, SD 2.55 °C vs mean 30.48, SD 2.46 °C at the instep; P=.08; mean 31.20, SD 2.36 °C vs mean 29.96, SD 2.61 °C at the sole; P=.02).

Additionally, a multivariable GEE model, adjusting for age, DM, CAD, congestive heart failure, Kt/V (Daugirdas formula), and the calcium × phosphorus product, revealed that an increased ABI was significantly associated with a reduced risk of abnormal distal PTA flow (odds ratio [OR] 0.73, 95% CI 0.63‐0.83) and ATA flow (OR 0.71, 95% CI 0.59‐0.85; Table 3). Moreover, wrist-to-foot temperature differences at the end of the HD session, evaluated at the instep and sole, respectively, were significantly associated with abnormal distal PTA flow at the sole (OR 0.81, 95% CI 0.68‐0.97), with a borderline, nonsignificant association at the instep (OR 0.84, 95% CI 0.70‐1.003), and with abnormal distal ATA flow at both the instep (OR 0.84, 95% CI 0.71‐0.99) and sole (OR 0.81, 95% CI 0.70‐0.95). A similar pattern of results was observed for foot temperature at the end of the HD session.

The diagnostic utility of intradialytic temperature monitoring was evaluated for its ability to predict an ABI value of less than 0.9 (Table 4). ROC analysis revealed that the temperature difference between the wrist and foot at the end of the HD session was the most effective parameter for this purpose. At the instep, this gradient achieved an AUC of 64.5% (95% CI 57.5%‐71.0%). Similarly, the gradient measured at the sole demonstrated a significant AUC of 64.2% (95% CI 57.2%‐70.8%; Figure 3). For the instep measurement, an optimal cutoff of 1.25 °C or less yielded a sensitivity of 63.2% (95% CI 46.0%‐78.2%) and a specificity of 60.0% (95% CI 52.1%‐67.5%). At the sole, a cutoff of 2.83 °C or less provided a higher sensitivity of 81.6% (95% CI 65.7%‐92.3%) but lower specificity of 42.4% (95% CI 34.8%‐50.3%). Other parameters, including absolute foot temperatures before or after HD and the ear-foot temperature gradient, did not reach statistical significance in discriminating limbs with an ABI value of less than 0.9 (Table 4).

ROC analysis was performed to further evaluate the diagnostic utility of various temperature parameters for identifying severe PAD (Table 5). The temperature difference between the wrist and the foot at the end of the HD session emerged as a significant indicator of arterial flow abnormalities. Specifically, for the detection of abnormal distal PTA flow, the wrist-foot temperature difference at the sole yielded an AUC of 62.0% (95% CI 55.2%‐68.5%; Figure 4). At the instep, this parameter showed an AUC of 59.1% (95% CI 52.3%‐65.7%). Notably, using a cutoff value of –2.2 °C or less for the wrist-foot difference at the instep provided a high specificity of 97.6% (95% CI 94.0%‐99.3%) for identifying abnormal PTA flow. The temperature parameters demonstrated even greater discriminative power for abnormal distal ATA flow. The wrist-foot temperature difference at the end of dialysis showed an AUC of 64.5% (95% CI 57.8%‐70.9%) at the sole and 62.0% (95% CI 55.2%‐68.5%) at the instep (Figure 5). Furthermore, absolute foot temperature at the end of the HD session was also a significant indicator of abnormal ATA flow, with an AUC of 64.5% (95% CI 57.7%‐70.8%) at the sole and 61.5% (95% CI 54.7%‐68.0%) at the instep. According to Youden index, a sole temperature cutoff of greater than 30.44 °C at the end of HD yielded a sensitivity of 70.8% (95% CI 55.9%‐83.0%) and a specificity of 58.2% (95% CI 50.4%‐65.7%) for detecting abnormal ATA flow (Table 5).

Lastly, we performed DeLong test to examine whether the addition of temperature monitoring yields incremental diagnostic value over the baseline ABI (Table 6). Regarding abnormal distal PTA flow, temperature parameters at both the instep and sole failed to offer incremental diagnostic value over an ABI value of less than 0.9. Conversely, relatively stronger diagnostic utility was observed for abnormal distal ATA flow. For the instep measurements, although the improvements did not achieve statistical significance, the ΔAUC values increased by roughly 1% (95% CI –5.5% to 7.5%) to 4% (95% CI –2.1% to 10.1%). Notably, for the sole measurements, the wrist-to-foot temperature difference at the end of dialysis yielded a significant increase in AUC (ΔAUC 5.6%, 95% CI 0.2%-11.0%). Additionally, a potential trend toward improved AUC was noted for foot temperature at the end of the HD session (ΔAUC 5.1%, 95% CI –1.0% to 11.2%).

In this prospective cohort study of patients with ESRD undergoing maintenance HD, we explored the diagnostic potential of intradialytic foot temperatures for detecting severe PAD. First, we found that in this population, where arterial calcification is prominent, the ABI may remain near-normal despite the presence of severe PAD as confirmed by DUS. Second, absolute foot temperatures at the end of the dialysis session were significantly higher in patients with severe ATA disease compared with those without. Finally, the temperature gradient between the wrist and foot served as a significant predictor of severe ATA or PTA disease, achieving an AUC of up to 64.5% (95% CI 57.8%-70.9%) for abnormal ATA flow. These results suggest that monitoring foot temperatures during HD provides an accessible and noninvasive means of screening for PAD in the vulnerable ESRD population. Nonetheless, the discriminative ability of these temperature parameters was modest (AUCs of approximately 0.59‐0.65), and because all estimates were derived and tested within the same cohort without internal or external validation, these findings should be interpreted as demonstrating an association rather than establishing a standalone diagnostic tool.

The diagnosis of PAD in patients with ESRD remains a significant clinical challenge due to the inherent limitations of conventional screening tools. The ABI, while widely used as a first-line diagnostic test, frequently yields falsely elevated or “normal” results in the HD population [17,18]. This is due to extensive arterial calcification, which renders the crural vessels noncompressible and prevents the accurate measurement of distal blood pressure [8]. This leads to a high rate of underdiagnosis in a population that is otherwise at extreme risk for CLTI and amputation. Furthermore, renal insufficiency is frequently associated with a more distal, crural distribution of atherosclerosis, which further complicates pressure-based assessments that are more sensitive to proximal disease. In our cohort, limbs confirmed to have severe PAD via DUS still exhibited mean ABI values as high as 0.89 (SD 0.27) to 0.90 (SD 0.26), suggesting that relying on ABI alone would have missed a substantial proportion of patients with severe tibial artery disease. We also demonstrated the feasibility of detecting severe PAD through the temperature gradient between the wrist and foot at the end of the HD session. For example, an instep wrist-foot temperature difference cutoff of –2.2 °C or less achieved a remarkable specificity of 97.6% (95% CI 94.0%-99.3%) for detecting abnormal flow in the PTA. By leveraging the hemodynamic stress during the dialysis process, this method provides a functional assessment of perfusion that is independent of vessel wall compressibility. Intradialytic thermography offers a potential screening adjunct, identifying high-risk patients who would otherwise be cleared by traditional, yet possibly less reliable, pressure-derived measurements.

Based on the established relationship between thermodynamics and hemodynamics, the use of skin temperature during HD as a proxy for arterial perfusion provides a novel physiological entry point for detecting PAD in patients with ESRD. Skin temperature has been previously recognized as an indicator of cutaneous blood flow, with infrared thermography and contact sensors being used to map ischemic zones in high-risk populations [19-21]. Prior studies have demonstrated that abnormal skin temperatures in the feet correlate significantly with reduced ABI values and abnormal limb arterial ultrasonography [20,22]. Our team previously reported that a significant postprocedural increase in foot temperature serves as a robust predictor of wound healing after EVT [16], reinforcing that thermal changes can directly reflect perfusion. However, interpretation of foot temperatures varies in different settings and can be challenging. In patients with DM, focal increases in foot temperatures are predictive of ulceration [15,23], while globally warmer feet suggest autosympathectomy related to diabetic neuropathy, with subsequent arteriovenous shunting and increased blood flow to the skin [19,22]. Thus, absolute foot temperature alone may be confounded and insufficient to reflect PAD severity. Similar to the use of myocardial stress testing to unmask underlying CAD, this present study advances this diagnostic concept by leveraging HD as a “thermal stress test” for the peripheral vasculature. To ensure diagnostic precision, the temperature at the wrist of the nonaccess arm served as a proximal, well-perfused reference. This approach allowed for the isolation of local thermal changes elicited specifically by vascular stress, thus controlling for systemic temperature fluctuations related to the dialysate itself. We found that while baseline temperatures before dialysis were similar between groups, the temperature at the end of the session significantly differentiated those with severe PAD. Specifically, the wrist-foot difference at the sole effectively discriminated abnormal flow in both the PTA and ATA. Interestingly, it is a lack of temperature decrease during HD that identifies severe PAD in our cohort (ie, those with severe PAD have higher foot temperatures). This paradoxical finding conforms with literature [18,19], suggesting disrupted vasoconstriction mechanisms in these patients. Our study demonstrates that wearable mHealth sensors provide a physiological foundation for future research into the underlying mechanisms of PAD in patients with ESRD.

Socioeconomic disparities significantly delay the diagnosis of PAD, leading to higher amputation rates in underserved populations [24]. Patients with ESRD, in particular, are frequently overlooked, with limited access to specialized care due to ambulatory difficulties and socioeconomic disadvantages [7]. Traditional screening for PAD via ABI or DUS is often limited by accessibility, resulting in a reactive approach where referral is initiated only after ulcers or gangrene appear. Intradialytic foot temperature monitoring offers a solution to overcome these barriers by using the Temp Pal, a wearable mHealth thermometer system that provides high-fidelity, continuous monitoring with minimal overhead. While studies have demonstrated the feasibility of smart-socks [21,25], such specialized apparel is unavailable in many regions, including Taiwan. In contrast, cutaneous thermometers are significantly cheaper, highly accessible, and reusable, making them a more practical choice for global implementation. More importantly, integrating this monitoring method into routine dialysis sessions allows patients to undergo screening without additional travel or time commitments, ensuring accessibility to those with limited mobility or resources. Our ultimate goal is to mitigate health disparities within the high-risk population with ESRD, thereby improving both the trajectory of vascular disease and long-term survival. These potential benefits, improved screening access, reduced disparities, and earlier identification, were not evaluated in the present study and remain hypotheses to be tested in future implementation and outcome studies.

While this study offers a novel approach to PAD screening, several limitations must be acknowledged. First, the sample size of 109 patients is relatively small and was recruited from 2 tertiary centers within a single geographic region. As our cohort consisted exclusively of an East Asian population in Taiwan, the generalizability of our findings to other ethnic groups remains uncertain. Given that racial and socioeconomic disparities significantly influence the presentation and progression of CLTI, further research in more diverse populations is necessary to validate our findings. Second, we used DUS flow patterns at 5 cm above the ankle as our gold standard for defining severe PAD. Although DUS is a clinically standard and reliable tool for assessing arterial patency, it primarily evaluates macrovascular flow. We did not incorporate other perfusion markers, such as transcutaneous oxygen tension or skin perfusion pressure, which might have provided a better understanding of how intradialytic temperature shifts correlate with actual tissue oxygenation. Finally, the sensitivity of intradialytic temperature monitoring for detecting severe tibial artery flow abnormalities varied considerably across parameters and measurement sites, with a reciprocal trade-off in specificity. In addition, this study was not designed to assess long-term outcomes. Although early detection is critical in PAD, there is a lack of evidence that this screening method leads to improved limb salvage rates or survival benefits. Future longitudinal studies are required to determine if initiating early interventions based on these thermal warning signs can successfully reduce major adverse limb events, major adverse cardiovascular events, or mortality associated with PAD in the population with ESRD. In addition, as most patients were asymptomatic and were screened irrespective of symptoms, formal clinical PAD staging (eg, Fontaine or Rutherford categories) was not performed; disease severity was instead defined objectively by DUS flow patterns. Moreover, several factors may have affected the accuracy of our temperature measurements: the ambient temperature of the dialysis unit was not systematically recorded, the access (fistula) arm was not measured, and a single nonaccess arm was used as the wrist reference for both legs, which could introduce bias if that reference limb itself had impaired perfusion.

This study demonstrated that intradialytic foot temperature monitoring is a potential diagnostic adjunct for detecting severe PAD in patients with ESRD. We found that in patients with ESRD, the ABI can remain near-normal despite severe disease, whereas the wrist-foot temperature difference at the end of the HD session was significantly associated with abnormal tibial flow, although sensitivity and specificity varied considerably across temperature parameters. Temperature monitoring with mHealth devices during HD may provide an accessible, low-cost screening solution that mitigates health disparities in patients with ESRD. Further longitudinal studies are required to confirm the survival benefits of this novel method. Because the observed discriminative ability was modest and has not yet been externally validated, intradialytic temperature monitoring should be regarded as a complementary adjunct to, rather than a replacement for, established vascular assessments such as the ABI.