Study protocol for a randomized controlled trial to test for preventive effects of diabetic foot ulceration by telemedicine that includes sensor-equipped insoles combined with photo documentation

According to the definition of the World Health Organization (WHO), diabetic foot syndrome (DFS) encompasses all foot complications, constituting an “ulceration of the foot (distally from the ankle and including the ankle) associated with neuropathy and different grades of ischemia and infection” [1]. It increases the risk of limb amputation, and even mortality, if left untreated [2]. In Germany, about 40,000 legs, feet, or toes are amputated, with 70% of major amputations and 85% of minor amputations due to DFS. In addition, foot lesions in diabetic patients impose an enormous social and economic burden across the world. In the US, Rogers et al. reported that $18 billion were spent on the care of diabetic foot ulcers (DFUs) and $11.7 billion sum up as consequences of lower extremity amputations [3].

Among the reasons for DFS, diabetic foot neuropathy is the major contributing factor for foot complications (50% as a single cause, 30–50% as a cause in combination with angiopathy [4]), because it affects the ability of the foot to feel and sense [5‐7]. This is why patients with diabetic neuropathy are not able to realize injuries to their feet. Most of the complications develop due to infection and ulceration in the foot [8, 9]. The early signs of DFS include fissures, blisters, abundant callus formation, redness, and increased temperature [10]. A physician may diagnose the exact cause by analyzing these physical features [11].

It is possible to delay or even avoid the development of DFUs with adequate treatment at early stages. Usually, clinicians assess the general condition through analyzing ankle brachial pressure indices, plantar pressure profiles, and testing for foot neuropathy [12]. Additionally, advanced technologies like corneal confocal microscopy, magnetic resonance tomography, and Doppler ultrasonography provide tools to diagnose the prevalence of peripheral neuropathy and angiopathy, foot ulcers, and its risks [13]. However, these methods are considered intrusive and are costly; patient compliance is lacking, especially with frequent doctor’s visits [2]. On the other hand, patient self-assessment has limitations such as lack of knowledge about this condition, difficulties using specialized equipment, and impaired physical mobility. More effective and advanced approaches need to be investigated to provide flexible and comprehensive foot care for patients at risk for the DFS.

Elevated plantar temperatures have been reported to be an early sign of incipient DFUs. In the studies of Lavery et al. and Armstrong et al., home temperature monitoring and reduced activities have been verified to be effective to reduce the incidence of DFUs in high-risk patients [14, 15]. In the study of Lazo-Porras et al., the effectiveness of foot thermometry (TempStat™ for thermal image capture) to prevent DFUs was investigated, together with mHealth reminders (SMS and voice messaging), in an evaluator-blinded randomized 12-month trial. The authors highlighted the importance to evaluate adherence to daily home-based measurements [16]. Furthermore, a left-to-right foot temperature difference of > 2.2 °C as a proposed threshold for an impending ulceration has been investigated comprehensively by Wijlens et al. in 20 patients with diabetes and peripheral neuropathy. Their conclusion was that the > 2.2 °C threshold is only acceptable if it is confirmed after 24 h in a repeated measure and if, in addition, the temperature difference is individually corrected depending on baseline measurements [17]. In addition to neuropathic ulcers [18, 19], one has to consider osteomyelitis [20, 21] and the disease termed Charcot foot [22] as differential diagnoses in the case of elevated plantar temperatures.

On the other hand, decreased foot temperatures may point to a vascular insufficiency in the foot [23]. Therefore, foot temperature monitoring with thermometers, thermal imaging techniques, wearable temperature techniques (socks, insoles, and shoes) has been widely tested to date. For example, Netten et al. explored the temperature discrimination thresholds between “no,” “local,” or “diffuse” DFUs with a high-resolution infrared thermal imaging technique [24]. Fraiwan et al. implemented a mobile thermal imaging system with an automated method to identify possible ulcers in diabetic patients [11]. These pioneering works may open a window for patients to check for their foot condition in a feasible and comfortable fashion in the future.

Moreover, in the study by Fryberg et al., a novel smart mat technology was evaluated for predicting impending DFUs in a 34-week cohort study that enrolled 132 patients with diabetes. Their results support the notion that the remote temperature-monitoring system could be a feasible and efficient strategy to early identify DFUs, but the asymmetry thresholds have a significant influence on the sensitivity and specificity. Comparing the 2.22 °C and 3.20 °C thresholds, sensitivity decreased from 97% to 70%, but the specificity increased from 43% to 68% [25]. Therefore, effective and convenient means of temperature measurements such as home-based wearable technologies, accurate and reasonable early warning mechanisms with disparate asymmetry thresholds, followed by timely and appropriate interventions are the main research focus in this field.

From our perspective, home-based monitoring of plantar foot temperatures may be regarded as an effective method in the early detection and possible prevention of DFUs. In this study, by utilizing a novel sensor-equipped insole, we aim to establish a telemedicine structure with a remote server and the corresponding smartphone app to timely monitor changes of plantar foot temperatures in diabetes patients. The evidence obtained will include a set-up with predefined standardized temperature recordings and a telemedicine aspect allowing for feedback and alarming as well as picture recordings. The outcome of our study will ultimately allow us to determine if and to which extent such an effort may reduce the number of diabetic foot ulcerations and other medical foot conditions in such a cohort.

This study makes three principal contributions concerning the prevention of DFUs. First, the introduction of sensor-equipped insoles to promote daily home-based measurements of foot temperatures. Second, the implementation of a telemedicine structure with a smartphone app to measure foot temperatures, provide photo documentation, and evaluate wellbeing (quality of life) using an interactive diary. These collected data will be transferred to a remote server for interpretation and adjustment of intervention measures. Thus, our system appears much more sophisticated and provides more reliable data compared to simple thermometric approaches. Ultimately, intelligent predictive models for DFUs will be built with the collected sensor data and interpretations, which may support medical care providers.

Instead of using a thermometer (TempTouch; Xilas Medical, San Antonio, TX, USA) [14, 15] or thermal imaging devices (TempStatTM) [16], our study innovatively introduces the sensor-equipped insole to help diabetic patients to perform daily home-based monitoring of foot temperatures. The insole can easily be inserted into house slippers or shoes and may record the temperature data continually for several hours if required. It provides a more convenient and comfortable way for frequent temperature measurements.

The telemedicine structure implemented in our study comprises a remote server as core controller in the study center and the smartphone application (SPDFA) as data collecting terminal. With the SPDFA, patients can immediately comprehend the initial analysis results of their measurements. Sensor data will be transmitted from the SPDFA to the study server, together with the initial evaluations, the requested photo series (at alarm levels 3 and 4) (Fig. 5), and a self-assessment about wellbeing and foot status using our interactive diary (Fig. 6). Compared to the approach by Lazo-Porras et al., patients in our study do not need to identify the pre-defined alarm signs by themselves and consult the study physicians or nurses for timely interpretation [16]. For medical interpretation, our approach provides more information by means of photo documentation and using a wellbeing score, instead of only collecting temperature data. The study server stores the data and provides physicians with an interface to visualize the status of the patients and to interpret the ulcer alarms. In the case of a confirmed alarm, the study server can exchange data with the SPDFA to perform suitable intervention measures for the patient and to collect the patient’s interactive feedback during intervention periods. This approach will evaluate the effectiveness of activity reduction in order to delay or even avoid the development of DFU. Based on this concept, efficiency and timely interventions will be significantly improved. In addition, our alarm algorithms with stepwise graded alarm levels are able to test and verify various temperature warning measures (apart from only measuring temperature differences between left and right corresponding sensor sites) [16], time intervals between two alarm levels, individual corrections based on baseline data [17], or even different asymmetry thresholds [25].

Based on these collected data and clinical interpretations, intelligent predictive models might be built in the future for machine learning algorithms. With the development of such algorithms, intelligent telemedicine technologies have already proven to be one of the most cost-effective solutions for the early detection of DFU. As exemplified in the study of Goyal et al., deep learning methods for real-time DFU localization were applied to an extensive database of 1775 images of DFUs. The deep learning model showed great potential in the real-time localization of DFUs on an NVIDIA Jetson TX2 and a smartphone app [36]. The data collected in the present study will be important to test for an alarming system with a preset temperature threshold, compliance of diabetes patients to a bi-daily recording rhythm, and the challenges of picture recordings with a mobile app. Therefore, a whole package of innovation is brought to the intervention group participants; however, an entire telemedicine system with auto-response of the database recording system is not yet intended. The study physician interprets the data at 24-h intervals.

In subsequent studies, we will be able to test for different thresholds concerning temperature and alarm evaluation. This will allow us to furthermore adjust algorithms to detect other temperatures abnormalities caused, e.g. by Charcot foot, or vascular insufficiency. Ultimately, machine-learning algorithms and decision tree classification will be used to train an automated predictive model of DFUs with the data that are collected in past periods.

We believe that the complexity of the retrieved data from our protocol offers the potential to tackle a difficult problem from a unique aspect and, therefore, possibly will have a substantial impact on DFUs prevention not only in Germany but also in many other parts of the world.

This manuscript is based on version 1.6 of the trial protocol, dated 18 February 2019. Recruitment for this study began on 30 January 2018 and should be completed by 30 December 2019.

At the time of submission, our study has already recruited 196 patients; 87 patients were randomized into the intervention group. Of the 87 patients, 72 are active by daily measurements of foot temperatures with our system. The study is widely known in the area of Saxony-Anhalt; currently, a growing number of people with diabetes are eager to participate in the trial.