Autonomic neuropathy is an incurable complication of diabetes with multiorgan presentation, including the cardiovascular system, called cardiovascular autonomic neuropathy (CAN). Prevalence estimates of CAN vary widely ranging from 12-73% and it contributes to the excess mortality from cardiovascular diseases (1, 2). The cardiovascular system is regulated bidirectionally by the brainstem, controlled by the sympathetic and parasympathetic nervous systems. When this regulation is impaired, the ability to constantly adapt to the physiological and environmental changes is diminished, potentially leading to hemodynamic instability, arrhythmias, and silent ischemia (3–5). The prevalence of diabetes is increasing at an alarming rate, particularly type 2 diabetes (T2D) (6), and with it the number of complications.

The pathophysiology underlying CAN is that the parasympathetic branch of the autonomic nervous system is initially affected, leading to parasympathetic withdrawal and sympathetic compensation, which ultimately progresses to sympathetic withdrawal (7). The late stage of CAN is considered irreversible due to the nerve damage being too severe at this point (8, 9). This emphasizes that the early stages with parasympathetic withdrawal should be of interest to increase early detection, where lifestyle and cardiovascular risk modifications are still effective (10).

The American Diabetes Association recommend annual assessment of autonomic neuropathy, whereas Danish guidelines only advise evaluation at clinical suspicion (11, 12). Yet, up to 44% of individuals with type 2 diabetes exhibited symptoms of autonomic neuropathy in a Danish study (13), suggesting that many cases may go undetected without routine screening. Nonetheless, clinical implementation of the recommendations remains limited, potentially contributing to underdiagnosis and delayed management.

The current gold standard is a test series assessing the integrity of the brainstem-heart connection, the cardiovascular autonomic reflex tests (CARTs). The primary parasympathetic test is a one-minute cycle of deep breathing, which demonstrates RR-interval variation from respiratory sinus arrhythmia, whereas the RR-interval response to postural change and the Valsalva maneuver is a mixed response of both sympathetic and parasympathetic innervation (14). Measures of heart rate variability, baroreflex sensitivity and 24-hour blood pressure monitoring are also common in the assessment of CAN. However, to advance the detection of early-stage CAN, it may be valuable to explore biomarkers that more specifically reflect parasympathetic innervation e.g., vagal rebound during early heart recovery after exercise or cardiac vagal tone (CVT) (15). CVT is a measure of parasympathetic activity derived from a continuous stream of electrocardiograms (ECGs) as pulse-synchronized phase shifts in consecutive cardiac cycles. The normal resting levels of CVT range from 5–15 on the linear vagal scale (LVS) and CVT has previously been shown to be diminished in type 1 diabetes and related to neuropathy status (16–18).

Unless very complicated, T2D is managed in the primary sector. New screening and diagnostic tools should therefore be easy to use to be successful in a primary care facility with limited specialist and technical expertise available. To this purpose, ProBiometrics has developed the ProCVT smart-sheet prototype for screening of dysautonomias such as CAN. The sheet is an ECG-recorder that enables wireless, surface electrode-free, long-term recordings of CVT during sleep, where the parasympathetic tone is at its highest. Thus, this study aimed to investigate whether the ProCVT smart-sheet is a reliable screening tool to assess CAN in T2D.

Table 1 lists the clinical characteristics of the study population. In general, the individuals with T2D had higher BMI, and both antihypertensive and cholesterol-lowering medications were more frequently used. Regarding the biochemical profile, individuals with T2D had higher HbA1C and fasting glucose levels than the controls. However, the individuals with T2D had lower, cholesterol, high-density lipoprotein cholesterol, and low-density lipoprotein cholesterol levels, indicating routine treatment of risk factors according to national guidelines. However, they had higher triglyceride levels, likely related to metabolic syndrome. No major differences were observed in kidney function, but the albumin/creatinine ratio was slightly elevated in the individuals with T2D. In terms of the autonomic profile, manifest CAN was more common in individuals with T2D, and they presented with higher COMPASS-31 scores, lower CVT, lower cardiac sensitivity to the baroreflex (CSB), and had smaller pupil diameter when dilated, all signs of a dysregulated autonomic nervous system.

This feasibility study demonstrated the successful use of a novel non-contact capacitive ECG method, the ProCVT smart-sheet, for home-based recordings without contact surface electrodes. The sheet performed reasonably well with robust measures across multiple nights, and as such, the innovation meets technical requirements. The CVT level was lower in individuals with T2D, and associated with varying stages of CAN. Furthermore, CVT showed potential as a biomarker for CAN, effectively distinguishing individuals with any or manifest CAN from those without. However, long-term CVT recordings did not outperform short-term measures in predicting CAN stages.

Traditionally, long-term ECG recordings are performed using Holter monitors – a small device worn around the neck and connected to chest electrodes. However, wearing them for several days can be cumbersome, with tangled cables (“cable spaghetti”) prone to movement-induced artifacts. Single-channel ECG patches like the ePatch offer a cable-free alternative for up to five days of monitoring, but issues such as skin irritation and motion artifacts remain common. These limitations can reduce patient compliance with long-term monitoring. Many newer wireless systems are patches, handheld devices, or use Bluetooth in combination with traditional surface electrodes (24). In contrast, the ProCVT smart-sheet was developed to mimic a multiple-lead system with its six electrodes without the issue of the ‘cable spaghetti’ or skin irritation. However, as a wireless system without traditional skin electrodes, the device’s data quality required validation. We showed a signal quality of 13 or higher on average, constituting 93% of a recording, which is deemed very acceptable. While 21% of the data was missing before processing, this likely reflects individual sleep patterns, such as rolling off the sheet. Still, 5 hours of ECG data is generally sufficient to get an in-depth analysis of the autonomic tone. Data processing excluded 36 recordings due to a strict, conservative approach chosen to minimize artifacts common in ECG signals.

ECG is vital for assessing cardiac function and is increasingly available through consumer wearables, many of which track heart rate variability (HRV). HRV reflects the variation between successive R-R intervals, making its accuracy dependent on the precision of R-peak detection. Low resolution (8-10-bit) can mask low amplitude R-peaks or distort timing, whereas higher bit depth reduces rounding errors during analogue to digital conversion. While some studies suggest that a sampling rate of 25-250 Hz is sufficient for HRV analysis (25-28), an analytical model shows that at least 1000 Hz is needed for acceptable accuracy (29). This highlights the importance of high sampling rates and resolution for diagnostic purposes, and wearables cannot yet compete with medical-grade devices. The ProCVT smart-sheet combines a 1000 Hz sampling rate with 24-bit resolution, providing an interpolated precision of ~100-200 µs. This makes it well-suited for assessing CVT and HRV metrics such as the Root Mean Square of Successive Differences (RMSSD) and the high-frequency band i.e., RR-interval fluctuations in the 0.15-0.4 Hz range. These reflect parasympathetic activity, where it is necessary to detect changes of 50-20 ms in R-R intervals, and even a 5-ms error in R-peak timing can distort the values and possibly underestimate vagal activity. The combination of a home-monitoring system with high sampling rate, high resolution, multiple electrodes, and automatic upload makes the system well-suited for both clinical and research applications. Particularly the large size of the electrodes makes it very suitable for sleep measures, as it allows the patients to move freely while maintaining contact with the electrodes. This reduces the amount missing data and enhances patient compliance.

HRV includes non-linear, time-, and frequency-domain measures, amounting to around 25 parameters, which can complicate its clinical use. In contrast, CVT offers a simpler and more intuitive alternative. This study found a reduced CVT in T2D as well as in those with any degree of CAN and manifest CAN. These results are consistent with findings from Brock et al. (2017) and Wegeberg et al. (2020), conducted in individuals with type 1 diabetes (16, 17). However, this study found no significant difference in CVT between Early CAN and No CAN, limiting its usefulness for early detection. Both groups showed considerable variation in CVT, which may suggest less parasympathetic withdrawal than expected, potentially challenging current pathophysiological theories. However, it also shows that there is a considerable individual variation in CVT. When comparing CVT across multiple nights, 91% of the variance was attributed to differences between individuals, while only 9% was due to night-to-night variation. This suggests that tracking deviations from an individual’s baseline – rather than relying on a fixed cut-off – may be more informative, especially in relation to factors like stress, sleep quality, or neuropathy. The low night-to-night variation demonstrates that CVT is a reliable marker of parasympathetic tone, likely due to the recordings being taken at home during sleep, a setting shown to provide more robust measures of the heart’s adaptability (30). One advantage of CVT over HRV is its insensitivity to the use of beta blockers. Beta blockers reduce sympathetic activity and subsequently prolong the RR intervals, which can artificially inflate HRV measures (31). In contrast, CVT remains robust under these conditions, as it reflects dynamic vagal modulation, rather than absolute RR-interval duration. Since beta blockers tend to prolong RR intervals uniformly, they do not affect the variability captured by CVT.

Heart rate variability (HRV) is increasingly applied to classify sleep phases, as it is far more accessible than polysomnographic recordings. HRV is fairly effective in distinguishing rapid eye movement (REM) from non-REM sleep, since REM sleep is characterized by higher sympathetic tone (i.e., higher heart rate) and consequently lower parasympathetic tone (i.e., lower CVT) (32–34). This demonstrates the dynamic vagal modulation that occurs during sleep, which we aimed to capture with the concept of ‘CVT capacity’. During our data processing, some sleep segments with either high or low CVT may be excluded due to poor signal quality or artifacts, which carries the risk of under- or overestimating mean CVT. The conservative approach used in this study, requiring that at least 70% of the original data be preserved, with the majority of recordings exceeding 90%, minimizes this risk, while ensuring reliable representation of dynamic vagal modulation across sleep stages. CAN has previously been linked to impaired sleep outcomes (35, 36), probably due to disruption of the sleep architecture, with reduced non-REM sleep, due to parasympathetic withdrawal and sympathetic dominance. Parasympathetic withdrawal would be reflected as a lower mean CVT. However, polysomnographic confirmation of sleep phases is necessary to establish the relationship between CVT and REM/non-REM sleep. Nevertheless, imbalances in the distribution of non-REM and REM sleep are likely reflected in mean CVT, as has been consistently shown in HRV studies (32–34). In healthy individuals, the suppression of deep non-REM sleep reduces glucose tolerance (37). Thus, CAN may not only reduce sleep quality, but also exacerbate nocturnal hyperglycaemia, potentially aggravating pre-existing autonomic neuropathy.

This study investigated the ability of different CVT parameters to distinguish various degrees of CAN. Short-term and long-term mean CVT showed the strongest performance for detecting both the degree of CAN and manifest CAN, with AUCs ranging from 0.64-0.79, balanced sensitivity and specificity, and high negative predictive values – an important feature for screening. For manifest CAN, optimal cut-off points were 2.7 LVS (short-term) and 5.0 LVS (long-term), aligning with Wegeberg et al., 2020, who proposed a short-term threshold of 3.2 LVS in type 1 diabetes based on 5-minute recordings. For early CAN, this study found cut-offs of 4.2 (short-term) and 7.5 (long-term), close to Wegeberg et al.’s 5.2 (17). While we could not confirm CVT’s utility in detecting early CAN, the similar thresholds suggest comparable underlying pathophysiology in both cohorts. Furthermore, the higher long-term cut-point likely reflects increased parasympathetic activity during sleep in a home environment, consistent with known day-night differences in HRV (38). Moreover, a disadvantage of the 5-minute recording is that the CVT may be influenced by the surrounding area, and the patient may be less comfortable in clinical setting which could result in a lower CVT. Whereas the ProCVT smart-sheet can provide a more reliable estimate of the patient’s parasympathetic capacity due to the home element. Additionally, the technology used in the 5-minute recording does not offer real-time recording, requiring uploading and processing, and the processing takes the same amount of time whether the recording is five minutes or eight hours long. The ProCVT smart-sheet, however, automatically uploads data and has the potential for implementing automatic processing, reducing clinical workload and improving feasibility. The ProCVT smart-sheet also supports real-time recording, making it suitable not only for long-term monitoring but also for short-term recordings.

Furthermore, mean CVT outperformed CVT capacity and showed minimal variation across multiple nights, eliminating the need to select a single recording subjectively. This simplifies clinical use, requiring less specialized knowledge. Surprisingly, short- and long-term mean CVT demonstrated similar performance despite being recorded on different devices, suggesting the robustness of the underlying algorithm. Both approaches may be suitable for evaluating CAN in individuals with diabetes, though further studies are needed to explore their ability to detect early-stage CAN, where intervention is still possible. When considering clinical implementation, health economic factors are important. Both methods require high-resolution ECG devices, which are costlier than standard equipment but differ in operational demands. The short-term method involves healthcare staff and yields immediate results, while the smart-sheet allows for patient-led long-term recordings at home, though data must be processed afterward. Ultimately, the choice of method should align with clinical workflows, staff capacity, and patient preferences.

While cross-sectional studies offer valuable insights, they also have inherent limitations. Although this study shows that CVT remains stable over multiple nights, a longitudinal design would better assess its ability to track CAN progression through declining CVT. We excluded recordings from the ProCVT smart-sheet if less than 70% of the original data remained. While this may seem stringent, we opted for a conservative approach to minimize entire sleep phases being removed and thus ensure reliable cut-off point determination We acknowledge that the selection of the ‘best’ recording introduces bias; however, further analysis showed limited night-to-night variation in mean CVT suggesting that this only had a small influence on the cut-off values. Additionally, the study included 40 participants with T2D, most without signs of CAN. Future research should involve a larger, more balanced cohort across CAN stages, particularly early and manifest CAN, to improve understanding of CVT dynamics. Given the predominance of the No CAN group, a 2:1 control-to-case ratio is recommended. Based on a target AUC of 0.75, 80% power, and a 5% significance level, at least 19 early CAN cases and 38 controls are needed in future studies to ensure adequate statistical power. Moreover, it is a limitation that, without polysomnographic confirmation of sleep phases, phase-specific central–autonomic nervous system interactions could not be explored. Unfortunately, polysomnography was not feasible in the home setting of the present study.

This preliminary study demonstrated that the novel ProCVT smart-sheet is a feasible non-invasive method for obtaining ECG recordings effortlessly without surface electrodes, easing the integration of digital medical devices in T2D care. CVT emerged as a reliable biomarker for CAN, effectively distinguishing individuals with any degree or manifest CAN from those without. Long-term recordings were not superior to short-term recordings, suggesting both are viable options. With further development, especially in early-stage CAN detection, the ProCVT smart-sheet could become a valuable clinical tool. Its robust technology may also hold potential for broader use in other neurodegenerative conditions involving autonomic dysfunction, pending further validation.