To our knowledge, this is the first randomised, double-blind, sham-controlled trial assessing the clinical effect of tVNS on gastrointestinal symptoms in individuals with diabetes and autonomic neuropathy. The active and sham stimulation similarly reduced gastrointestinal symptoms during a short-term high-intensity stimulation period and a long-term medium-intensity stimulation period. A minor increase was seen for gastric emptying in the active group, but otherwise, segmental gastrointestinal transit times, CAN score and CVT were similar between the treatment arms.

Potential mode of action of vagal neuromodulation

The gammaCore device is optimised to induce signals in the afferent myelinated Aβ vagal nerve fibres [16]. Assessing somatosensory evoked brain potentials and performing functional magnetic resonance scans of the brain and brainstem during cervical tVNS has demonstrated the activation of vagal fibres in healthy individuals [14, 36, 37]. Studies in animals and healthy individuals have shown that tVNS modulates fundamental patterns of gastrointestinal motility and increases heart rate variability measurements [11,12,13, 38, 39]. Thus, tVNS is believed to stimulate afferent vagal nerve fibres, transmitting the signals to the brainstem and the brain, where an efferent vagal nerve signal is conveyed to the gastrointestinal system [40].

Effect of tVNS on gastrointestinal symptoms

There is a lack of standardisation for tVNS regarding the intensity, frequency, pulse duration, length, anatomical site and type of device. As a result, it is challenging to compare results across studies [41]. Two smaller observational, non-randomised, uncontrolled studies investigated the effect of cervical tVNS in individuals with medically refractory and severely symptomatic gastroparesis of various aetiologies, including diabetes [17, 18]. Both studies reported a decreased symptomatic burden in 35–40% of the participants. The present trial shows a comparable proportion of responders in both treatment arms, stressing the importance of having a placebo group. Our participants are different from those in the above-mentioned studies. First, we only included participants with diabetes. Second, a verified gastroparesis diagnosis was not a criterion for inclusion as this measure correlates poorly with symptoms [6] and shows large intra-study variability (see below). Third, as diabetic gastroenteropathy is a pan-enteric condition, we included patients with a broader spectrum of symptoms than just the cardinal symptoms of gastroparesis. In the present study, most participants also complained of upper gastrointestinal symptoms as evaluated by the overall GCSI and by the symptom category sub-scores for nausea/vomiting, bloating and fullness. The bloating and fullness sub-scores were the primary drivers for the symptom decrease in the active group but no difference in these sub-scores was seen when comparing the treatment groups.

The cut-off value defining a symptomatic response to a given treatment lacks consensus. We defined treatment success as a 30% decrease in one or both symptom scores. This threshold was chosen to reflect the fact that a higher baseline symptom score may lead to a more prominent symptom reduction than a lower baseline score. The proportion of responders increased twofold when applying the 0.75-point GCSI cut-off used in the above-mentioned observational studies. However, the proportion of responders in the active and sham groups remained comparable. A randomised, double-blind pilot study has investigated the effect of cervical tVNS on lower gastrointestinal symptoms in 19 individuals with Parkinson’s disease and gastrointestinal complaints. It showed a GSRS symptom reduction in the active arm, but no differences were observed when comparing the post-treatment symptom scores between the active and sham groups [42]. In line with this study, we also observed a degree of GSRS symptom reduction across treatment groups.

Anatomical variations between men and women, such as differences in subcutaneous adipose tissue, muscular density and subjective compliance, may influence these outcomes. Therefore, sex was included as a covariate in the multiple linear regression analyses, but it did not alter the overall conclusions, and the results are most likely applicable to both sexes in the population.

Effect of tVNS on gastrointestinal transit times

Most studies have not demonstrated changes in gastric emptying following the implantation of gastric neurostimulators [7]. In contrast, we found a minor increase in gastric emptying time in the active group, but no changes were seen for small-bowel or colonic transit times. A longer gastric emptying time has the potential to increase symptoms such as nausea and vomiting. Thus, these findings contradicted the anticipated outcome of the tVNS treatment.

We observed quite large intra-individual and inter-individual transit time variations, especially for gastric emptying time, which are unsurprising. Previous studies have demonstrated that individuals with diabetes and upper gastrointestinal symptoms have more significant transit time fluctuations than healthy individuals [21]. One study investigated diabetic and non-diabetic individuals with gastrointestinal symptoms. It showed that, across two subsequent measurements 14–16 days apart, 30% of the participants had a different gastric emptying time (normal, more rapid or delayed) [43]. This is in line with the present study, in which 24% of participants in the active group and 31% in the sham group had different gastric emptying times (pathological or normal) when comparing assessments before and after study period 2.

Effect of tVNS on cardiac autonomic function

In the present study, tVNS changed neither the CAN score nor the CVT. In healthy individuals, tVNS was shown to increase CVT [13]. In individuals with chronic pancreatitis, a study showed increased CVT when combining tVNS with deep, slow breathing, but another study showed no effect [44, 45]. In individuals with gastroparesis of various aetiologies, cervical tVNS did not improve heart rate variability [18]. In the present study, the median CVT was in the lowest part of the normal range and below the cut-off for recognising established CAN [34, 35]. Furthermore, more than 50% of the participants had an abnormal CAN score at baseline, representing early or manifest CAN. Thus, tVNS may not effectively induce a sufficient efferent nervous signal to change the heart rate variability in individuals with autonomic neuropathy.

Strengths and limitations of the study

The present study has notable strengths, including examining short-term high-intensity and long-term medium-intensity tVNS using a randomised, sham-controlled and double-blind design. The involvement of multiple centres allowed inclusion of more participants, further strengthening the study. Additionally, compliance levels were high and consistent across the active and sham groups.

There are also several limitations. First, the tVNS was self-administered without real-time assessment of vagal nerve fibre activation. Therefore, it is impossible to directly determine the ‘exposure’ to the active stimulation. Second, the intended vagal modulation may be hampered by autonomic neuropathy in any of the vagal neurocircuits within the brainstem, midbrain and higher cortical centres controlling gastrointestinal motility [46]. Third, no established standards exist for stimulation duration and daily treatment frequency for treating gastrointestinal symptoms. However, the dosage of two consecutive 120 s stimulations twice daily has proven effective as a prophylactic approach for treating primary headaches [16, 47]. Previous studies targeting gastrointestinal symptoms have often used two to four daily stimulations [17, 18, 42]. Long-term treatment compliance may be limited by the time-consuming nature of the handheld application, especially with treatment frequencies exceeding two daily stimulations. Fourth, the participants had either type 1 or type 2 diabetes, reducing the homogeneity of the cohort. However, studies have shown a comparable prevalence of diabetic autonomic neuropathy between the diabetes categories [6]. Adjusting for diabetes type in the multiple linear regression analyses did not affect the symptom score difference between treatment arms. Fifth, autonomic neuropathy was determined using either the COMPASS31 questionnaire to assess autonomic symptoms, the VAGUS device to evaluate cardiovascular autonomic reflex test, or the SUDOSCAN device to estimate sudomotor function by measuring electrochemical skin conductance. These methods are all broadly accessible and easy to use. While more accurate methods exist, they are more complex and time consuming, making them unsuitable for this study [33, 48]. In addition, the SUDOSCAN measurement is only a surrogate measure of autonomic function, and the sweat response may not be exclusively induced by sympathetic autonomic fibres [49]. Sixth, measures of transit times were based on wireless motility capsule data. Thus, the accuracy of the gastric emptying time may be lower than that of scintigraphic measures, as the solid capsule can be expelled from the stomach in the fasting state after emptying the solid meal [50]. However, gastric emptying times obtained using wireless motility capsules have been shown to correlate with 4 h scintigraphy measurements (sensitivity 0.87 and specificity 0.92) [50, 51]. The rationale for choosing the wireless motility capsules in the present study was to obtain a comprehensive evaluation of pan-enteric transit times in these individuals with multiregional dysmotility [1]. Seventh, the incorporation of a 2-week wash-out period between the study periods aimed to assess the efficacy of each treatment regimen independently, preventing potential carry-over effects from study period 1 that may interfere with study period 2. However, introducing a wash-out period also increases the risk of relative unblinding, particularly if the tVNS treatment demonstrated clinical effectiveness. Eighth, the observed symptom scores were generally lower than in previous studies investigating individuals with gastroparesis. Nonetheless, when only analysing participants with more than ‘mild symptoms’, the symptom changes were still comparable between groups. Lastly, the sham device only produces a humming sound, and the active device often induces twitching in the superficial facial muscles unilaterally. Thus, unintentional unblinding of participants during the trial is a potential risk.

In conclusion, the present randomised, sham-controlled, double-blind study provided no evidence of gastrointestinal symptom relief when applying short-term high-intensity or long-term medium-intensity cervical tVNS in individuals with diabetic gastroenteropathy compared with sham stimulation. Hence, tVNS in this format is probably not a recommendable adjuvant treatment to ease the burden of gastrointestinal symptoms in these individuals.