Diabetic diarrhoea: A study on gastrointestinal motility, pH levels and autonomic function

Chronic diarrhoea is a common, but poorly investigated diabetes complication. Autonomic neuropathy is a leading pathophysiological theory founded on old, small studies. Studies of gastrointestinal motility and pH levels are lacking.


Introduction
Chronic diarrhoea affects more than 10% of diabetic patients and often leads to impaired quality of life [1,2]. Evaluation can be challenging, as diabetic patients are at increased risk for developing other conditions leading to diarrhoea, like coeliac disease, pancreatic exocrine insufficiency and inflammatory bowel disease [3][4][5][6][7]. Diarrhoea may also come from dietary factors or common antidiabetic drugs like metformin and glucagon-like peptide-1 agonists [8][9][10][11]. However, in half of all patients, diarrhoea will be attributed to alterations in intestinal motility and secretion secondary to diabetic gastroenteropathy [12]. Diabetic gastroenteropathy can affect any portion of the gastrointestinal tract, leading to manifestations like oesophageal dysmotility, gastroparesis and intestinal hyper-or hypomotility [11,13,14]. When these patients present with chronic diarrhoea, it has been termed diabetic diarrhoea [12]. Here, the diarrhoea is typically nonbloody and painless, with a high-volume, watery consistency. It may be nocturnal and can lead to faecal incontinence [1,12].
Although being recognized for almost a century, little is known about the pathophysiological mechanisms behind diabetic diarrhoea [13]. The leading theory of autonomic neuropathy was established after finding high covariance between neuropathy and diarrhoea in case studies and a clinical resemblance with patients who had undergone vagotomy or sympathectomy [15,16]. Later studies are limited in numbers, inconclusive, and there are no studies using new technology for assessing autonomic function [13,15,[17][18][19]. Studies investigating intestinal transit and contractility are also lacking. One explanation may be that these measurements previously have been laborious and patient unfriendly, limiting their availability to specialized centres. Recently, the wireless motility capsule has emerged as a promising method, simultaneously measuring transit times and contractility throughout the gastrointestinal tract whilst patients are ambulant [20]. The capsule also measures pH levels [21]. This is relevant since diarrhoeal disorders are associated with both intraluminal and systemic pH level alterations [22,23]. So far, these measurements are unexplored in diabetic diarrhoea, but there are noteworthy findings from studies on patients with irritable bowel syndrome and small intestinal bacterial overgrowth [24,25].
In this study, our main hypothesis was that diabetic patients with diarrhoea had altered gastrointestinal transit times. We also hypothesized that they had altered intraluminal pH levels, reduced contractility and autonomic dysfunction. To investigate this, we examined a cohort of diabetic patients with gastrointestinal symptoms and suspected gastroenteropathy using wireless motility capsule and autonomic function tests.

Study population
Between 2014 and 2018, we prospectively included diabetic patients with symptoms suggestive of gastroenteropathy into a cross-sectional, observational study. All patients had been referred for diagnostic evaluation at a tertiary centre at Haukeland University Hospital, Bergen, Norway. Inclusion criteria were type 1 or type 2 diabetes, age over 18 years and normal upper endoscopy. Exclusion criteria were pregnancy, breastfeeding, active malignancy (defined as any cancer not in complete remission for the last six months) and lack of ability to comply with the study protocol.
Patients were admitted to the hospital during the study, where they were interviewed and examined by a physician, and delivered blood, urinary and faecal samples (Table 1). They were kept on intravenous glucose-insulin infusion during fasting and examinations, with a target glucose level between 4 and 10 mmol/L.

Wireless motility capsule
The wireless motility capsule (SmartPill; Medtronic, Minneapolis, USA) is a 26 × 13 mm indigestible, single-use capsule. It registers temperature (range 25-49°C), pH (0.5-9.0) and pressure (0-350 mmHg) throughout the gastrointestinal tract. During the test, data are transferred to a portable data receiver and afterwards downloaded to a computer. For analysis, we used MotiliGI ® software version 3.0 (Medtronic).
We used stereotypical pH profiles to define transit times [21]: Gastric emptying time (capsule ingestion -pylorus), small bowel transit time (pylorus -ileocaecal junction), colonic transit time (ileocaecal junction -capsule expulsion); and whole gut transit time (capsule ingestion -capsule expulsion). Antral pH was defined as median pH for the last 15 min before the pylorus; duodenal pH the first 15 min after the pylorus; ileum the last 15 min before the ileocaecal junction; caecum the first 15 min after the ileocaecal junction; and rectum the last 15 min before capsule expulsion. Delta pylorus was defined as the difference between duodenal and antral pH; delta ileocaecal junction the difference between ileal and caecal pH.
We also measured the motility index and contractions per minute in the whole stomach, small bowel and colon [20]. To determine the ileocaecal junction pressure, we used the method proposed by Chander Roland et al., identifying the maximum pressure for the last 4 min prior to the ileocaecal junction pH drop [25].
All patients had to pause medications potentially altering intestinal function before and during the study. We have specified details in a previous article, together with a description of the test meal and initiation protocol [26]. Patients continued other regular medications, provided doses had been stable for 3 months. Intake of alcohol was prohibited, and patients were asked to refrain from smoking and strenuous physical activity.

Autonomic function tests
We assessed heart rate variability at rest and baroreflex sensitivity using the Heart Rhythm Scanner PE and the Biocom 5000 Bluetooth ECG Recorder (Biocom Technologies, Poulsbo, USA). We have described the heart rate variability protocol in a previous paper [27]. To measure baroreflex sensitivity, patients took deep breaths at a rate of five per minute. Thereafter, actual values were compared with predicted normative age-adjusted values by the software. Finally, we assessed orthostatic hypotension using Welch Allyn ProBP 3400 (Welch Allyn Inc., Skaneateles Falls, USA) following a standardized protocol measuring supine, resting blood pressure and standing blood pressure after 1 and 3 min. Orthostatic hypotension was defined as a drop in systolic blood pressure of ≥20 mm Hg or diastolic blood pressure ≥10 mm Hg from supine to standing position [28].

Symptom assessments
Symptoms were evaluated by physician interview and using the Gastrointestinal Symptom Rating Scale (GSRS), a questionnaire validated for assessing the occurrence and severity of upper and lower gastrointestinal symptoms during the last week [29]. GSRS includes 15 questions, each rated from no discomfort (zero points) to very severe discomfort (six points). Diarrhoea syndrome (hereafter named 'diarrhoea') is derived by taking the mean of the individual symptoms: increased passage of stools, loose stools and urgent need for defecation [29]. We used a cut-off value of ≥4 points, corresponding to the 75th percentile, to define cases with diarrhoea. Those scoring <4 were used as controls. We also looked at correlations between diarrhoea score and each wireless motility capsule and autonomic function test parameter.

Statistical analysis
Continuous variables are stated as median (quartiles) and categorical variables as n (%). We used the Mann-Whitney U test to compare two continuous variables, using r as an effect size estimate (r = z/square root of the total number of cases, N).

Ethical considerations
All participants submitted oral and written consent prior to study-related procedures. The study was approved by The Western Norway Regional Medical Ethics Committee (2015/58) and conducted in accordance with the Declaration of Helsinki.

Results
Seventy-two patients were included in the study, of which 68 were examined with wireless motility capsule. We were unable to identify the ileocaecal junction in three patients, precluding determination of small bowel and colonic transit. Of the remaining 65 patients, eight had missing GSRS data, leaving 57 available for comparisons. An inclusion flowchart is shown in Fig. 1.

Transit times, pH levels and contractility parameters
A comparison of transit times, pH levels and contractility parameters between cases and controls are presented in Table 2. Transit times are also displayed in Fig. 2, pH levels in Fig. 3. Correlations between all wireless motility capsule measurements and the continuous GSRS diarrhoea score are shown in Table S2.
We found that cases had slower gastric emptying (p = 0.03) and faster colonic transit (p < 0.001) than controls. We found no difference in small bowel transit (p = 0.11) nor whole gut transit (p = 0.16). Colonic transit correlated with the diarrhoea score, p = 0.006.

Autonomic function tests
In Table 3, we present a comparison of autonomic function tests between cases and controls, as defined by the GSRS cut-off value. Correlations between autonomic function test parameters and the GSRS diarrhoea score are shown in Table S3.
In cases, we found a trend towards increased diastolic blood pressure drop at 3 min, p = 0.054. We found no difference in any of the other parameters (all p > 0.10) and no significant correlations (all p > 0.10). Thirteen controls (33%) and eight cases (57%) had orthostatic hypotension, χ 2 (1) = 2.65, p = 0.10, ɸ c = 0.22. Of all autonomic function test parameters, only diastolic blood pressure drop at 0 min correlated significantly (p = 0.04) with colonic transit time, Table S4.

Discussion
In this study we aimed to investigate the association between diabetic diarrhoea, intestinal motility, pH levels and autonomic dysfunction. By examining diabetic patients with wireless motility capsule, we found that patients with diarrhoea had slower gastric emptying and faster colonic transit than controls. They also had an increased pH level in the stomach's antrum, caecum and entire colon and decreased pH difference across the pylorus and ileocaecal junction. We found a moderate negative correlation between diastolic blood pressure drop and colonic transit time, but no other associations between diabetic diarrhoea and autonomic dysfunction. Previous studies of intestinal dysmotility in diabetic diarrhoea have shown divergent results [31][32][33][34][35][36]. Some have found an association between diabetic diarrhoea, prolonged transit time and small intestinal bacterial overgrowth [31,33,34]. Others have found results similar to ours, with shortened intestinal transit, some also identifying a correlation with autonomic dysfunction [12,32,35,36]. Theoretically, autonomic dysfunction may induce intestinal dysmotility through several pathways. Loss of inhibitory input through damaged sympathetic innervation could explain the rapid transit seen in our diarrhoea patients [17]. Stimulation of alpha-adrenergic receptors on enterocytes is also important for intestinal fluid absorption, and autonomic dysfunction could lead to increased colonic fluid levels and watery diarrhoea [1,13]. Alterations in the sympathetic and parasympathetic nervous systems have been found in several human pathological studies [17,37]. Despite this, Whalen and colleagues demonstrated intact efferent autonomic function in patients with diabetic diarrhoea when investigating intestinal motility in response to intravenous stimulation by adrenergic and cholinergic agents [38]. They did, however, find reduced pain response to intrajejunal balloon distention, indicating afferent dysfunction [38]. Similar findings have been made in the oesophagus, duodenum and rectum, whilst gastric barostat studies have found the opposite in diabetic gastroparesis: visceral hypersensitivity [39][40][41]. Studies utilizing cardiac autonomic function tests in patients with diabetic gastroenteropathy have also provided conflicting results [36,42,43].
In this study, we were unable to find any differences between cases and controls in heart rate variability or baroreflex sensitivity. Neither did we find any correlations between these parameters and diarrhoea score. We did, however, find a trend towards an increased orthostatic blood pressure drop in cases. In addition, we found a moder-ate negative correlation with colonic transit time. These results could indicate a possible impairment of the sympathetic nervous system, although in such case, we would have expected to find differences in the high-frequency spectres of the heart rate variability as well [28,44]. Overall, our results imply that mechanisms other than autonomic dysfunction are more prominent in the pathophysiology of diabetic diarrhoea. One explanation for our findings could be that some patients in the comparator group also had enteric dysmotility. All patients had gastrointestinal symptoms and a clinical suspicion of gastroenteropathy, but controls differed with respect to not reporting diarrhoea. To investigate this further, we found that a small proportion of controls had slow-transit constipation, as defined by GSRS and prolonged colonic transit [21,29]. However, excluding these patients did not alter the statistical significance of our original results.
Another pathophysiological theory potentially explaining our findings is the loss of enteric neurons [17]. Through the production of nitrogen monoxide, these neurons have an important inhibitory effect on gastrointestinal peristalsis, and their depletion may lead to accelerated transit [1]. Apoptosis of enteric glial cells may aggravate neuronal loss [1]. Another possible mechanism may be reduced synthesis of sodium hydrosulphide, which acts as an inhibitor of intestinal smooth muscles [45]. There are conflicting results regarding the role of bile acid malabsorption in diabetic diarrhoea, but the theory has recently gained new impetus [12,46]. Increased levels of colonic bile acids might explain diarrhoea through several mechanisms, including a direct stimulatory effect on motility [47]. Small intestinal carbohydrate malabsorption may accelerate colonic transit through an increased fluid load, but short-chain fatty acids produced by fermentation of carbohydrates slow down transit [48]. Additional mechanisms possibly contributing to dysmotility are neuroendocrine dysregulation, alterations of smooth muscle cells and loss of interstitial cells of Cajal [17]. Interestingly, the effect of hyperglycaemia is somewhat paradoxical: whilst chronic hyperglycaemia is central in the development of enteric neuropathy and hence leads to accelerated transit, acute hyperglycaemia leads to delayed transit throughout the entire gastrointestinal tract [13,14,17].
Our study is the first to report intestinal pH alterations in patients with diabetic diarrhoea. Previously, this has been investigated in asymptomatic type 1 diabetic patients with peripheral neuropathy, findings from studies show a decreased colonic pH levels and an increased pH difference across the ileocaecal junction compared to healthy controls [49,50]. Similar findings have been demonstrated in irritable bowel syndrome [24]. Normally, the pH level decreases more than one unit across the ileocaecal junction as a consequence of the more acidic environment in the caecum compared to the ileum [21]. This is mostly due to bacterial fermentation and production of short-chain fatty acids [49]. The magnitude of the ileocaecal pH drop has therefore been suggested as a proxy for the degree of fermentation in the proximal colon [51]. This may be increased in carbohydrate malabsorption or with heightened intake of fibre or other nonabsorbable sugars, the latter being a common cause of diarrhoea in diabetes [8,11,48]. When our cases had a decreased ileocaecal pH drop and an increased intracolonic pH profile than controls, this may reflect another microbial profile [21,49]. A number of factors may influence the microbial composition, including diet, stool consistency, intestinal transit times and bile acids [52,53]. Theoretically, bile acid malabsorption may lead to colonic pH alterations directly, but this is so far not reported in studies. Different types of nutrients can influence pH levels indirectly, where increased production of ammonium in high-protein diets may lead to an alkaline intracolonic milieu [52]. In contrast to the fermentation of carbohydrates, protein fermentation is most pronounced in the lower parts of the colon, thus being a less likely cause of our caecal pH findings [53]. The interrelationship between intestinal transit and pH levels may be unpredictable: when colonic transit is rapid, pH levels may increase as bacteria have less time to ferment carbohydrates. At the same time, rapid transit may induce a shift towards lactate production, potentially lowering pH levels [52]. In this study, we found no association between pH levels and transit times.
Another possible explanation for our findings of a more alkaline caecal micromilieu may be altered activation of receptors facilitating bicarbonate secretion [54]. Interestingly, a study administering linaclotide to patients with irritable bowel syndrome with constipation, increased caecal pH, reduced colonic transit time and improved symptoms [51]. Linaclotide exerts its effect through increased luminal secretion of chloride and bicarbonate, in the next case leading to increased efflux of water [51]. Ileocaecal valve dysfunction could lead to a decreased pH drop across the ileocaecal junction, as shown in patients with Crohn's disease who had undergone ileocaecal resection. Compared to controls, patients had increased pH in the caecum, whilst ileal pH levels were similar [55]. Ileocaecal valve dysfunction has also been associated with bacterial overgrowth [25]. However, we did not find any differences in ileocaecal junction pressure between cases and controls. Neither did we find any other differences in contractility parameters, but this should be explored in more detail in future studies. New studies are also needed to investigate the many potential causes of pH level alterations in diabetic diarrhoea, including characterization of the microbiome and tests for bile acid malabsorption and bacterial overgrowth.
Previous studies have shown that wireless motility capsule examinations have large therapeutic consequences, providing new diagnoses in 50% of patients and changing treatment in 75% [56]. Our results also suggest that tests of gastrointestinal motility and pH levels have a role in the evaluation of diabetic diarrhoea, potentially guiding medical treatment. As an example, the patient with slow small bowel transit secondary to bacterial overgrowth needs a different therapeutic approach than the patient with rapid colonic transit caused by enteric neuropathy. Many diabetic patients also have concurrent dysmotility in more than one gastrointestinal segment, evidenced by our diarrhoea patients having both delayed gastric emptying and rapid colonic transit [12,35,56]. Here, motility testing may help to tailor pharmacological treatment [13,35,57,58]. Furthermore, alterations in pH levels or changes in luminal water content may affect intestinal drug delivery and absorption, being especially relevant for the release of active substances from drugs with controlled release formulations [14].
Although not yet investigated in diabetes, intestinal pH level alterations may also be linked to visceral sensitivity [59]. Finally, and crucially, the attention to this underreported and undertreated diabetes complication should be increased in health care providers. It is worrying that 30% of our study patients had diarrhoea, but only 7% used antidiarrhoeal medications.
There are some methodological considerations regarding our study. We used a validated questionnaire to assess bowel function [29]. As there are no predefined dichotomous cut-off values for the GSRS diarrhoea syndrome, we chose to define ≥4 points as cases with diarrhoea. This cut-off value was intentionally conservative, to maximize sensitivity for detecting true diarrhoea cases. A post hoc Kappa analysis demonstrated a substantial agreement between our chosen cut-off for diarrhoea and clinical information gathered from physician interviews. Additionally, we performed correlation analyses showing similar results, thus strengthening our findings. Furthermore, the exact localization of the wireless motility capsule is only possible when it passes the pylorus, ileocaecal junction or is expelled from the body [60]. The definition of gastrointestinal subsegments is therefore based on temporal measurements in relation to these physiological landmarks. We utilized pH measurements 15 min before and after the pylorus and ileocaecal junction to determine pH in the adjacent subsegments, similar to the reference study by Wang and colleagues [21]. Other studies have used 30-min measurements or split the intestines into quartiles [49,50,61]. Compared to these approaches, 15-min measurements are preferential in patients with rapid transit. Due to the large variance in transit times, it also has an advantage over the quartile approach when it comes to interindividual comparisons. As stabilized pH values for >10 min is a criterion for manually determining the physiological landmarks, and the capsule has a negligible lag phase for detecting pH changes, 15-min measurements are likely sufficient [60]. Nevertheless, we support further validation studies to establish a consensus. Lastly, to investigate the association between diabetic diarrhoea and autonomic dysfunction, we measured heart rate variability, baroreflex sensitivity and orthostatic hypotension [28]. These are validated methods for assessing cardiac autonomic function and are often used as a proxy for visceral autonomic neuropathy due to the lack of ideal tests for evaluating gastrointestinal autonomic function [27,62]. We have previously demonstrated an association between impaired rectal sensitivity, indicating autonomic neuropathy, and reduced cardiac autonomic function [40]. Others have also found an association between cardiac autonomic neuropathy and gastric vagal neuropathy [63].
Our study had some limitations. Being an exploratory study, we did not perform an a priori power analysis, but our main findings still had moderate effect sizes. However, our study may have been insufficient to identify a minor difference in small bowel transit. We also included patients having comorbidities or using drugs associated with diarrhoea (Table S1). Due to their frequency, excluding these patients would potentially introduce a selection bias. To assess eventual influence from comorbidities, we compared GSRS scores, only finding a marginally lower free thyroxine in diarrhoea patients (Table S1). Since the difference between groups was within the biological variation of free thyroxine, and both groups were in a euthyroid state, we find this unlikely to have had an influence on symptoms [64]. As for medications, we found higher diarrhoea scores in patients using opioids and antiepileptic drugs, both drug classes common in the treatment of painful neuropathy. Considering this, we find it unlikely that our main findings could be explained by medications.
The main strength of our study was the use of state-of-the-art technology to assess gastrointestinal motility, pH levels and autonomic function. To our knowledge, this is the largest experimental study to date investigating diabetic diarrhoea. Whilst similar studies often have a retrospective design, we used prospective inclusion. Thereby we limited potential biases and were able to standardize patient characterization using structured interviews, review of medicine lists and measurement of biochemical parameters. Another strength was the measurement of faecal calprotectin and faecal elastase-1 to exclude previously undiagnosed inflammatory bowel disease and pancreatic exocrine insufficiency, respectively.
To conclude, we found that patients with diabetic diarrhoea had slower gastric emptying, faster colonic transit and altered gastrointestinal pH levels. Overall, our findings do not support the association between diabetic diarrhoea and autonomic dysfunction. Our results add increased knowledge to a field largely devoid of research for the last two decades. Hopefully, they provide the groundwork for further studies into the pathophysiology of diabetic diarrhoea. Our study also proves that measurement of transit times and intestinal pH levels can be a valuable guide for individualized treatment and may warrant a more central role in the evaluation of diabetic diarrhoea.