Gustatory sweating in people with type 1 and type 2 diabetes mellitus: Prevalence and risk factors

Abstract Objective Gustatory sweating (GS) is characterized by profuse sweating during or immediately after ingestion of food and is known as a complication of diabetes mellitus (DM). This study aimed to determine the prevalence of GS and to characterize the sweating in a cohort of patients with type 1 and type 2 diabetes mellitus (T1DM and T2DM) as compared with a control group. Methods In a cross‐sectional study, 665 outpatients with T1DM and 505 outpatients with T2DM filled in an 8‐point questionnaire about GS. Answers were paired with medical data from the electronic patient records to explore associations with DM complications. The control group consisted of 1158 persons without DM answering the same questionnaire in an online version. Results In people with T1DM and T2DM, the prevalence of GS was 10% (95% CI 7%–12%) and 13% (95% CI 10%–16%), respectively. In the control group, the prevalence of GS was 5% (95% CI 3%–6%). Most commonly, people sweat on the face and/or head and upper body with a duration of 10–30 min albeit in the control group <10 min. In patients with T1DM, increased HbA1c was associated with GS (OR 1.3 [95% CI 1.05–1.6], p = .016), and in T2DM, younger age (OR 0.95 [95% CI 0.92–0.99), p = .006), presence of severe peripheral neuropathy (OR 2.33 [95% CI 1.04–5.2], p = .039) and absence of proliferative retinopathy were associated with GS (OR 0.22 [95% CI 0.07–0.71], p = .011). Conclusion We found the prevalence of gustatory sweating of 11% in a hospital‐based cohort of patients with T1DM and T2DM. This was twice as high as in non‐diabetic control persons. Associations between GS and known diabetes complications could only be demonstrated in T2DM. Compared with a control group without DM, odds for GS are higher in people with DM and age >45.


| INTRODUC TI ON
Gustatory sweating (GS) is a secondary form of focal hyperhidrosis triggered by food intake. GS can either be physiological such as when eating spicy foods or be non-physiological where the response is independent of food type. 1 There is no internationally agreed definition of GS. In a position statement from the American Diabetes Association (ADA) on diabetic neuropathy, GS is described as a sudomotor dysfunction with profuse sweating on the face and neck in relation to food intake (or in some cases the smell of food). 2 In the current study, we have expanded the definition to include sweating also from other parts of the body in relation to food intake. Although considered a harmless condition, GS can cause distress and strong feelings of shame. Many patients report withdrawal from eating in social settings, which has a strong negative impact on their quality of life. [3][4][5] In some cases, GS disrupts normal eating patterns, which can lead to poorer glycaemic control and potentially life-threatening hypoglycaemia in patients with insulin-treated or sulphonylurea-treated diabetes (DM). 6,7 The pathophysiology behind GS in diabetes is unknown. Some studies hypothesize that it is a manifestation of autonomic dysfunction due to aberrant nerve fibre regeneration. 4,5,8 Others discuss evidence of separate aetiologies like compensatory thermoregulation, anti-sympathetic ganglia antibodies, neuropathic loss of suppression of nerval tonus that controls sweating and the role of reversible molecular changes due to nephropathy. [9][10][11] The distribution of sweating is equivalent to the territory of the superior cervical ganglion 5 and can be objectified, by applying quinizarin powder to the face, head and upper torso/extremities (turning blue when getting wet) 5 or by weighing absorbent dressings worn during meals. 3 Both methods are impractical in larger cohorts. 3 As an alternative, Shaw et al 10 used a questionnaire for self-reported GS. To test the reliability of the questionnaire, they made 25 random samples and found that it was a reliable way of reporting GS. 10 The prevalence of a gustatory sweat response in people with DM compared to the background population has not yet been established. Neither has its potential relation to other DM complications been investigated.
In this study, we assessed the prevalence of GS in cohorts of people with type 1 and type 2 DM (T1DM and T2DM) and in a control group without DM. Furthermore, we describe duration and body location of GS during meals, and to generate hypotheses on the pathophysiology of GS we looked for associations between GS and late diabetic complications as well as different indices of glycaemic control.

| Design
In a cross-sectional study, 745 outpatients with T1DM at Steno Diabetes Center (now Steno Diabetes Center Copenhagen), Denmark, and 991 outpatients with T2DM in the diabetes clinic at Nordsjaellands Hospital, Denmark, received a questionnaire by mail with eight questions regarding GS ( Figure 1). Answers were paired with medical data from the electronic patient records. The study was approved by the Danish Data Protection Agency (#2012-58-0004). Patients were informed about the scope, purpose and design of the study in an accompanying letter and gave consent to use previously collected clinical data by sending back a signed filled-in questionnaire.
Inclusion criteria were a diagnosis of DM based on the diagnoses (ICD-10 diagnosis code: DE10.X = T1DM and DE11.X = T2DM) recorded in the electronic patient record, and age >16 years.
Exclusion criteria were questionnaires received too late, a lack of ID-number or signature.
A control group of 1158 people with no prior or current diagnosis of diabetes answered the same 8-point questionnaire in an online version distributed through social media. The questionnaire was made available to an unselected cohort on several social platforms.
Exclusion criteria (age below 16 years or yes to current or prior DM) were asked as the first questions, and the GS questionnaire only opened if no exclusion criteria were met.

| Clinical data collection
To test for association between the presence of GS and diabetic complications, clinical data from patient records were collected as close to the date of the return of the questionnaire as possible and at least within a year. The following data were collected: duration of diabetes (years), HbA1c (mmol/mol), peripheral neuropathy (assessed by biothesiometry and defined as present if vibration perception threshold (VPT) was ≥50 V on one foot), nephropathy (urinary albumin/creatinine-ratio [UACR] subdivided into normoalbuminuria [<30 mg/g], microalbuminuria [30-300 mg/g] and macroalbuminuria [>300 mg/g]), diabetic retinopathy (classified in none, non-proliferative diabetic retinopathy [NPDR] and proliferative retinopathy based on digital fundus photography). In the control group, age, sex, height, weight and yes/no to kidney disease were self-reported.

| Primary endpoint
The prevalence of GS was the primary end-point. GS was defined as present if a patient answered yes in box 1 and 2 in the questionnaire ( Figure 1). The cohort with GS was further split up into physiological and true GS. A sweat response was considered physiological if any known spicy foods were reported in question 5. True GS was defined as sweating not only triggered by spicy foods. Patients that mentioned inconclusive GS food triggers were excluded from the true GS group.

| Secondary endpoints
The other items on the questionnaire (duration, location, time of sweating regarding meals and start of GS symptoms) were secondary end-points used to further characterize the sweating. The true GS and physiological GS groups were compared. Correlations between known diabetic complications and the risk of GS were additional explorative end-points. Gender, age, duration of DM, HbA1c, nephropathy, retinopathy and neuropathy were included in the analysis.

| Statistical analyses
Data are presented for T1DM, T2DM and the control group separately. All numerical data were assessed for normality by using a one-sample Kolmogorov-Smirnov test, which indicated that all data were normally distributed, except for age at diagnosis of DM.
Consequently, a Mann-Whitney U test was performed for this variable. Comparing other continuous variables was done using independent Welsh t tests. Continuous data are presented as means (±1SD). Categorical data were compared with a chi-square test and are presented as percentage.
To identify variables associated with GS, the following variables were included: sex, age, duration of diabetes, HbA1c, albuminuria, retinopathy and neuropathy. A multiple logistic regression was performed to model the interdependent influence of these variables on F I G U R E 1 Questionnaire sent to 745 patients with type 1 diabetes from Steno Diabetes Center and 991 patients with type 2 diabetes from Nordsjaellands Hospital. A total of 1158 controls completed the exact same questionnaire in an online form Questionnaire regarding sweat production in relation to meals: risk of GS, and odds ratios were calculated. To minimize collinearity between variables in the multiple logistic regression analysis, duration of DM was included and age at DM diagnosis was excluded. A two-tailed p-value of ≤.05 was considered statistically significant. All statistical analyses were performed using SPSS 25.0 (SPSS).
To control for overall age differences in the DM cohort and the control group, five age groups were defined as 16-29, 30-44, 45-59, 60-74 and 75+ years of age. Moreover, to clarify whether perimenopausal sweating in woman contributes to the prevalence of GS, we did a supplementary analysis comparing prevalence of GS in men and women aged 40-60 years in both the DM group and in the control group.

| RE SULTS
Questionnaires were sent out to totally 1736 people with T1DM and T2DM and 1204 valid questionnaires (69%) were returned ( Figure 2).

| Primary endpoint
A total of 1170 people, 665 with T1DM and 505 with T2DM, were included in the analysis of the primary endpoint. In the T1DM cohort, 13% of the patients had physiological GS (they answered yes to question 1 and 2) and 10% (95% CI 7%-12%) had true GS. In the T2DM cohort, 22% of the patients had physiological GS and 13% (95% CI 10%-16%) had true GS. In the control group, 1158 people were included, 9% had physiological GS and the prevalence of true GS was 5% (95% CI 3%-6%).

| Secondary endpoints
The analysis of the questionnaire data regarding the sweat characteristics used data of all 1170 people with DM, but for the logistic regression analysis, only data of 1158 people (653 with T1DM and 505 with T2DM) were analysed because clinical data were missing in 12 patients with T1DM.

| Characteristics of sweating
The debut of GS, bodily distribution of sweating, triggers and duration for both T1DM, T2DM and the control group are presented in Table 2.
Most people with DM had a history of GS symptoms for 1-5 years (44% in T1DM and 48% in T2DM) but in the control group, most people had GS symptoms for >10 years (37%). Some patients, however, report having experienced the symptoms for less than 1 year and for more than 10 years. In most people, GS starts by the end of or after a meal. Most people with T1DM and T2DM respondents reported duration of sweating to be 10-30 min (58% in T1DM and 37% in T2DM). In the control group, the most reported duration was <10 min (44%). There was a tendency for people with T2DM to sweat longer than both people with T1DM and the control group (T2DM 23%, T1DM5% and control group 11% sweat for 30-60 min, respectively).
The non-spicy trigger foods for GS that were mentioned by more than one patient were soup and fruit in T1DM, fatty foods and meat or beef in T2DM, and candy in the control group. Foods that were mentioned by both patients with T1DM, T2DM and in the control group as triggers were soups, fatty foods, sugary foods and cheese.

| Explorative analysis of associations between gustatory sweating and DM complications
In patients with T1DM, logistic regression analysis showed that increasing HbA1c was associated with increasing probability of GS (OR   Table  5 were back of the neck (6) (Table S1).

| Menopause
To clarify whether perimenopausal sweating in woman contributed to

| DISCUSS ION
We found the overall prevalence of GS of around 11% in a large, hospital-based cohort of patients with T1DM and T2DM. GS was associated with higher HbA1c in T1DM and with lower age, severe peripheral neuropathy and absence of proliferative retinopathy in T2DM. In the control group, we found the overall prevalence of GS of 5%. When comparing in age groups, we found that people with DM and age above 45 years had increased risk of true GS. To our knowledge, this is the first study to look at the prevalence of GS in both people with DM and without DM. The results support the general view that GS is a complication to DM. specific trigger foods, particularly cheese, and in one patient by the thought of food alone. 5 The distribution of GS correlating to the territory of the superior cervical ganglion made the authors suggest that GS is due to abnormal regrowth of damaged vagal nerve ends to sympathetic cholinergic sweat fibres at the level of the superior ganglion. When atropine was administered to three patients, symptoms of GS disappeared, supporting their hypothesis. 5 We were able to confirm the distribution of sweating to mainly head, face and upper body through a self-reported questionnaire in both people with diabetes and the control group. The trigger foods mentioned in previous studies, such as cheese or chocolate, could not be confirmed as strong triggers (albeit mentioned by a few patients) in our study but patients mentioned a wide range of trigger foods with fatty foods, meat/beef and soup being mentioned most often. 5,11 The similarity of characteristics of GS in people with physiological and non-physiological sweating supports the idea of similar pathways as the physiological sweat response to spicy food. We found that duration of sweating in connection to a meal is often long, for some patients (with type 2 diabetes) up to 2 h and that GS for some patients is a long-standing complication being present for years.
In a study from 1996 using patient interviews and includ- and GS also report high HbA1c as associated with GS. 15,16 Poor glycaemic control may impact the risk for GS, and the potential mechanism should be addressed in further studies.
Proliferative retinopathy-which was associated with a lower probability of having GS in our T2DM cohort-is a microvascular diabetes complication and is often present in an asymptomatic state long before diagnosis. 22,23 This could indicate that a microvascular aetiology of GS is unlikely, but the observed association needs fur-

| Strengths and limitations
The strengths of our study are primarily the large number of patients included, the access to clinical data, the inclusion of a control group and the relatively high questionnaire response rate, especially in the T1DM cohort. Moreover, the answers to the questions about duration and location of GS add to our knowledge about this diabetes complication. More clinical information is required to make further thorough analyses of autonomic neuropathy, possible sweat as sideeffects of medication, lipid-profile, duration of symptoms of GS and macrovascular complications.
A study limitation is that the questionnaire has not been validated. There may be a difference between those answering paper questionnaires (DM groups) and those answering online questionnaires (control group). Also, self-reported information from the control group about diabetes status, kidney status and height and weight may be inaccurate and patients with unclear food triggers were not categorized as having true GS which may have underestimated its prevalence. We also did not have complete retinopathy data for patients with T1DM, and we did not have eGFR data for either group. In the ADA statement on diabetic neuropathy from 2017, GS is limited exclusively to the head and neck region, 2 while we included sweating from anywhere on the body. Therefore, we may have overestimated the prevalence of GS.

| Future aspects
The highly individual perception of the symptoms and varied clini-

ACK N OWLED G EM ENTS
We thank associate professor in statistics Andreas Kryger Jensen for the invaluable input regarding statistical questions.

CO N FLI C T O F I NTE R E S T
None.

DATA AVA I L A B I L I T Y S TAT E M E N T
The data that support the findings of this study are available from the corresponding author upon reasonable request.