Passive heating and glycaemic control in non-diabetic and diabetic individuals: A systematic review and meta-analysis

Objective Passive heating (PH) has begun to gain research attention as an alternative therapy for cardio-metabolic diseases. Whether PH improves glycaemic control in diabetic and non-diabetic individuals is unknown. This study aims to review and conduct a meta-analysis of published literature relating to PH and glycaemic control. Methods Electronic data sources, PubMed, Embase and Web of Science from inception to July 2018 were searched for randomised controlled trials (RCT) studying the effect of PH on glycaemic control in diabetic or non-diabetic individuals. To measure the treatment effect, standardised mean differences (SMD) with 95% confidence intervals (CI) were calculated. Results Fourteen articles were included in the meta-analysis. Following a glucose load, glucose concentration was greater during PH in non-diabetic (SMD 0.75, 95% CI 1.02 to 0.48, P < 0.001) and diabetic individuals (SMD 0.27, 95% CI 0.52 to 0.02, P = 0.030). In non-diabetic individuals, glycaemic control did not differ between PH and control only (SMD 0.11, 95% CI 0.44 to -0.22, P > 0.050) and a glucose challenge given within 24 hours post-heating (SMD 0.30, 95% CI 0.62 to -0.02, P > 0.050). Conclusion PH preceded by a glucose load results in acute glucose intolerance in non-diabetic and diabetic individuals. However, heating a non-diabetic individual without a glucose load appears not to affect glycaemic control. Likewise, a glucose challenge given within 24 hours of a single-bout of heating does not affect glucose tolerance in non-diabetic individuals. Despite the promise PH may hold, no short-term benefit to glucose tolerance is observed in non-diabetic individuals. More research is needed to elucidate whether this alternative therapy benefits diabetic individuals.

Introduction fasting glucose and HbA 1C may be multi-factorial, but heat shock proteins (HSP) may play a pivotal role [35]. Upon physiological stress (e.g. heat, hypoxia, cancer) there is an increase in the amount of unfolded and misfolded proteins, causing cell damage [36][37][38]. The physiological stress is accompanied by a heat shock response, which triggers the release of HSP. The increased number of HSP available help facilitate correct folding of proteins, thus preventing cell damage. Both human and murine diabetic models are characterised by low intracellular (i) and high extracellular (e) HSP levels [39][40][41][42][43][44][45], promoting a pro-inflammatory state that reduces insulin sensitivity [46,47]. iHSP has direct protective effects whereas high eHSP is linked with insulin resistance [48,49]. Restoration of iHSP levels in diabetic models and subsequent favourable glycaemic control are mediated, in part, through reductions in inflammatory cytokines, c-Jun N-terminal kinase and IkappaB kinase, both associated with the inhibition of insulin signalling [50,51]. Importantly, PH has been shown to increase iHSP levels in diabetic and obese human and murine models [39,52,53].
Considering these findings, it would seem prudent to study PH and its effect on glycaemic control further. Indeed, reviews have highlighted how PH may benefit individuals with diabetes or those who are insulin resistant [27,35,45,46,[54][55][56][57][58], but to date, there is no systematic search, review and meta-analysis of PH and glycaemic control in diabetic and non-diabetic individuals. Therefore, the purpose of this study was to review and conduct a meta-analysis of published literature relating to PH and glycaemic control.

Search strategy
An electronic literature search was conducted in July 2018 using PubMed, Embase, and Web of Science. Searches were performed using Boolean operators and the following key terms and their combinations: glucose, insulin, diabetes, passive heating, sauna, thermal therapy, warm water, warm air, thermotherapy, and hot water immersion. The reference lists of all included studies were also examined to identify potentially relevant data sets that were not found in the original search (Fig 1).

Study inclusion and exclusion criteria
Studies were included in the review where: 1) a PH intervention, defined as any technique designed to increase body temperature using non-exercise models, was applied; 2) the experimental design included a non-heating control trial; 3) at least one primary outcome measure (i.e. glucose or insulin) was reported; 4) participants were human adults (i.e. aged �18 years), and 5) data were published in a peer-reviewed journal. There were no restrictions applied to the PH mode (e.g. water or air), exposure duration, participant sex, health status or study setting. Studies involving exercise in combination with PH were excluded. by two authors (MM, AH). The outcome measures extracted included 1) blood glucose and insulin; and 2) complications or adverse effects experienced attributable to the intervention. The authors of original investigations were contacted via email, as required, to clarify any queries relating to data or study characteristics as required. Any disagreement between the review authors extracting data was resolved by consensus or a third party (GM).

Risk of bias assessment
Risk of bias assessment was independently conducted in accordance with the Cochrane Handbook for Systematic Reviews of Interventions [59] by two authors (MM, AH). Potential sources of bias were classified as high, low or unclear in the areas of sequence generation, allocation concealment, blinding of participants and personnel, blinding of outcome assessment, incomplete outcome data, selective outcome reporting, and other bias. These outcomes  were visually summarised (Fig 2), with any disagreement between the authors' interpretation resolved by consensus or a third party (GM).

Statistical analyses
To measure the treatment effect, standardised mean differences (SMD) with 95% confidence intervals (CI) were calculated and analysed using a random effects model. Missing data prompted an email request to the study authors seeking this data and/or clarification as to why data were missing. The absence of standard deviations that could not be sourced from authors were calculated from available statistics as per Higgins and Green [59]. Further, where necessary, data were manually extracted from figures using WebPlotDigitizer [60]. Heterogeneity between comparable trials was evaluated using the I 2 statistic. Values of I 2 were interpreted using the following scale [59]: 0% to 40%, might not be important; 30% to 60%, moderate heterogeneity; 50% to 90%, substantial heterogeneity; and 75% to 100%, considerable heterogeneity.

Results
The search strategy identified a total of 2038 records (Fig 1). We also found nine potentially eligible studies from additional reference searches in these papers. Following removal of duplicates (n = 988), 1059 titles and abstracts were screened which resulted in 49 full-text articles being retrieved for eligibility. Articles were included in full-text screening if the abstract discussed PH in humans relating to either non-diabetic or diabetic individuals. After full-text screening, another 35 articles were excluded, mostly due to no relevant primary outcome or inappropriate control trial. Consequently, 14 articles were included in the final meta-analysis. Table 1 shows the characteristics of the included studies. The sample size varied from six [30, 61,62] to 32 [63]. In two studies, both non-diabetic and diabetic individuals were examined [63,64], while nine studies focused on non-diabetic individuals [30,62,[65][66][67][68][69][70][71] and three studies focused on diabetic individuals [61,72,73]. The studies using diabetic participants described their cohort as non-insulin-dependent [63], insulin-dependent [61,72], type 1 [73] and T2DM   [64]. The participants described as non-insulin-dependent received treatment with small doses of oral sulfonylureas [63]. The studies utilising insulin-dependent diabetics were described as receiving between 20 U to 27 U of intermediate or intermediate plus rapid-acting insulin in one of two injections per day [61,72], or 26 U to 56 U of long or long plus rapid-acting insulin in one or two injections per day [73]. The T2DM participants were taking Metformin [64].
All studies measured glucose concentration from venous blood samples. Insulin concentration was measured in five studies using non-diabetic individuals during PH and control trials following a glucose load [62,[65][66][67]69].

Risk of bias
Risk of bias via the Cochrane Collaboration's tool indicated that most of the information was from studies with low or unclear bias (Fig 2). Despite not being able to blind participants to a hot environment, performance bias and detection bias was low for all studies as participants were considered unable to change their glucose concentration consciously. Other biases included unclear or inappropriate statistical analyses and unclear handling, storage and analyses of blood samples.

Meta-analysis outcome
Glucose concentration. A summary of individual studies and meta-analysis for glycaemic control following a glucose load are shown in Figs 3 and 4. Fasting glucose concentration did not differ between control and PH in both non-diabetic (Fig 3) and diabetic individuals (Fig  4). Compared with control, glucose concentration was greater after 20-30 minutes of PH in non-diabetic individuals, which was a consistent observation at 40-60 minutes and 120 minutes of PH (Fig 3). In diabetic individuals, glycaemic control did not differ at any time point between PH and control trials, but the pooled overall effect was statistically significant, highlighting the potentially greater glucose concentration during PH (Fig 4).
When analysing data from experiments that did not administer a glucose load, glycaemic control did not differ between PH and control trials (Fig 5). Similarly, glycaemic control did not differ between PH and control trials when a glucose challenge was administered post-heating (Fig 6).
Insulin concentration. A summary of individual studies and meta-analysis of insulin concentration following a glucose challenge is shown in Fig 7. At each time point, insulin concentration was similar between control and PH.

Discussion
The primary aim of this meta-analysis was to investigate the effect of PH on glycaemic control in diabetic and non-diabetic individuals. Collectively, the meta-analysis showed PH resulted in a greater glucose concentration in diabetic and non-diabetic individuals (Figs 3 and 4). In contrast, glycaemic control does not differ between PH and control trials without a glucose load ( Fig 5). Finally, no favourable glycaemic control was observed following a glucose challenge within 24 hours of a single bout of PH (Fig 6).
Hormone changes may, in part, affect blood glucose concentrations during PH [4,29]. Considering glycaemic control does not differ between PH and control trials conducted without a glucose challenge (Fig 5) and insulin concentration does not differ during PH preceded with a glucose load (Fig 7), it is likely other factors may be at play. Arterialisation of venous blood [67] may be another contributing factor. Arterial blood glucose is consistently greater than venous blood glucose, but the difference between a vein in a heated hand and arterial samples is substantially smaller [74] owing to the opening of the arterio-venous anastomoses. This phenomenon presents a problem for oral glucose tolerance tests conducted in varying environments as the sampling technique may be a limiting factor. No study included in this metaanalysis used the heated hand technique [75]. Future studies should be mindful that arterialisation of venous blood may provide a methodological limitation to venous blood glucose sampling when comparing PH and thermoneutral trials. Heating the hand in both thermoneutral and PH trials may circumvent this limitation.
In contrast to non-diabetic individuals, glycaemic control was similar in PH and control trials at individual time points in diabetic individuals (Fig 4). However, the pooled overall effect indicated PH may elicit greater glucose concentrations. It is possible the reduced glucose extraction due to insulin resistance is partly responsible; thus arterialisation of venous blood yields minimal difference. Nevertheless, differences in hormonal responses influencing glucose output and uptake cannot be ruled out.
Providing a glucose challenge post-heating may be a more appropriate research design considering hormonal changes and arterialisation of venous blood may confound outcomes during PH. Glucose tolerance tests are also regularly conducted post-exercise, highlighting an insulin sensitising effect improving glycaemic control that is a relatively short-lived phenomenon (<48 hours) [76]. Limited studies were available for meta-analysis of post-heating glucose  [64] in non-diabetic individuals. Rivas et al. [64] did also measure glucose tolerance post-heating in T2DM individuals but, again, found PH not to influence glycaemic control. It is possible, as the author's state [64], they may have missed the window for improved insulin sensitivity or participant's continued medication may have influenced the glycaemic response. It is clear more work is needed with diabetic individuals to investigate glycaemic control following a single-bout of PH.
Our systematic search of the literature found no chronic randomised control trials investigating the effect of PH on glycaemic control in people with diabetes. Considering Hooper [6] provided evidence of the benefits of PH for those with T2DM nearly 20 years ago, it is surprising to find no other chronic study has been conducted. In the meantime, pharmaceutical agents have been developed to stimulate HSP production, replicating the response observed during PH and exercise [27,35,38,54]. Aside from Hooper [6], only three studies have investigated the effect of chronic PH in individuals with diabetes and other diseased states [15,77]. Specifically, PH (20 minutes, 3 days per week, over 3 months) improved perceived quality of life in people with T2DM [77], while Imamura et al., [15] reported fasting glucose was reduced with PH (45 minutes, 7 days a week, over 2 weeks) in those with coronary risk factors (e.g.  Passive heating and glycaemic control obesity, diabetes, hypertension). However, Masuda [21] reported no change in fasting glucose associated with PH (45 minutes, 7 days a week, over 2 weeks) in those with coronary risk factors. The reason for the discrepancy between the latter two studies is not clear.
Favourable glycaemic control following PH was initially attributed to an increased muscle blood flow facilitating glucose uptake [6,15]. Given muscle blood flow may increase during PH [78] and an increase in muscle blood flow may independently facilitate glucose uptake [34], the mechanism is logical. However, the present data show that PH does not acutely benefit glycaemic control in diabetic individuals, thus not supporting the proposed mechanism. Despite this, it is important to highlight the scarcity of data investigating PH and glycaemic control and, in particular, the PH use in diabetic individuals. Based on the limited human data available and current animal models it could be hypothesised that as an individual continues PH sessions over days and weeks, basal iHSP levels rise, eHSP levels fall. Alongside the increased release of nitric oxide and glucose transporter expression, the ratio change between iHSP/ eHSP associated with PH will reduce inflammatory cytokines which may improve insulin signalling to aid glycaemic control and reduce the high insulin output [39,45,50-52, [79][80][81][82][83]. Other factors such as improved appetite regulation (i.e. increased leptin concentration) and fat mass loss have been shown to occur with PH [6,64] and may work synergistically to improve glucose homeostasis. If an increase in iHSP is necessary for improvements in glucose homeostasis in diabetic individuals, then a sufficient heating stimulus is required. Exposure to 43˚C-53˚C air for 60 minutes increased body temperature by �0.5˚C [65,70], while immersion in 39˚C water for the same time increased temperature to~38.5˚C (~1.6˚C difference versus control) [64]. A deep body temperature rise of >0.8˚C has been shown to increase eHSP concentrations [84][85][86], but it is unknown whether this modest rise in deep body temperature is a great enough stimulus to increase iHSP concentrations. iHSP was increased in murine models where rectal temperature was held at 41.5˚C for~20 minutes [39,53], and where human whole-blood was incubated for two hours at 42˚C [52]. However, it is not ethically acceptable to maintain rectal temperature >39.5˚C in an attempt to elicit iHSP. Considering the impaired thermoregulatory control in individuals with diabetes [87], finding the lowest thermal stress required for health benefits should be a focus of future work.
No adverse events were noted in the studies reported in this meta-analysis. It is possible, however, with inappropriate PH protocols such an event may occur. For example, PH increases absorption of exogenously delivered insulin which could increase the likelihood of hypoglycaemic events [72,88]. Thermal sensations and thermoregulation may also be impaired in diabetic individuals [87,89] which could lead to burns and heat-related illness.
The strength of the meta-analysis presented here is that it combined data from 14 studies to estimate the effect of PH on glycaemic control with more accuracy than could be achieved in a single study. The risk of bias of the included studies appears low to unclear (Fig 2). If blinding of the participant and outcome assessment are ignored, then the risk of bias is mostly unclear. Importantly, there was a high risk of bias in selective reporting where authors simply failed to report variables or reported only 'good responders' [30]. However, the main limitation of this meta-analysis was the methodological differences amongst studies, including participants (e.g. duration of diabetes, age, sex), protocols (e.g. duration and modality of heating) and outcome assessment (e.g. timing of glucose measurement). These differences increase the heterogeneity in outcome measures. To account for heterogeneity, a random-effects model was utilised in the present meta-analysis; with more research in this area future analyses will be more robust and allow subgroup analyses.
In conclusion, this meta-analysis reveals an unclear picture of how PH may benefit glycaemic control in non-diabetic and diabetic humans. In non-diabetic individuals, glucose intolerance may occur when PH follows a glucose load. Importantly, however, this work highlights the paucity of research that has been conducted on this potentially beneficial, low-cost, intervention. Future research should focus on diabetic humans, using a randomised controlled trial design to measure glycaemic control in response to chronic PH. The benefits of PH may follow a dose-response relationship between the temperature and duration of heating [3]. Therefore, future work should determine the appropriate heat exposure to benefit glycaemic control in people with diabetes. If PH is found to be beneficial, then guidelines should then be developed with practical end-user constraints in mind.
Supporting information S1