Impact of sleep and physical activity habits on real‐life glycaemic variability in patients with type 2 diabetes

The aim of this study was to better characterise whether sleep habits, eating schedule and physical activity in real‐life are associated with glycaemic control in patients with type 2 diabetes. A total of 28 patients (aged 60 years [58; 66], 54% female) with type 2 diabetes treated with basal‐bolus insulin therapy administered by insulin pumps were analysed. Glycaemic data measured by Flash Glucose Monitor System, physical activity and sleep data measured by accelerometer, and meal schedules were simultaneously collated with insulin pump administration data, for 7 days in real‐life. Their respective impact on the time spent in target, in hypoglycaemia, in hyperglycaemia and on glycaemic variability was evaluated. Multiple regressions showed that the total daily dose of meal boluses of insulin was inversely associated with the coefficient of variation (CV; coefficient β = −0.073; 95% confidence interval: −0.130, −0.015; p = 0.016), as well as sleep duration. The higher the sleep duration, the lower the glycaemic variability (coefficient β = −0.012; 95% confidence interval: −0.023, −0.002; p = 0.027). The mean 7 days physical activity of the subjects was very low and was not associated with glycaemic control on the 7 days mean values. However, days with at least 1 hr spent in physical activity higher than 1.5 METs were associated with less glycaemic variability that same day. This real‐life observation highlights the importance of sufficient sleep duration and regular physical activity to lessen the glycaemic variability of patients with type 2 diabetes.


| INTRODUCTION
Type 2 diabetes is a combination of insulin resistance and a relative deficiency of insulin secretion. This relative lack of insulin secretion increases over time and may lead to the need to introduce exogenous insulin. Usually, patients initially require the introduction of basal insulin, and then insulin requirements may increase further until basal-bolus therapy is required, like that used in the treatment of type 1 diabetes. This treatment is most often administered by multi-injection, but can be administered by insulin pump when the glycated haemoglobin (HbA1c) is not sufficiently improved by multi-injection (Metzger et al., 2017). Patients needing an insulin pump are mainly those with the longest disease duration and the most severe insulin resistance.
The HbA1c is a valuable measure of population health, and remains a validated indicator of glycation as a risk factor for complications.
However, it is not as helpful for personalised diabetes management mainly because it provides only an average of glucose levels over the previous past 3 months, and does not detect hypoglycaemia or hyperglycaemia daily. For some years now, patients have been able to benefit from tools that allow them to monitor their blood glucose levels with intermittently scanned continuous glucose monitors (CGMs). This technology provides the current glucose value plus retrospective glucose data for a specified 8-hr period upon "scanning", that is by swiping the device close to the sensor. This system is also known as "flash" monitoring. The glycaemic control of patients, beyond the mean results represented by HbA1c every 3 months, is now assessed through several glucose metrics, for example, the time spent in a range of glycaemia from 70 mg dl À1 to 180 mg dl À1 ; the time spent below and above this range; the glycaemic variability (Danne et al., 2017).
Several lifestyle habits that may be associated with glycaemic control, that is, physical activity, sleep and feeding schedules, have not been studied in this specific group of patients with type 2 diabetes who require insulin administered by an insulin pump.
In the present work, we sought to document whether physical activity, sleep habits and feeding schedules can influence glycaemic metrics in patients living with type 2 diabetes and treated by insulin pump therapy.

| PATIENTS AND METHODS
A total of 25 patients were included from the population of patients with type 2 diabetes followed at home for their insulin pump treatment by the same medical-technical home care provider in the Grenoble area, France. These patients had to be on stand-alone pumps, without sensor integration. A total of 28 patients with complete data were analysed. The reasons for exclusion from final analyses of these seven patients were the following: one patient had 306 IU per day of insulin doses and thus had to regularly change his insulin pump reservoir. As a result, there were frequent interruptions in pump data acquisition that did not allow for correct data acquisition. The remaining six patients were excluded from the analysis because they had more than 21% of data not acquired by the Flash Glucose Monitor System (FGMS), due to an insufficient number of scans. The median age of these seven unanalysed patients was 67 (60; 70) years, 14% female sex. The median disease duration was 21 (18; 27) years. The median HbA1c was 7.0 (6.9; 7.5)%.

| Inclusion criteria
Patients were aged 40-75 years, with type 2 diabetes treated by an insulin pump for at least 6 months, and with a body mass index (BMI) between 27 kg m À2 and 40 kg m À2 . The total daily insulin dose had to be between 40 U per 24 hr and 300 U per 24 hr to avoid extreme heterogeneity of the group. The patient had to be equipped with a FGMS FreeStyle Libre ® (Abbott Laboratories).
The study was approved by ethic committees (IDRCB de DT2_1: 2020-A01710-39), and all patients signed an informed consent. The protocol was published on clinicaltrial.com (NCT04522882).

| Objective
To measure the potential associations of sleep, physical activity and feeding schedules with glucose metrics in patients with type 2 diabetes treated with insulin pumps for seven consecutive days.

| End-points
Search for associations between the variables of physical activity, sleep and mealtimes on the time in range (TIR) 70-180 mg dl À1 over 7 days of continuous recording of interstitial glucose level measured by FGMS, percentage of time spent in hypoglycaemia and hyperglycaemia, glycaemic variability measured by the CV and standard deviation (SD) of glucose levels.

| Continuous 7-day interstitial glucose measurements
Interstitial glucose measurements were collected from data of the freestyle FGMS (Abbott ® ) over a 7-day period. This system allows intermittently scanned CGM. Patients using this system are blinded from continuous data. They have access to their glucose level and to an arrow indicating the evolutionary trend when they voluntarily scan themselves.

| Glucose metrics
Glucose metrics are extracted from CGM measurements (Danne et al., 2017). TIR refers to the time spent in a target glucose range

| Continuous measurement of physical activity over 7 days
Over the same 7-day period, physical activity was collected by actimetry (ActiGraph ® GT3X Actimeter). It was expressed as the average 7-day diurnal physical activity in metabolic equivalents of task (METs).

| Measurement of sleep patterns over 7 days
Total sleep time and chronotype (defined by the midnight time point) were collected by actimetry and compared for validation of the automatic analyses with a sleep diary. Figure 1 shows the simultaneous collection of all these data.

| Collection of meal and snack schedules over 7 days
Meal and snack schedules were collected in a diary. Food content and quantity were voluntarily not collected to avoid desirability bias that could affect food choices (Novotny et al., 2003;Rumpler et al., 2008).
Evening snack was defined by any kind of food intake taken after dinner and before night sleep, ≥ 4 days out of 7.

| Statistics
Descriptive data are reported by median and interquartile range for quantitative variables, and by number and frequency for qualitative data.
Univariate Spearman correlations were performed to look for an association between the independent variables and time spent in the 70-180 mg dl À1 target, time spent in hyperglycaemia and hypoglycaemia, with the CV of glucose levels and SD of glucose levels. A Mann-Witney test was used for categorical variables.
Multivariate analyses were then performed including the independent variables that were associated with the dependent variables with a value of p < 0.2 in univariate analyses. Log transformation of the variables was performed if necessary to obtain normality of the distribution of the residuals.

F I G U R E 1
The overlay of actimetry data with physical activity in metabolic equivalents of task (METs), sleep schedules (blue panel), blood glucose levels measured every 5 min by the Flash Glucose Monitor System (FGMS). Mealtimes are indicated by the vertical black bars with the amount of carbohydrates at meals because this particular patient used the "bolus assistant" function. The basal insulin rate from the pump and the amount of insulin injected at mealtimes and for additional "catch-up" boluses are also reported To further study the relationships between physical activity, sleep behaviour and glycaemic metrics, days with physical activity were compared with days without. Physical activity periods were defined from actigraphy data as a physical activity above 1.5 METs that occurred for at least 1 hr. In addition, glucose metrics were compared between sleeping and wake periods. Sleeping periods were extracted from the actigraphy data. Awake periods were defined as non-sleeping periods.
Comparison of glucose metrics was performed by Mann-Witney tests.
Statistics were performed with the Python software Python 3.7.5 with the packages pandas 1.0.5, numpy 1.19.5, scipy 1.5.2, statsmodel 0.12.1. A value of p < 0.05 was chosen as the significance threshold.

| RESULTS
Patient characteristics are reported in Table 1.

| Univariate analyses
Univariate analyses between explanatory and dependent variables: time spent in the 70-180 mg dl À1 target, percentage of time spent in hypoglycaemia, percentage of time spent in hyperglycaemia, CV and SD are reported in Table S1. All variables that were associated with these dependent variables with a value of p < 0.2 were then included in multivariate regression models to determine independent associations.

| Multivariate analyses
Results are reported in Table 2

| Additional analyses
Comparison of the glycaemic metrics between sleep and wake periods showed that the glucose control was less good during awake time compared with sleep periods: TIR was lower, time in hyperglycaemia was higher as well as the CV while patients were awake. There was no difference for time spent in hypoglycaemia ( Table 3).
Comparison of the glycaemic metrics on days without any physical activity with days with at least 1 hr of physical activity above 1.5 METs shower a lower CV on days with physical activity (Table 4).

| Sleep, physical activity and meal schedule in glycaemic control
In type 1 and type 2 diabetes, several studies have shown that sleep patterns have an impact on glycaemic control (Larcher et al., 2015;Larcher et al., 2016;Reutrakul et al., 2016). Some studies have shown the importance of sleep duration, with an association of short (< 6-7 hr) and long (> 8-9 hr) sleep duration and poorer glycaemic control (Brouwer et al., 2020;Martorina & Tavares, 2019;Mokhlesi et al., 2019;Whitaker et al., 2018). The results of our study seem to agree with this, showing that short sleep duration was associated with more glycaemic variability. On the other hand, we did not find an association between long sleep duration and a worse glycaemic profile.
Physical activity is a component that is reported to be a major determinant in the glycaemic control (Igarashi et al., 2021). For example, interrupting a sedentary lifestyle with 3 min of moderate-intensity walking or resistance exercise every 30 min improves glycaemic control, including the following night (Dempsey et al., 2017). We observed indeed that having a physical activity above 1.5 METs for at least 1 hr was associated with less glycaemic variability that day.
However, in this 7-day real-life study, a very low level of overall physical activity among patients was observed, with a lack of impact on the 7-days glycaemic metrics. A previous study investigated in patients with type 2 diabetes whether the addition of an accelerometer to the continuous glucose measurement data was of interest to feed a closed-loop algorithm. The contribution of the accelerometer was negligible in improving the algorithm, which seems to be confirmed by the results presented here (van Doorn et al., 2021). This finding shows that the physical activity habits of the subjects in real-life remain low despite health education and recommendations.
T A B L E 3 Comparison of blood glucose metrics between sleep and wake periods Meal timing has been associated with diabetes control in patients with type 2 diabetes in previous studies. Indeed, a late chronotype was associated with poor glycaemic control (Reutrakul et al., 2013).
This effect was mediated in part by the intake of more calories at the evening meal. In the study by Vera et al. (2018), it was shown that overweight or obese subjects had a poorer metabolic profile when they had a late chronotype, with notably greater insulin resistance.
The late chronotype was associated with less favourable eating behaviour, more emotional eating and, of course, later mealtimes. In contrast, the late chronotype was not associated with higher caloric intake. In the present study, however, we found no association of chronotype or mealtime with glucose measurements in these patients with type 2 diabetes and treated by an insulin pump.

| Impact of lowering glycaemic variability for patients
Glycaemic variability is a process characterised by the amplitude, frequency and duration of the fluctuation. Both the amplitude and the timing of blood glucose fluctuations contribute to the risks for hypoglycaemia and hyperglycaemia associated with diabetes (Danne et al., 2017;Kovatchev & Cobelli, 2016). Glucose variability in type 2 diabetes has been associated with atherosclerosis (Temelkova-Kurktschiev et al., 2000) and oxidative stress (Monnier et al., 2006). It is also related to the frequency of hypoglycaemia (overall and severe; Cox et al., 1994;Monnier et al., 2017). A recent review of the relationships between these new glycaemic metrics derived from CGM data showed that glycaemic variability and low TIR showed associations with all microvascular and macrovascular complications of diabetes. Notably, a higher TIR was associated with reduced risk of albuminuria, retinopathy, cardiovascular disease mortality, all-cause mortality and abnormal carotid intima-media thickness. Peripheral neuropathy was predominantly associated with SD of blood glucose levels and mean amplitude of glycaemic excursions, another marker of glucose variability (Yapanis et al., 2022). In this respect, the patients in our study had low glycaemic variability and a low risk of hypoglycaemia. However, we found that physical activity and longer sleep duration were associated with lower glycaemic variability. Increasing sleep duration by 1 min was associated with a reduction of 0.012% of the CV. Thus, one more hour in sleep duration could lower the glycaemic variability by 0.72%. A recent meta-analysis of intervention studies aiming at extending sleep duration found that in 26 randomised clinical trials, interventions led to a standard mean difference of 0.80 (95% CI 0.28-1.31; p < 0.01; I 2 = 99.2%) hr of sleep duration between intervention and control groups (Baron et al., 2021). There is a hope that intervention studies to extend sleep duration may lower glycaemic variability to an extent that could be clinically significant, in particular to prevent peripheral neuropathy.

| Strengths and weaknesses of the study
The strengths of this study are that it simultaneously collected objective sleep patterns, physical activity levels and meal schedules in real-life, in association with continuous subcutaneous glucose measurement in a particular population of patients having type 2 diabetes and treated by insulin pump therapy. A limitation of this study is that the content of food intake, both qualitatively and quantitatively, was not collected, but only the timing of meals. We cannot therefore measure their impact on glycaemic variability. We made this methodological choice to avoid various possible biases that are known to occur in dietary surveys: indeed, the desirability bias may lead either to underreporting of consumption or to the subject eating differently during the week of collection, in order to adhere to what he/she imagines to be the investigator's expectations (Novotny et al., 2003;Rumpler et al., 2008). It could also be questioned whether we should have used more detailed information regarding sleep habits. Indeed, actimetry measurement allows to distinguish naps from night sleep, as well as to measure social jetlag. We chose to restrain our analyses to sleep duration and chronotype, excluding naps and social jetlag for two reasons: most of our patients were inactive (mostly retired); and social jetlag was less relevant without working days. The presence of naps was highly variable between subjects, with some not taking any and others taking several per day or having days with and without naps, leading to very heterogeneous information on a limited number of patients.
We therefore added naps to the total sleep time of each day. For the same reason, this heterogeneity of sleep pattern between naps and nights did not allow us to compare the glycaemic variability during the day according to the duration of sleep the previous night. This interesting information would require a larger number of patients observed over a longer period.

| CONCLUSION
In this real-life observational study, we found that physical activity was overall very low and, thus, had no impact on subjects' 7-days glycaemic metrics. However, a physical activity of at least 1 hr above 1.5 METs was associated with less glucose variability that same day. Short sleep duration was associated with greater glycaemic variability. These observations underline the importance of considering sleep hygiene for the metabolic health of patients with type 2 diabetes, in addition to promoting healthier eating habits and regular physical activity.

AUTHOR CONTRIBUTIONS
Pierre Gauthier performed data management and statistical analyses.
Chesner Desir contributed to statistical analyses. Maud Plombas participated in study design. Eloïse Joffray participated in the study design and follow-up, patients' recruitment and data collection.
Pierre-Yves Benhamou participated in study design and paper reviewing for important intellectual content. Anne-Laure Borel participated in study design, patient recruitment and wrote the present paper.

DATA AVAILABILITY STATEMENT
Embargo on data due to commercial restrictions.