Introduction

Type 2 diabetes mellitus (T2DM) is prevalent in people living with HIV (PLWH) [1, 2]. In Thailand, the prevalence of T2DM and prediabetes among PLWH are 14% [3] and 27.5% [4], respectively. The prediabetes prevalence in this group was higher than the 10.6% rate in the general Thai population [5].

Intensive lifestyle interventions (ILI) are effective in reducing diabetes rates and are generally recommended to the general population and PLWH with prediabetes [6,7,8]. Whether or not the unique characteristics of PLWH affects the outcomes is unknown. To date, there have been only a few studies that explored the effectiveness of ILI in PLWH with impaired glucose tolerance. A non-randomized study of 28 PLWH that had impaired fasting glucose showed that ILI for 6 months improved glucose, blood pressure and lipids levels [9]. Another study showed that a supplementation of pioglitazone to an exercise program for 4 months improved insulin sensitivity in PLWH that had insulin resistance and central obesity [10]. Therefore, understanding the acceptability and efficacy of ILI along with factors that may modulate the responses, is crucial in reducing the diabetes risk in this group.

In people with normoglycemia, prediabetes and T2DM, sleep disturbances, including insufficient sleep, poor sleep quality and increased sleep variability (varying day-to-day sleep timing or duration), have been linked to abnormal glucose metabolism [11] and cardiometabolic risks [12,13,14]. Studies in non-HIV people have shown that prediabetes individuals had shorter sleep durations (< 5–6 hours (h)/night) [15, 16] when compared to normoglycemic individuals. When compared to non-HIV infected people, PLWH had a higher prevalence of poor sleep quality [17, 13]. For each individual, the mean across all available nights was used.

Sleep quality in the previous month was assessed using a validated Thai version of the Pittsburgh Sleep Quality Index (PSQI) [31].

Physical activity assessment

The physical activity was assessed by the Global Physical Activity Questionnaire (GPAQ) version 2 and measured as Metabolic Equivalents (METs) [32, 33].

Quality of life

Quality of life was measured using the European Quality of Life Measure-5 Domain-5-Level (EQ-5D-5L) [34, 35] and the visual analog scale [VAS; scale ranged from 0 (the worst) to 100 (the best)].

The six-month intensive lifestyle intervention (ILI) program

This pilot six-month ILI program was led by a multidisciplinary team of diabetes nurses and dieticians, and included a follow up at 12 months. The program was a group-based activity adapted from the Diabetes Prevention Program (DPP)-Thailand project, which was shown to be effective in improving glycemic parameters [36]. The goals of the program were to target three main aspects of a healthy lifestyle including healthy eating (aiming 7% weight loss in overweight individuals), exercise (150 min of aerobic exercise weekly) and healthy co** (for example, practicing in spiritual and mindfulness, emotional management). This modified program consisted of five one-day sessions which met every six weeks over the 6-month period, and supplementary weekly telephone follow-ups (10–15 minutes). Each session included instructions from the health care team, through workshop activities and the setting of mutually-agreed behavioral goals. In-between-visit telephone follow-ups were aimed at encouraging goal achievement, discussion of problems and enhancing health knowledge.

To evaluate the acceptability of the intervention, questionnaires which addressed the participants’ satisfaction of their experience were administered 4 times, at the end of each session (1, 2, 3 and 4).

Statistical analysis

Comparisons of the differences between normoglycemia and prediabetes groups were performed by Student t Test, Mann–Whitney U test and Pearson’s chi-squared test, as appropriate. Comparison of metabolic parameters at baseline and 6 months, and baseline and 12 months, were performed by paired-samples T-test or Wilcoxon Signed Ranks Test. Spearman’s correlation was used to investigate the association between changes in body mass index (BMI), physical activity, sleep parameters and changes in metabolic parameters at 6 months compared to baseline. A multiple linear regression analysis was further performed to investigate the association between changes in sleep parameters and changes in metabolic parameters after adjusting for the changes in BMI and physical activity. Analyses were performed using SPSS statistical software package, version 18.0 (SPSS, Chicago IL, USA).

Results

Baseline demographics, metabolic and sleep characteristics in normoglycemic and prediabetic individuals

Thirty-nine PLWH (20 normoglycemia and 19 prediabetes) with a mean age of 51.5 ± 6.0 years were included (Table 1). There were no differences in clinical characteristics between the two groups (Table 1). As expected, individuals with prediabetes had higher FPG, 2 h-PG, HbA1c, and HOMA-IR than those in the normoglycemia group. Other indices obtained from OGTT (Matsuda, Insulinogenic and Disposition Indices), lipid profiles and all sleep parameters were similar between the two groups.

Table 1 Baseline clinical characteristics of 20 PLWH with normoglycemia and 19 PLWH with prediabetes
Full size table

Results of ILI

Thirteen PLWH with prediabetes completed the six-month ILI program and a 12-month follow-up visit. At 6 months, there was a significant reduction in BW and BMI from baseline, which was maintained at 12 months (Table 2). Waist circumferences (WC) significantly decreased at 12 months (Table 2). There were no changes in physical activity during the follow-up period. FPG non-significantly decreased at 6 months (p = 0.051), whereas there were no changes in 2 h-PG and HbA1c at 6 and 12 months. For OGTT-derived indices, when compared to baseline, the Matsuda index non-significantly increased at 12 months (p = 0.064), HOMA-IR significantly decreased at both 6 and 12 months, and an insulinogenic index non-significantly decreased at 6 months (p = 0.055).

Table 2 Compared parameters at baseline, 6 months (at the end of the program) and 12 months in 13 PLWH with prediabetes
Full size table

For sleep parameters, sleep efficiency (a marker of sleep quality) significantly increased at 6 months, whereas there were no changes in other parameters (Table 2).

Quality of life assessed by EQ-5D-5L did not change. However, quality of life as assessed by VAS significantly increased at 6 months.

Relationship between changes in BMI, physical activity, sleep parameters and changes in metabolic parameters during ILI

We further investigated the association between changes in BMI, physical activity, sleep parameters and changes in metabolic parameters at 6 months (Table 3). A decrease in BMI correlated to an increase in the Matsuda index. An increase in physical activity was associated with a reduction in HbA1c and insulinogenic index. For changes in sleep parameters, an increase in sleep duration was associated with a reduction in HbA1c and insulinogenic index. However, the associations became non-significant after controlling for changes in BMI or physical activity. In addition, an increase in sleep variability (SD sleep duration and SD MST) was significantly associated with an increase in 2 h-PG. The association between ∆SD sleep duration and 2 h-PG remained significant after adjusting for changes in BMI (b = 0.603, 95% CI: 0.146, 1.059, p = 0.015) or physical activity (b = 0.774, 95% CI: 0.339, 1.209, p = 0.003), but the ∆SD MST and 2 h-PG became non-significant after adjustment.

Table 3 Correlation between changes in BMI, physical activity, sleep parameters and changes in metabolic parameters between baseline and 6 months in 13 PLWH with prediabetes
Full size table

This ILI program was well-accepted. Over 90% of participants were satisfied. The program attendance was 84.6–100%, suggesting that the program was feasible in this setting.

Discussion

In this pilot study of 20 normoglycemic and 19 prediabetic PLWH, we did not find differences in sleep characteristics between the two groups. The pilot six-month ILI program for PLWH with prediabetes was effective in improving diabetes risks, including a reduction in BW, BMI, WC and HOMA-IR, which were maintained at 12 months. The program was feasible and acceptable, as evaluated by the participants’ attendance and satisfaction. Furthermore, the changes in sleep variability were a novel factor associated with changes in metabolic parameters after ILI. An increase in SD sleep duration was significantly associated with increased 2 h-PG regardless of the changes in BMI and physical activity. These results from the present study suggest that ILI is effective and feasible in reducing diabetes risks in PLWH with prediabetes, and that sleep modification could possibly be complimentary to ILI. Therefore, a larger dedicated RCT should be performed to confirm these findings.

Of note, the magnitude of BW reductions in the present study was smaller than the pre-specified goal of the program. However, this was comparable to the findings from a Community-Based DPP in Thailand [36]. This might be due to the lower baseline weight. Nevertheless, the weight reduction was accompanied by a significant reduction in HOMA-IR over the 12-month period.

This is the first study demonstrating that objectively measured sleep duration and sleep variability may play a role in the effectiveness of ILI in PLWH. However, the findings should be interpreted with caution due to the small number of subjects. Insufficient sleep demonstrated to increase diabetes risk in some studies [11, 37]. Furthermore, maintaining adequate sleep was also shown to be beneficial during a weight loss study [38]. In a study of ILI, individuals with prediabetes who reported ≤ 6 h/night of sleep at enrollment had a significantly higher rate of incident diabetes, as well as less weight loss [39], compared to those slee** 7 h. Our results in PLWH are in agreement with this data, as we demonstrated that a decrease in objectively measured sleep duration was associated with an increase in HbA1c. In addition, this is the first study that demonstrated that increased sleep variability was associated with an increase in 2 h-PG after ILI. Increasing ten minutes of SD sleep duration increased 2 h-PG 6 mg/dL after adjusting for changes in BMI, and increased 2 h-PG 7.7 mg/dL after adjusting for changes in physical activity. Sleep variability has been increasingly recognized to be related to metabolic health. This could potentially be related to the shift in timing of sleep which could affect the body’s circadian regulation, and hence, affect metabolism. Sleep variability has been shown to be associated with glycemic control in people with type 1 diabetes [13, 14], less physical activity [40], higher prevalence and incidence of metabolic abnormalities [41], and an unhealthy diet [42]. Our pilot results suggested that maintaining adequate and regular sleep could potentially be beneficial in reducing diabetes risks during ILI in PLWH.

In conclusion, intensive lifestyle interventions in PLWH with prediabetes is feasible and effective in improving metabolic control. The preliminary results suggest that the effects of ILI are modulated by sleep duration and sleep variability. This highlights a possibility to apply the ILI program and sleep adjustment as a diabetic prevention program in this high-risk group for T2DM.

Limitation

We recruited only PLWH in this study, therefore, we lacked a comparison of the differences in sleep variability and glucose metabolism in individuals without HIV. In this study, participants in normoglycemia and prediabetes groups were not matched, however, the demographic and anthropometric parameters were similar. In addition, we recruited PLWH with prediabetes who received ART with complete viral suppression. Thus, our results may not be applicable to PLWH without successful ART. Since the cultural and socioeconomic needs vary between different ethnic groups, the results may not be implied for non-Asian populations. The number of study participants was small. Lastly, more detailed objective sleep and metabolic measurements such as body fat composition, hyperinsulinemic euglycemic clamp and OSA assessment were not performed in this study.”