Given that sleep disorders are known to be related to insulin resistance, and metformin has favourable effects on insulin resistance and on ventilatory drive, we sought to determine whether metformin therapy was related to sleep variables in a group of patients with Type 2 diabetes.
We performed a retrospective, observational study of our centre's database for patients referred for potential sleep disorders and then compared metformin-treated patients with those not treated with the drug. All study patients had undergone the same standard polysomnographic procedure. A multivariate analysis was performed to establish whether or not there was an independent relationship between metformin use and sleep variables (after adjusting for age, gender, BMI, neck circumference, cumulated risk factors and insulin use).
We studied 387 patients (mean ± sd age: 58.4 ± 10.8 years), of whom 314 had been treated with metformin. Total sleep time and sleep efficiency were higher in metformin-treated patients than in patients not treated with metformin [total sleep time: 6 h 39 min vs. 6 h 3 min, respectively (P = 0.002); sleep efficiency: 77.9 ± 12.3 vs. 71.5 ± 17.2%, respectively (P = 0.003)]. These differences persisted after adjustment for covariates and were observed even although metformin users had a higher BMI than did non-users (median 37.5 vs. 34.8 kg/m2; P = 0.045).
We showed that metformin therapy is associated with a longer sleep duration and better sleep efficiency. Randomized clinical trials are needed to confirm metformin's favourable effect on sleep quality and quantity.
Sleep disorders are particular frequent in states of insulin resistance [1-3]. Given that metformin exerts well-known effects on insulin resistance in humans [4] and directly increases ventilatory drive in non-obese rats [5, 6], the drug should (in theory) relieve sleep disorders. Our observations of a small series of patients referred for sleep disorders (n = 26) [7] showed that all but one of the patients not suffering from sleep apnoea syndrome were being treated with metformin; this finding prompted us to investigate the putative relationship between metformin and the relief of sleep disorders in a much larger population.
To this end, we studied patients with Type 2 diabetes referred for potential sleep disorders in our university medical centre and compared metformin-treated and metformin-naive patients in terms of sleep quantity and quality while taking account of possible confounding factors (BMI, neck circumference, insulin use, etc.).
We included consecutive, confirmed outpatients with Type 2 diabetes who had been referred for screening for sleep apnoea syndrome at Amiens University Hospital (Amiens, France) because of excessive daytime sleepiness, nocturnal snoring, disturbed sleep, nycturia or other symptoms. We excluded patients with unstable or uncontrolled cardiopulmonary diseases, a history of upper airway surgery, uncontrolled thyroid diseases and the sequelae of stroke.
When comparing metformin users and non-users, we systemically screened for the use of drugs with a sedative effect (hypnotics, opioids, neuroleptics and antidepressive, anti-dopaminergic and antihistaminic agents).
Patients were admitted for standard polysomnography from 19.00 to 08.00 h using the Brainnet System (Medatec, Brussels, Belgium) coupled to an electroencephalogram, an electrooculogram, an electromyogram, continuous nasal airflow measurement, measurement of thoracic and abdominal movements (using strain gauges) and oxygen saturation measurement with a pulse oximeter (Medatec).
Sleep apnoea syndrome was diagnosed according to the International Classification of Sleep Disorders criteria [8], on the basis of an apnoea/hypopnoea index over 15 events/h. Sleep efficiency was defined as the total sleep time divided by the total sleep period.
Multivariate analysis was performed to establish whether or not there was an independent relationship between metformin use and the prevalence of sleep disorders (after adjusting for several factors conventionally associated with disturbed sleep including age, gender, BMI, neck circumference, cumulated cardiovascular risk factors and insulin use).
Data are expressed as either the mean ± SD and the median (interquartile range) (quantitative data) or a percentage (qualitative data). Quantitative variables were compared between metformin users and non-users with a Student t-test or a Wilcoxon test according to the data distribution and a χ2- or Fisher test (as appropriate) were used for the comparison of qualitative variables between metformin users and non-users. The univariate association between sleeping variables (total sleep and sleep efficiency) and clinical factors were assessed using a Pearson or Spearman correlation (quantitative variables) or a Student t-test (qualitative variable). The independent association between metformin use and each of the sleep variables was assessed with an analysis of covariance and backward selection, after controlling for clinical factors with a univariate P-value < 0.2.
For comparison between metformin users and non-users, P-values were adjusted with the Bonferroni–Holm method to avoid inflation of a type I error. Test results with P-values ≤ 0.05 or > 0.05 were considered to be significant and non-significant, respectively. All statistical analyses were performed with SAS software (version 9.2; SAS Institute Inc., Cary, NC, USA).
Three hundred and eighty-seven patients were included in the study (mean ± SD age: 58.4 ± 10.8 years; male/female gender ratio: 0.64). Of these, 314 were being treated with metformin. There was no intergroup difference in demographic characteristics between metformin users and non-users, with the exception of BMI, number of anti-diabetic medications and proportion of patients on sulphonylurea therapy (Table 1). In contrast, the two groups differed significantly in terms of total sleep time and sleep efficiency [total sleep time: 6 h 39 min vs. 6 h 3 min for metformin users and non-users, respectively (P = 0.002); sleep efficacy: 71.5 ± 17.2 vs. 77.9 ± 12.3%, respectively (P = 0.003)] (Table 1).
| Demographic and sleep-related characteristics | Univariate analysis P-value | Mutivariate analysis P-value | |||||
|---|---|---|---|---|---|---|---|
| Metformin-treated patients n = 314 | Patients not treated with metformin n = 73 | P-value | Total sleep time | Sleep efficiency | Total sleep time | Sleep efficiency | |
| Age, years | 57.9 ± 10.7 (21–79) | 60.7 ± 11.4 (30–83) | 0.061 | 0.0024 | < 0.0001 | < 0.0001 | |
| Gender ratio, male:female | 64.6 | 61.6 | 0.630 | < 0.0001 | < 0.0001 | < 0.0001 | < 0.0001 |
| BMI, kg/m2 | 37.5 (32.5–43.3) | 34.8 (31.2–41.0) | 0.045 | 0.6325 | 0.8372 | ||
| Neck circumference, cm | 45.0 (42.0–47.0) | 43.0 (41.0–47.5) | 0.206 | 0.0431 | 0.0100 | ||
| Arterial hypertension,% | 77.3 | 76.7 | 0.912 | ||||
| Tobacco,% | 62.4 | 58.9 | 0.583 | 0.0039 | < 0.0001 | ||
| Hypercholesterolaemia,% | 56.7 | 54.8 | 0.769 | ||||
| Metabolic syndrome,% | 82.5 | 75.3 | 0.160 | ||||
| Number of risk factors | 4 (4–5) | 4 (4–5) | 0.115 | 0.7791 | 0.4252 | ||
| Number of anti-diabetic medications | 1.8 ± 0.8 (1–5) | 1.2 ± 0.5 (0–3) | < 0.0001 | ||||
| Metformin therapy | — | — | — | 0.0001 | 0.0002 | 0.0001 | 0.0004 |
| Sulphonylurea therapy,% | 28.3 | 54.8 | < 0.0001 | ||||
| Insulin therapy,% | 17.3 | 15.1 | 0.646 | 0.1892 | 0.6690 | ||
| Sedative/hypnotic drugs,% | 15.3 | 11.0 | 0.344 | ||||
| Total sleep period, h | 8 h 8 min (4 h 22 min–9 h 55 min) | 7 h 58 min (3 h 34 min–9 h 45 min) | 0.189 | ||||
| Total sleep time, h | 6 h 39 min (1 h 51 min–9 h 4 min) | 6 h 3 min (1 h 32 min–8 h 32 min) | 0.002 | ||||
| Sleep efficiency,% | 77.9 ± 12.3 (25.0–99.4) | 71.5 ± 17.2 (16.8–95.2) | 0.003 | ||||
| Apnoea/hypopnoea index | 17.0 ± 23.1 (0–112.9) | 19.9 ± 26.7 (0–112) | 0.397 | ||||
The groups of metformin-treated and metformin-naive patients did not differ significantly in terms of the prevalence of apnoea or hypopnoea (mainly obstructive; data not shown), although the BMI was higher in metformin-treated patients (median 37.5 vs. 34.8 kg/m2, P = 0.045).
The results of the univariate and multivariate analyses are presented in Table 1. After adjusting for covariates, the metformin-treated and metformin-naive groups still differed significantly in terms of total sleep time and sleep efficiency (P = 0.0001 and 0.0004, respectively).
Cross-sectional and longitudinal studies have shown a high prevalence of glucose intolerance, insulin resistance and diabetes in subjects with sleep disorders [9, 10] and, conversely, revealed that sleep disorders are independently associated with impairments in glucose metabolism [11]. Several mechanistic explanations have been suggested, including intermittent hypoxia, sleep fragmentation, sleep deprivation with secondary sympathetic activation, impairments of the hypothalamus–pituitary axis, generation of reactive oxygen species and elevated activity of inflammatory pathways—all of which ultimately lead to an insulin-resistant state and worsened glucose tolerance [12-14]. In view of metformin's well-known effects on insulin resistance, this anti-diabetic drug may therefore exert favourable effects on sleep. Data from animal experiments support this hypothesis [4, 5, 15]. However, these preclinical data did not have clinical counterparts. Here, we present the first evidence of a favourable, independent association between metformin therapy and sleep-related characteristics in a fairly large population.
The present study population was characterized by severe obesity (with a BMI close to 40 kg/m2 in metformin-treated patients) and a high cardiovascular risk (with more than four risk factors present per patient, on average). Most of the patients had the metabolic syndrome. However, whereas our study population was characterized by severe obesity, the mean numbers of apnoea/hypopnoea episodes were suggestive of mild sleep apnoea syndrome. This should be borne in mind when considering that the metformin-treated and metformin-naive groups did not differ significantly in terms of sleep apnoea syndrome.
The two study groups differed significantly in terms of two important sleep variables—one related to sleep quantity (total sleep time) and the other related to sleep quality (sleep efficiency) (P < 0.002 and 0.003, respectively). Our multivariate analysis provided the most striking finding of the present study. As expected, we found independent associations between sleep variables and some demographic variables (but not BMI, surprisingly). It is important to note that, after adjustment for covariates, this association persisted only for age, gender and metformin therapy—even although the metformin-treated patients had a higher BMI than did the metformin-naive patients (P = 0.045).
Our study had several limitations: (1) some data (such as the metformin dose, the duration of metformin therapy and the HbA1c value) were not available; (2) there were relatively few metformin-naive patients; and (3) the metformin and metformin-naive study groups differed significantly in terms of the number of concomitant anti-diabetic medications.
In conclusion, we provide new evidence of an independent effect of metformin on sleep quantity and quality. Randomized clinical trials are needed to confirm this beneficial association and to determine whether metformin's ability to lower blood glucose levels might also be related to its effect on sleep (along with its known mechanisms of action), as sleep disorders are associated with elevated glycaemia [2, 3, 7].
None.
None declared.
1464-5491/asset/olalertbanner.jpg?v=1&s=e2ae1ff5ce421223aa62253d943b6d9e6d7ee9e6)