Background
Diagnostic-level sleep disorders affect between 5% and 20% of the general population in Western countries [
1,
2]. Despite current guidelines that recommend non-medication-based treatment methods for long-term sleep disorder [
1], the rate and use of prescribed sleep medication is frequent, especially among older populations [
3‐
7]. Patients with type 2 diabetes (T2D) are an important patient group in this regard, as they are more likely than patients without T2D to suffer from sleep problems [
8,
9].
As is well acknowledged, long-term exposure to sleep medication is associated with pervasive adverse health outcomes, including an increased risk of falls and fractures [
10], cognitive impairment [
11], and suicidal attempts [
12]. It also often leads to dependence [
13], which is one reason why the majority of patients report ceasing long-term sleep medication to be difficult [
5]. With respect to T2D, melatonin [
14,
15], which is commonly used sleep medication, may also play a role in glucose metabolism, although no clear consensus on this has been reached. Moreover, the sleep problems behind prescribed sleep medication tend to confuse conclusions; that is, sleep problems have a negative impact on glucose balance [
16], whereas improved sleep due to treatment may result in better outcomes [
8].
A recent cohort study estimated that around 16% of all T2D patients have been prescribed recurrent sleep medication [
9]. However, less data exist on the prevalence of recurrent sleep medication prescriptions in a primary care setting. This knowledge would be highly valuable in the design of T2D treatment guidelines targeted at primary care clinicians. Moreover, there may be cultural differences in the prescription tendency of sleep medications, so guidelines on recommended medication are not necessarily fully comparable [
1,
17,
18].
Therefore, we sought to primarily explore (a) what is the prevalence of recurrent sleep medication prescriptions among Finnish primary care T2D patients and (b) which are the most common medications prescribed. As a secondary aim, we studied whether the T2D patients who had been prescribed recurrent sleep medication (divided into three subcategories: previous, ongoing, subsequent medication), differed from those who had not been prescribed sleep medication in terms of sleep disorder diagnosis and several background and treatment target variables, including the achievement of glycosylated hemoglobin A1c (HbA1c), low-density lipoprotein (LDL), and systolic blood pressure (sBP) goals. We hypothesized that recurrent sleep medication prescriptions are common, and that melatonin is the most often prescribed medication.
Results
The final study sample consisted of 4,508 T2D patients who had data on sleep medication, and for whom HbA1c and LDL target achievements were available. Their average age was 69.9 years (SD: 11.9) and 45.8% were females (Table
1). Sleep apnea was diagnosed among 12.5% and a non-apnea sleep disorder among 12.2% of all the patients. A total of 70.7% and 54.9% had achieved the treatment targets of HbA1c and LDL, respectively, whereas 46.1% (n = 620/1,346) had attained the sBP goal.
Table 1
Characteristics of study sample, stratified by recurrent sleep medication prescription. Percentages and numbers presented and full data available, unless otherwise indicated
Age, mean (SD) | 73.4 (11.8) | 68.5 (11.7) | 69.9 (11.9) | < 0.001 |
Sex | | | | < 0.001 |
Females | 53.9 (683) | 42.6 (1,381) | 45.8 (2,064) | |
Males | 46.1 (583) | 57.4 (1,861) | 54.2 (2,444) | |
Sleep medication (percentage of ‘yes’ responses are presented) | | | | |
Benzodiazepine-like medication | 56.9 (720) | 1.0 (33) | 16.7 (753) | < 0.001 |
Melatonin | 44.4 (562) | 0.9 (28) | 13.1 (590) | < 0.001 |
Mirtazapine | 35.8 (453) | 0.5 (17) | 10.4 (470) | < 0.001 |
Temazepam | 19.2 (243) | 0.3 (11) | 5.6 (254) | < 0.001 |
Nitrazepam | 0.2 (3) | 0 (0) | 0.1 (3) | 0.006 |
Doxylamine | 0.5 (6) | 0 (0) | 0.1 (6) | < 0.001 |
Doxepin | 0 (0) | 0 (0) | 0 (0) | |
Trimipramine | 5.4 (68) | 0.1 (4) | 1.6 (72) | < 0.001 |
Trazodone | 0 (0) | 0 (0) | 0 (0) | |
Number of sleep medications, median (IQR) | 10.0 (4.0–25.0) | NA | | |
Time in months between sleep medication prescriptions, median (IQR) | 2.94 (1.19–6.17) | NA | | |
Sleep apnea diagnosis | | | | < 0.001 |
Yes | 23.4 (296) | 8.3 (268) | 12.5 (564) | |
No | 76.6 (970) | 91.7 (2,974) | 87.5 (3,944) | |
Non-apnea sleep disorder diagnosis | | | | < 0.001 |
Yes | 22.0 (278) | 8.3 (270) | 12.2 (548) | |
No | 78.0 (988) | 91.7 (2,972) | 87.8 (3,960) | |
Body mass index, mean (SD) | 29.3 (6.3) | 30.0 (5.8) | 29.8 (6.0) | < 0.001 |
Missing
| n = 231 | n = 784 | n = 1,015 | |
Depression diagnosis | | | | < 0.001 |
Yes | 20.1 (255) | 7.2 (234) | 10.8 (489) | |
No | 79.9 (1,011) | 92.8 (3,008) | 89.2 (4,019) | |
Long-acting insulin | | | | < 0.001 |
Yes | 32.1 (406) | 21.8 (707) | 24.7 (1,113) | |
No | 67.9 (860) | 78.2 (2,535) | 75.3 (3,395) | |
Antihypertensive medication of any kind | | | | < 0.001 |
Yes | 85.6 (1,084) | 78.5 (2,544) | 80.5 (3,628) | |
No | 14.4 (182) | 21,5 (698) | 19.5 (880) | |
Lipid-lowering medication of any kind | | | | 0.219 |
Yes | 67.9 (860) | 66.0 (2,140) | 66.5 (3,000) | |
No | 32.1 (406) | 34.0 (1,102) | 33.5 (1,508) | |
Achieved HbA1c target (< 53 mmol/mol) | | | | < 0.001 |
Yes | 66.4 (841) | 72.4 (2,348) | 70.7 (3,189) | |
No | 33.6 (425) | 27.6 (894) | 29.3 (1,319) | |
Achieved LDL target (< 2.5 mmol/l) | | | | < 0.001 |
Yes | 59.2 (750) | 53.2 (1,724) | 54.9 (2,474) | |
No | 40.8 (516) | 46.8 (1,518) | 45.1 (2,034) | |
Achieved sBP target (< 135 mmHg) | | | | 0.398 |
Yes | 44.7 (243) | 47.0 (377) | 46.1 (620) | |
No | 55.3 (301) | 53.0 (425) | 53.9 (726) | |
Missing
| n = 722 | n = 2,440 | n = 3,162 | |
A total 28.1% of the T2D patients had been prescribed recurrent sleep medication within the study period (Table
1). The most often prescribed medications were benzodiazepine-like medications (56.9%), followed by melatonin (44.4%) and mirtazapine (35.8%). The patients who had been prescribed recurrent sleep medication were slightly older than those who had not (73.4 [SD 11.8] vs. 68.5 [SD11.7], p < 0.001), were more likely to be female (53.9% vs. 42.6%, p < 0.001), and had more likely been diagnosed with depression (20.1% vs. 7.2%, p < 0.001) (Table
1). Only 22.0% of these patients had been diagnosed with a non-apnea sleep disorder.
In terms of subcategories of recurrent sleep medication (Supplement 1), patients who had ongoing sleep medication were more likely to be female (55.8%) and had more likely been diagnosed with sleep apnea (24.8%) and depression (22.7%), compared with the other subcategories and ‘no recurrent sleep medication prescription’ category (p < 0.001 for all). The HbA1c target was most likely to be achieved by the ‘no recurrent sleep medication prescription’ category patients (72.4%). In turn, the LDL target was most likely to be attained by the ‘recurrent sleep medication prescription’ category patients (59.2%) (Table
1), and particularly by patients who had previously been prescribed recurrent sleep medication (63.3%) (Supplement 1). There were no significant differences between the two main sleep medication categories in terms of achieving their sBP target (Table
1), but over half the patients in the ‘subsequent medication’ subcategory had achieved the goal (54.3%), compared less than a half in other subcategories and in the ‘no recurrent sleep medication prescription’ category (Supplement 1).
Discussion
In extensive, real-world data on 4,508 T2D patients, we found that 28.1% had been prescribed recurrent sleep medication during the study period between 2011 and 2019. The most commonly prescribed medications were benzodiazepine-like medication (56.9%), melatonin (44.4%), and mirtazapine (35.8%). Interestingly, a non-apnea sleep disorder was only diagnosed among 22.0% of the T2D patients with recurrent prescriptions. None of the sleep medication (sub-)categories clearly outperformed in all treatment balance measurements.
In our data, nearly one-third of the T2D patients had been prescribed sleep medication at least twice. This finding is particularly alarming as the most often prescribed medications were benzodiazepine-like medications which are related to adverse health effects [
10,
12] and hence, only recommended for short-term use [
18]. Whether sedative sleeping drugs (such as benzodiazepine-like medication) should be entirely avoided, particularly among older patients [
21], which the present study sample also included, is under debate. Unfortunately, it seems that older patients aged 65 or over have long-term prescriptions for sedative sleep medication five to six times more often than younger ones [
22,
23].
As for the present prescription estimates, it is worth noting that the prevalence of recurrent sleep medication users may be even higher than can be indirectly estimated from the records, as the availability of melatonin products is not regulated in Finland and they are also available as self-care drugs. There are no other Finnish studies on recurrent sleep medication prescriptions among T2D patients, but one Scandinavian study has detected a clearly lower prevalence rate of recurrent prescriptions than that of ours when evaluating prescriptions of benzodiazepine-like medication and melatonin in a T2D sample [
9]. This difference may be related to the older study sample (70 years on average vs. median age 63 years) or a higher number of medications being included in the present study (nine vs. two).
Still, the extent to which the treatment of sleep problems relies on medications and on non-pharmacological methods should be explored. Based on comparisons to other reports, we can speculate that medication may play a crucial part in the treatment of sleep problems among primary care T2D patients. According to the current European guidelines [
1], non-medication-based treatment methods, such as cognitive behavioral therapy, should be first-line treatment for sleep problems, and this should also be the case in terms of T2D patients. On the other hand, the prescription prevalence of sleep medication may reflect a substantially high frequency of sleep problems in this population even though the sleep disorder diagnoses themselves were at lower levels; only slightly over one-fourth of patients with recurrent sleep medication prescriptions had been diagnosed with a non-apnea sleep disorder. In this sense, a sleep disorder diagnosis may not represent the magnitude of sleep problems in real life. A similar phenomenon was observed in a previous registry-based study of the Danish population, in which only 1% of all patients with a sleep disorder were identified on the basis of their diagnoses, and 99% on the basis of recurrent sleep medication [
9]. These observations raise questions about whether primary care physicians follow the recommended procedures of the diagnostic management of sleep disorders [
1]. Although a limited number of non-apnea sleep disorders in relation to the prescription prevalence may be related to under-recording, i.e., diagnoses not being appropriately recorded in patient records, it is possible that sleep disorders are not properly diagnosed among T2D patients in the primary care context.
Contrary to expectations, patients with ongoing sleep medication only slightly differed from those without recurrent sleep medication prescriptions in terms of achieving their HbAc1, LDL, and sBP goals; that is, they were less likely to fulfill their HbA1c target, but more likely to attain their LDL target. The discrepancy in the HbA1c goal may be due to the observation that a higher percentage of patients in the ‘ongoing medication’ subcategory than in the ‘no recurrent sleep medication prescription’ category had been prescribed long-acting insulin, which is indicative of worse T2D. As discussed above, systematic reviews and meta-analyses repeatedly demonstrate that sleep medication prescriptions likely reflect sleep problems related to higher glucose levels among T2D patients [
24,
25]. In contrast, reports relating sleep problems to cholesterol are more conflicting [
24,
26,
27], with a recent meta-analysis showing no such association [
24]. A slightly higher prevalence of lipid-lowering prescriptions in the ‘ongoing medication’ subcategory than in the ‘no recurrent sleep medication prescription’ category may indicate that more of the former use lipid-lowering medication, which could at least partly explain the LDL finding. In future, a case-control study in which patients’ age, gender, BMI, and severity of diabetes are controlled for, or a randomized control trial would serve as better grounds for clarifying the role of recurrent ongoing sleep medication in the T2D treatment balance in more detail.
Sleep apnea has been recognized as a highly common comorbidity among T2D patients, with the highest estimates being up to 90% for overall prevalence and 59% for moderate-to-severe sleep apnea [
28,
29]. Therefore, the finding that sleep apnea was only diagnosed among 12.5% of our study population was surprising. On the other hand, this is in line with the literature that shows that sleep apnea is highly underdiagnosed among primary care T2D patients [
28]. In their extensive primary care study, Heffner et al., [
28] found that 18% of T2D patients had a sleep apnea diagnosis, which is close to our estimates, but far from those expected. Primary care physicians should pay more attention to symptoms of sleep apnea and screen them regularly when treating T2D patients [
30], particularly due to the increased morbidity and mortality risks related to these co-existing diseases [
31].
Its substantial registry-based data is the main strength of this T2D study. This is also the first study on recurrent sleep medication prescriptions among the Finnish primary care T2D population. In terms of sex distribution and average age [
32], the study sample was similar to that from which T2D data was collected in several primary care clinics throughout Finland. This increases the generalizability of the current results to Finnish primary care patients with T2D. However, several elements need to be regarded as limitations in this study. First, the design was cross-sectional and thus no cause-and-effect conclusions can be drawn. Secondly, the study sample included mainly older patients, which may limit the generalizability of the current results to working-aged samples. On the other hand, a substantial part of primary care T2D patients are older in Finland because employed individuals are treated by occupational health services. Thus, we believe that the study sample likely represents the Finnish primary care population of T2D patients quite well in this regard. Thirdly, due to the structure of our available registers, we had no data on the reasons for which the medications were prescribed, thus some of them may have been prescribed for reasons other than sleep problems. Moreover, we lacked information on whether the patients had actually taken the drugs, which can be seen as a limitation and restricts conclusions on sleep medication prescriptions only.
Publisher’s Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.