The effect of depressive symptomatology on the association of vitamin D and sleep

Vitamin D is a steroid hormone playing an important role in both calcium-phosphorus and bone homeostasis [1]. Moreover, vitamin D has strong effects on brain development and adult brain function and there are close links between low levels of vitamin D and neuropsychiatric disease [2]. According to a recent review [3], vitamin D deficiency is associated with many acute and chronic diseases including periodontitis, autoimmune disorders, infectious and cardiovascular diseases, cancer, type 2 diabetes and neurological diseases. The discovery of vitamin D receptors in different cerebral areas (e.g., central gray matter in the midbrain, substantia nigra, nucleus tractus optici, nucleus interstitialis striae terminalis) suggested associations between vitamin D concentrations and mental health. High prevalence of vitamin D deficiency and insufficiency has been found in subjects with mental disorders [4‐6]. Furthermore, low levels of 25-hydroxyvitamin D (25(OH) D) in patients suffering from depressive disorders were well documented (for reviews see [7, 8]) and could be replicated in several more recent studies (e.g., [9‐11]). Moreover, Zhu et al. [12] found a significant inverse association of vitamin D concentrations and the severity of depressive symptoms in patients with major depressive disorder. However, there are less intervention studies examining the preventive effects of vitamin D supplementation [13]. A moderate intake of fruits/vegetables and calcium/high vitamin D sources has been shown to predict lower likelihood of depression in men [14], and subjects reporting supplementation with vitamin D had a reduced depression risk in a cross-sectional study [15]. However, in a prospective study with chronic dialysis patients with depression, calcitriol supplementation did not enhance depressive symptoms [16] and a two-sample Mendelian randomization study [17] also revealed no causal effects of serum 25(OH) D concentration on depressive symptoms or broad depression.

The influence of vitamin D deficiency on sleep has also been investigated. Sleep problems were associated with low vitamin D levels [18‐20] and vitamin D supplementation has been shown to improve sleep disorders [21]. However, only few studies examined the association of vitamin D concentrations and objectively measured sleep via actigraphy so far [22‐24]. Although a higher probability for short sleep (< 5 h) and low sleep efficiency (< 70%) were found to be associated with lower vitamin D concentrations [22], Bertisch and collegues could not demonstrate this association in adjusted models [23] including age, race/ethnicity, sex, and various other covariates. However, these findings were only present in older populations [22, 23]; moreover in the study of Massa and colleagues [22] the sample comprises only male participants. Using data from a large German population-based cohort study (LIFE-Adult-Study, [25]) our group could demonstrate that higher vitamin D concentrations in the general population are significantly associated with longer and earlier night sleep as assessed by actigraphic measures – even if the life-time clinical diagnosis of depressive disorders had been considered as covariate [24]. These findings were mainly in line with the previous studies examing the association of vitamin D levels and objective sleep parameters. Contrary to previous findings, we found no significant association between sleep problems assessed by PSQI (Pittsburgh Sleep Quality Index) and ESS (Epworth Sleepiness Scale).

Despite of body of evidence showing complex relationship between sleep and vitamin D the trilateral relations between serum 25(OH) D concentrations, night sleep and current depressive symptomatology remained unclear. Depressive symptomatology could be a possible mediator or moderator for the association of vitamin D and night sleep parameters. First, night sleep disturbances are part of the depressive symptomatology. Second, there are findings suggesting vitamin D deficiency in patients suffering from depression [7] and a significant negative association of vitamin D concentrations and depressive symptomatology [12]. These findings are related to patients suffering from depressive disorders; however, there is a lack of corresponding results for the total spectrum of depressive symptomatology in the general population.

The present re-analysis of data used in our previous publication [24] aimed to examine the associations between vitamin D concentrations, night-sleep parameters and depressive symptomatology in an exploratory way using a largely identical, but slightly expanded sample. It will be explored whether the association of vitamin D concentrations and night sleep parameters (night sleep duration and midsleep time) is mediated or moderated by depressive symptomatology. In view of of the existing literature it was assumed that the direction of the association goes from vitamin D concentrations over depressive symptomatology to night sleep.

The characteristics of the final sample are shown in Table 1. The mean serum 25(OH) D concentration in the sample (23.44 ng/ml; SD: 11.15 ng/ml) reflects vitamin D insufficiency [45]. The mean night sleep duration was rather low (06:16 hh:mm; SD: 00:59 hh:mm). According to the original CES-D sum scores, depressive symptomatology was present in 16.9% of the final sample.

Table 2 summarized the trilateral associations between serum 25 (OH) D concentrations, CES-D sum scores (after removing item 11) and night sleep parameters (night sleep duration, midsleep time). Only one association was found to be statistically significant: the positive correlation of 25(OH) D concentrations with night sleep duration (rho = 0.08; p = 0.006).

As shown in Fig. 1a, the direct effect of serum 25(OH) D concentrations on night sleep duration in the simple mediation model without covariates (c = 0.007; p = 0.009) was significant. The same was true for the effect of serum 25(OH) D concentrations on CES-D sum scores in this model (a = − 0.036; p = 0.021) whereas the effect of the CES-D sum scores on night sleep duration failed to be significant (b = − 0.0001; p = 0.978). The indirect effect of serum 25(OH) D concentrations on night sleep duration via depressive symptomatology was non-significant, too (ab = 0.000; 95% CI [− 0.0004, 0.0003]). In the corresponding mediation model with covariates (age, the BMI score and the season in which vitamin D concentrations had been measured) all afore-mentioned effects failed to be statistically significant (see Fig. 1b).

Figure 2a demonstrates that the direct effect of serum 25(OH) D concentrations on midsleep time in the simple mediation model without covariates was negative, but non-significant (c = − 0.001; p = 0.641). Instead, there was a significant negative effect of serum 25(OH) D concentrations on CES-D sum scores in this model (a = − 0.0363; p = 0.021) whereas the effect of the CES-D sum scores on midsleep time failed to be significant (b = 0.004; p = 0.325). The indirect effect of serum 25(OH) D concentrations on midsleep time via CES-D sum scores was non-significant, too (ab = − 0.0001; 95% CI [− 0.0005, 0.0002]). In the corresponding mediation model with covariates (age, the BMI score and the season in which vitamin D concentrations had been measured) all above-mentioned effects were non-significant (see Fig. 2b).

According to moderator analyses, the effects of the interaction of the factors “vitamin D concentration” and “CES-D sum scores without item 11 (sleep problems)” on night sleep duration failed to be significant. The same was true for corresponding regression models with midsleep time as dependent variable (see Table S1 in Supplement S2).

The main aim of this study was to examine the trilateral associations between serum vitamin D concentrations, current depressive symptomatology and night sleep duration in the general population of a German city as a follow-up study to our previous findings [24]. In addition, it was investigated whether vitamin D concentrations had significant effects on night sleep duration and midsleep time (as suggested by [24]) through the mediation of depressive symptomatology and whether it represents a relevant moderator variable for this association. According to the main results of this study, serum 25(OH) D concentrations did not have an indirect effect on night sleep duration and midsleep time through a mediation effect of depressive symptomatology: There was a significantly negative association of vitamin D concentrations with depressive symptomatology in the simple mediation models as expected due to previous findings suggesting a close link between depression and low levels of serum 25(OH) D [7‐12]. However, the sum score in the CES-D (after removing the item concerning sleep problems) was neither significantly associated with night sleep duration nor midsleep time which would have been required for a mediator between vitamin D concentrations and night sleep.

In the simple mediation model (without any covariates), serum 25(OH) D concentrations had a direct and significantly positive effect on night sleep duration in a German community sample. This finding was in line with previous studies, in which sleep parameters had been measured in an objective way by using actigraphy and a positive association between serum 25(OH) D concentrations and night sleep duration had been found [22‐24]). However, this association was found to be not mediated by depressive symptomatology and failed to be significant if the effects of relevant covariates (age, the BMI score and the season in which the vitamin D concentrations had been measured) were controlled for. Thus, a specific effect of vitamin D concentrations on night sleep duration seems to be quite improbable.

Another picture is present for the trilateral associations between serum 25(OH) D concentrations, current depressive symptomatology and midsleep time: Midsleep time was neither significantly associated with serum 25(OH) D concentrations nor with the sum score in the CES-D scale. In line with these findings, vitamin D concentrations had neither significant direct nor indirect effects on midsleep time in our sample. This result was unexpected since in another study using a very similar sample [24] an inverse significant association between midsleep time and serum 25(OH) D concentrations had been reported. In our sample, this association was also negative, but failed to be significant. The discrepancies might be due to different statistical approaches (multiple linear regression analyses with centered vitamin D concentrations [24] versus mediation and correlation analyses in the present study). Differences between this study and our previous publication [24] regarding mean vitamin D concentrations (23.4 ng/ml versus 21.9 ng/ml), mean night sleep duration (376 min versus 379 min), mean midsleep time (03:10 (hh:mm) versus 03:08 (hh:mm)) and the proportion of patients with a lifetime history of depressive disorders (10.5% versus 9.7%) cannot sufficiently explain these differences.

Further analyses revealed that the association between vitamin D concentrations on the one hand and night sleep parameters (night sleep duration and midsleep time) on the other hand was not moderated by the severity of depressive symptoms (without sleep problems).

Overall, the literature about sleep problems, depression and vitamin D is scarce with large methodological heterogeneity. Sleep problems are in most cases measured only subjectively via questionnaires, thus limiting the comparability wth our results, since objective sleep parameters were used for this mediation analysis. In a case series with depressed adolescents, an improvement in well-being and other depressive symptoms, including sleep difficulties, was shown after vitamin D supplementation [46]. In a recent meta-analysis [47], it has been demonstrated that vitamin D supplementation decreased depressive symptoms as well as sleep problems among patients with psychiatric diseases. Furthermore, Fallah and colleagues showed in a systematic review, that sleep quality and postpartum depression are associated to vitamin D deficiency [48]. In our study using a community sample we failed to show a significant association between night sleep and midsleep time and vitamin D in the mediation models with covariates. Furthermore, no association between depressive symptoms and night sleep and midsleep time was found. This is in line with a study where no effect of daily vitamin D supplementation for a year on depressive symptoms and sleep quality was found among overweight and obese postmenopausal women [49].

In this study, we hypothesized that depression might have a mediator effect on the relationship of night sleep and vitamin D levels according to the previous findings reporting vitamin D deficiency as a risk factor for depression and sleep disorders [50‐52]. However, a reverse association is also conceivable. It could be assumed that a sunlight-avoiding lifestyle as well as lack of physical activity caused by a depressive disorder can play a role in both vitamin D deficiency and sleep disorders. However, since vitamin D deficiency is also associated with low physical performance [45] and fatigue in several patient groups (e.g. [53]), an only monocausal association cannot be drawn based on existing literature.

The present study has some methodological limitations: First, the data were collected within a cross-sectional cohort study; thus, the temporal associations between vitamin D concentrations, depressive symptomatology and night sleep parameters could not be examined in an adequate manner. Second, although the mediation analyses considered relevant covariates (age, the BMI score and the season in which the vitamin D concentrations had been measured) we cannot exclude the possibility that other potential covariates had a significant effect on the night sleep outcomes (night sleep duration and midsleep time), too. Third, only 12.5% of the total sample of the Life-Adult-Study could be included in the present study. The reason for this was that actigraphy was only an optional additional examination, so that in total both vitamin D level and actigraphy data were only available for about ¼ of the 10,000 LIFE participants. Although the examination was offered in principle, selection effects cannot be excluded, e.g., subjects who had experienced sleep disturbances might have been more approachable, as might subjects who were more physically active and interested in quantifying their physical activity. To make matters worse, after all exclusion criteria were taken into account, only just under half of the available data sets could actually be included in the analysis. Again, bias cannot be ruled out; however, the selection criteria were defined with the intention of excluding disease- or medication-related changes in sleep. Fourth, it must be mentioned that lack of statistical significance regarding an association is not a proof of absence of that association because the negative finding could have been due to a type II statistical error. Fifth, the CES-D sum score might have shortcomings as a measure of depression. It would have been interesting to use a more comprehensive inventory of depressive symptoms. Major strengths of this study include a large sample size with the study population being not restricted to a certain age group or gender. Thus, the external validity of the findings was quite high. The interval between the measurement of vitamin D levels and the assessment of night sleep parameters was rather short (maximally 15 days). For the measurement of vitamin D concentrations and sleep parameters, high-quality methods had been used.

In conclusion, the associations between vitamin D concentrations and night sleep duration as well as midsleep time seem to be neither mediated nor moderated by depressive symptomatology. It will be important to confirm this finding in the context of a large representative longitudinal cohort study. Such a study would be of special interest because the temporal associations between serum vitamin D concentrations, the intensity of depressive symptoms and night sleep parameters could be investigated within this study in an adequate way.