INTRODUCTION

Depression in older people is common and substantially disabling (¹). It is almost invariably found to be associated with worse general health, probably both as a cause and consequence (^1,2). Micronutrient deficiencies (whether relatively low levels or clinical deficiency states) have been found to be associated with depression (^3,4) and are likely to be particularly important in older people because of higher levels of morbidity and frailty. Recent studies (^5,6) have suggested that depression may be associated with vitamin D deficiency. This may reflect influences of depressive states on sunlight exposure, diet, and nutritional supplementation. However, there are also plausible biological pathways for a role of vitamin D deficiency in the pathogenesis of depression, including effects on nerve growth factor synthesis (⁷) and a variety of potential neurotransmitter targets (⁸). Regardless of the direction of causation, the relationship is potentially of high public health importance because of the adverse impact of both states on well-being. In a nationally representative British survey of health in older people, we investigated the association between 25-hydroxy vitamin D (25(OH)D) levels and depressive symptoms, as well as the extent to which these were modified by general health status and (given the known influence of sunlight exposure on 25(OH)D levels) by the time of year at which examinations had been carried out.

METHODS

RESULTS

After sampling, further confirmatory analyses showed that participants for whom vitamin D data had been obtained were representative of those interviewed. For example, mean age was 74.5 years in those interviewed and was 73.7 years in those who gave a blood sample. Furthermore, 44% of those interviewed and 46% of those providing a blood sample were male. Income, region, and social class were also similar between groups.

Characteristics of the analyzed sample are summarized in Table 1. Sixty percent were aged between 65 and 74 years and <10% described poor physical health; however, an appreciable proportion had one or more limiting long-standing illness. Examinations were evenly distributed across seasons. The prevalence of case level depressive symptoms was 25.2% in the total analyzed sample. In participants with 25(OH)D levels of <30 ng/mL (85.4%), prevalence of depressive symptoms was 22.6%; in participants with 25(OH)D levels of <20 ng/mL (51.4%), prevalence of depression was 25.8%; and in participants with 25(OH)D levels of <10 ng/mL (9.8%), prevalence of depression was 35.0%. With respect to this latter deficiency category compared with the remainder of the sample, the prevalence ratio for depressive symptoms was 1.45 and the population attributable fraction calculated from this was 4.2%. The Spearman correlation coefficient between 25(OH)D levels and the number of depressive symptoms (GDS score) was −0.14 (p < .001) with evidence of nonlinearity. With respect to distributions, 25(OH)D levels were normally distributed, whereas GDS score showed pronounced positive skew.

Logistic regression analyses of the associations between the three deficiency states and depressive symptoms are summarized in Table 2. All three associations were significant before adjustment. Associations with the broader two deficiency categories were no longer significant after adjustment, but those with the most severe deficiency syndrome (25(OH)D levels of <10 ng/mL) remained independent. The most marked reductions in the strength of association occurred post adjustment for general health status. Stratification by season of examination (Table 3) did not reveal any marked or consistent variation (p value for interaction term between season and 25(OH)D deficiency 0.84). Finally, the dose-response association between 25(OH)D level and depressive symptoms was investigated in a linear model and found to be strongly significant (B-value, −1.94; 95% CI, −2.67, −1.20). In a secondary analysis, further adjustment for alcohol consumption (amount consumed on heaviest day in the past week) did not change the results meaningfully (data not shown).

DISCUSSION

In an analysis of data from the 2005 HSE, we investigated the association between vitamin D deficiency and depressive symptoms. These were found to be associated but only for conventionally defined clinical vitamin D deficiency states and not for milder levels of deficiency. The association was not accounted for by physical health or other potential confounding factors and was not substantially modified by the time of the year at which the examination was carried out. Particular advantages of this study were the presence of directly measured vitamin D levels, a widely used scale for measuring depressive symptoms in older people, and a large and nationally representative sample.

The importance of vitamin D in calcium absorption and metabolism for bone health is well known, but increasing evidence (^16,17) also suggests that adequate vitamin D status may be protective against noncommunicable diseases like diabetes, cancers, cardiovascular disease, rheumatoid arthritis, and autoimmune conditions. The amount of sun exposure (a primary source of serum vitamin D) necessary to meet requirements depends on factors, such as age, latitude, season, time of day, time of year, clothing, and skin pigmentation (¹⁸). Maximum vitamin D production occurs in summer months, and above latitude 37° North, the sun is not strong enough to provide any vitamin D in winter (for example, at the latitude of London, little vitamin D is generated from sun exposure between the middle of October and the middle of April). In the absence of sunlight, 1000 IU of vitamin D is necessary to maintain a healthy level of 25-hydroxyvitamin D >30 ng/mL (¹⁹). The half-life of vitamin D has been estimated to be of the order of 3 months (²⁰). Older people are at higher risk of poor vitamin D status due to a decline in efficiency of vitamin D synthesis and a lowered renal conversion to its active form (²¹). Low endogenous production can be compensated for by dietary intake and dietary supplement use, but in the United Kingdom, dietary intake of vitamin D-containing foods (oily fish, such as herring and mackerel, fortified margarines, meat and meat products, and eggs) among older people is also poor (²²). Winter vitamin D levels in older people in the United Kingdom have been found to be associated with diet and overseas holidays within the last 6 months but not with winter sun exposure (²³). The importance of vitamin D deficiency and variation in levels by season have been reported previously for the sample analyzed here (²⁴).

Although vitamin D deficiency has been investigated in relationship to mental disorders in younger adults (^25,26), relatively little research has investigated this association in older people, despite the higher potential impact. In a case-control study of older people with or without mild Alzheimer’s disease, lower 25(OH)D levels were associated with mood disorder as well as with cognitive impairment (⁵). The Longitudinal Aging Study Amsterdam also found lower 25(OH)D levels, as well as higher parathyroid hormone levels, to be associated with both major and minor depression (⁶). Our findings are consistent with these results, although increased depressive symptoms were only seen in the most severe deficiency state with no marked increases in association with more mild relative deficiency.

Several processes may underlie the association. Long-standing physical health problems may influence both risk of depression and a person’s access to sun exposure. However, adjustment for this revealed only modest confounding effects. Direction of causation is perhaps the most important consideration and cannot be concluded from a cross-sectional design. In particular, it is possible that depressive states were a cause, rather than a consequence, of vitamin D deficiency although, if this were the case, the association with depression would be expected to be to a similar extent with any relative 25(OH)D deficiency rather than, as observed in this sample, restricted to the 10% lowest levels (i.e., show associations across the range of 25(OH)D levels). Also, if depression caused vitamin D deficiency due to lack of sunlight exposure, then the association between 25(OH)D levels and contemporaneous depressive symptoms would be expected to be stronger in the summer months when levels are most strongly influenced, which was not evident in this sample. However, ultimately prospective research is required to clarify the direction of cause and effect.

Considering methodological issues, an advantage of the study was that the sample was drawn specifically to generalize to the national population. However, it should be borne in mind that participants were all community residents who represent a relatively healthier subpopulation of older adults because those requiring institutional care were not included. However, we do not believe that this compromises the principal findings for this subpopulation—it simply restricts the generalizability to more dependent groups. Conventional confounding factors were taken into account in this analysis. Other unmeasured potential confounding factors might include general life-style (for example, the ability/willingness to travel to overseas destinations with higher sunlight exposure) and cognitive impairment (for example, causing both depressive symptoms and vitamin D deficiency), neither of which were measured in this wave of the study. However, were data to be present for cognitive function, its role would probably remain difficult to determine, as cognitive impairment could also be a consequence of both vitamin D deficiency and depression. Dietary intake of vitamin D was also not measured, and it was not possible to determine vitamin D supplementation (because of varying doses in multivitamin preparations) but these are likely to be important, particularly in people with less than adequate sun exposure/absorption. However, the role of diet, if causal, is potentially complex because of the difficulty distinguishing micronutrients of interest. For example, oily fish are key sources of vitamin D but are also of interest in depression as a source of essential fatty acids (²⁷). The role of somatic comorbidity is another salient consideration. Long-standing illness, which was present in a large proportion and which was potentially associated with both depression and (through reduced sunlight exposure) vitamin D deficiency, did not seem to have a substantial influence on the association of interest. A limitation was that comorbidity relied on self-report information which, in addition to measurement inaccuracy, may be influenced by depression status. Its inclusion in regression models could represent overadjustment which was why it (as well as BMI for the same reasons) was adjusted for as a separate procedure. Finally, the nonfasting nature of the blood sample will not have influenced vitamin D levels meaningfully, as sunlight exposure is a much more important determinant (²⁸).

If vitamin D deficiency is demonstrated to be a cause of depression, the calculated population attributable fraction suggests that removing vitamin D deficiency could be an effective public health measure to reduce depression prevalence in later life. As summarized earlier, it is becoming increasingly recognized that this exposure has wider adverse effects than those on bone structure and skeletal integrity. Previous trial findings have suggested positive effects of vitamin D supplementation on seasonal affective disorder and mood (^29,30) and, as stated earlier, there are plausible biological pathways that might account for a causal association, including involvement in nerve growth factor synthesis (⁷) and a variety of potential neurotransmitter targets (⁸). Hoogendijk and colleagues (⁶) investigated a mediating role of hyperparathyroidism, but their evidence did not support this.

Regardless of the direction of causation, the higher than expected co-occurrence of vitamin D deficiency and depression is an important public health issue for older populations in northern latitudes because both are common, both have substantial adverse health consequences, and both are potentially reversible.