Background: Hypovitaminosis D has been cross-sectionally associated with dementia and stroke. The objective of this longitudinal study was to determine whether serum vitamin D deficiency at baseline could predict the onset of non-Alzheimer dementias (NAD) within 7 years among older women. Methods: Forty high-functioning older women (78.4 years, 76.4/82.0; median, 25th/75th percentile) from the EPIDOS Toulouse study were divided into two groups based on vitamin D deficiency (i.e., serum 25-hydroxyvitamin D <10 ng/ml) at baseline. At the end of the 7-year follow-up period, women matching the DSM-IV but not the NINCDS-ADRDA criteria were diagnosed with NAD while those matching the NINCDS-ADRDA criteria were considered to have Alzheimer’s disease (AD). Subtle cognitive impairments at baseline, cardiovascular risk factors and Parkinson’s disease were used as potential confounders. Results: NAD was reported in 6 women (82.8 years, 80.6/86.0) after 7 years of follow-up. More NAD were observed in women with vitamin D deficiency (p = 0.023). There was no between-group difference regarding the onset of AD (p = 0.332). We found an association between vitamin D deficiency at baseline and the onset of NAD (adjusted odds ratio = 19.57, p = 0.042). Conversely, vitamin D deficiency was not associated with AD (p = 0.222). Conclusion: Baseline vitamin D deficiency predicted the onset of NAD within 7 years among older women.
Low serum 25-hydroxyvitamin D (25OHD) concentrations are highly prevalent among older adults with prevalence reaching 90% , and have been associated with deterioration of global cognitive performance [2,3,4,5,6,7]. For instance, a significant association between severe 25OHD deficiency <10 ng/ml and advanced-stage dementia was found among 288 elderly inpatients aged 86 years on average, regardless of the type of dementia . In line with this result, a 2.2-fold increased risk of dementia was also shown among community-dwellers with hypovitaminosis D . More precisely, the authors found that hypovitaminosis D was associated with ischemic infarcts and white matter hyperintensities, suggesting a plausible link between hypovitaminosis D and the vascular component of dementia . Unfortunately, these two previous studies were limited by a cross-sectional design that prevented the finding of a causal relationship. It thus remains unclear whether dementia precipitates vitamin D deficiency – due to low dietary vitamin D intakes or a lack of sunlight exposure for example – or whether vitamin D deficiency plays a role in the onset of dementia. The latter assumption seems likely given that vitamin D deficiency was recently associated with faster cognitive decline compared to normal vitamin D status . Anyway, the type of dementia specifically associated with vitamin D deficiency (i.e., Alzheimer’s disease, AD, or non-Alzheimer dementias, NAD) is not conclusively clarified. Since vitamin D has exhibited protective mechanisms not only against neuronal degeneration in experimentation, but also against vascular damage [8,9,10], we hypothesized that vitamin D deficiency could result in NAD among older individuals. The objective of this prospective longitudinal study was to determine whether vitamin D deficiency (i.e., serum 25OHD <10 ng/ml) measured at baseline could predict the onset of NAD among community-dwelling older women participating in a large cohort study called the EPIDémiologie de l’OStéoporose (EPIDOS) study.
Materials and Methods
EPIDOS Study. The EPIDOS Study was a large French observational prospective multicenter cohort study designed to evaluate during a 4-year follow-up the risk factors for hip fractures among high-functioning women aged 75 years and older. The sampling and data collection procedures have been described elsewhere in detail . In summary, from 1992 to 1994, 7,598 subjects were recruited in five French cities from electoral lists (Amiens, Lyon, Montpellier, Paris and Toulouse). Exclusion criteria included inability to walk independently, institutionalization, previous history of hip fracture or bilateral hip replacement and inability to understand or answer the study questionnaires. All included study participants had at the same time a blood test and a full medical examination by trained nurses in each local clinical center, which consisted of structured questionnaires, information about health and chronic diseases and a clinical examination. Sera were stored at –100°C until analyses were performed. A randomized sample of 752 women was then drawn, including 90 subjects in the center of Toulouse. This choice of 752 women was based on our budgetary capabilities to perform the laboratory measure of serum 25OHD concentration. The randomization process was based on the use of a random number table which generated in an unpredictable, haphazard manner a sequence of number corresponding to the number of subjects included in the study .
Toulouse Cohort Study. At the end of the 4-year EPIDOS study, all participants included in the center of Toulouse were invited to take part in an additional 3-year follow-up study. The data collection procedures and flow diagram have been described elsewhere in detail . In summary, those who had given informed consent were assessed either at home or at the Department of Internal Medicine and Clinical Geriatrics of Toulouse University Hospital. Initially, 1,462 women were included in the EPIDOS study in Toulouse. At the end of the 7-year follow-up, data on cognitive status (i.e., no dementia, or AD, or NAD) were available for 714 women (48.8% of the initial cohort) including 40 women with a measure of serum 25OHD concentration at baseline. Of the other 748 women whose cognitive status remained undetermined, 193 (25.8%) died during follow-up, 414 (54.7%) were lost to follow-up, and 141 (18.9%) withdrew from the follow-up study .
Dependent Variable: Cognitive Status after 7 Years of Follow-Up
During the final visit, cognitive function was tested independently by a geriatrician and a neurologist from the Department of Internal Medicine and Clinical Geriatrics of Toulouse University Hospital, France, who had expertise in dementia and were unaware of medication taken. DSM-IV criteria were used to establish a clinical diagnosis of dementia  in the absence of delirium and regardless of the length and stage of dementia . AD was diagnosed according to the criteria of the NINCDS-ADRDA work group [16,17]. Subjects matching the DSM-IV but not the NINCDS-ADRDA criteria were considered as NAD. When the two assessors disagreed on a diagnosis, a joint decision was reached after additional information was obtained from the general practitioner. Subjects were then classified into three groups: nondemented subjects, subjects with AD, and subjects with NAD.
Explanatory Variable: Serum Vitamin D Deficiency
Venous blood was collected from resting subjects for the measurement of serum 25OHD. Serum concentrations of 25OHD were measured by radioimmunoassay (Incstar Corp., Stillwater, Minn., USA). With this method there is no lipid interference, which is often observed in other nonchromatographic assays of 25OHD. Intra- and interassay precisions were 5.2 and 11.3%, respectively (range 12–50 ng/ml in normal adults aged 20–60 years). Vitamin D deficiency was strictly defined for serum 25OHD concentrations lower than 10 ng/ml . All measurements were performed at Lyon University Hospital.
The sample size of 40 subjects assumed accounting for up to four covariables in multivariate analyses, including serum vitamin D deficiency at baseline. The following three potential confounders were included in our data analysis: subtle cognitive impairments at baseline, presence of cardiovascular risk factors at baseline, and diagnosis of Parkinson’s disease (PD) at baseline .
No diagnosis of dementia was made at baseline according to DSM-IV criteria . Initial cognitive functioning was assessed as a whole among this high-functioning cohort with the use of the Pfeiffer’s Short Portable Mental Status Questionnaire (SPMSQ) [19,20], a reliable standardized validated screening test of brain function for community-dwellers . It consists of a 10-item composite questionnaire with a score ranging from 0 to 10, with 10 representing the best cognitive performance. The validated cutoff value for normal cognitive functioning is a score of 8 or above . In the present study, subtle cognitive impairments at baseline assessment were thus defined as a SPMSQ score below 8.
Cardiovascular risk factors (i.e., older age, overweight, hypertension, diabetes mellitus, smoking and lack of regular physical activity) have also been sought [21,22]. Older age was defined as beyond 85 in this elderly population . Hypertension, diabetes mellitus, and smoking were reported by direct inquiry. Women were also asked whether they participated in a sport or physical leisure activity regularly (i.e., at least 1 h a week for at least the past month) . Overweight was defined as a body mass index above 25 according to the National Heart, Lung, and Blood Institute .
Information about PD was obtained from the standardized question: ‘Do you currently suffer from Parkinson disease?’ . In addition, women were asked to bring all the medication they were regularly taking to the clinical center, and those usually taking antiparkinsonian medications were also considered as having PD.
The subjects’ characteristics were summarized using medians (25th/75th percentile) or frequencies and percentages, as appropriate. Firstly, comparisons between subjects separated into two groups based on serum 25OHD concentration (<10 ng/ml or ≥10 ng/ml) were performed using the Mann-Whitney U test or the χ2 test, as appropriate. Secondly, univariate and multiple (i.e., fully adjusted model, and stepwise backward model) logistic regressions were used to examine the association between vitamin D deficiency (independent variable) and the onset of NAD (dependent variable), while adjusting for potential confounders; p values less than 0.05 were considered significant. All statistics were performed using SPSS (version 15.0; SPSS, Inc., Chicago, Ill., USA).
Among 40 women included in the analysis (78.4 years, 76.4/82.0), the median serum 25OHD concentration at baseline was 14.5 ng/ml (11.0/18.0; range: 6.7–47.0).
As indicated in table 1, 30 women (75.0%; 78.2 years, 76.3/81.1) were nondemented at the end of the follow-up, 6 (15.0%; 82.8 years, 80.6/86.0) presented with NAD, and 4 (10.0%; 77.14 years, 75.8/82.1) with AD. Seven women (81.59 years, 76.7/83.7) had vitamin D deficiency defined as 25OHD <10 ng/ml. More NAD were observed among women with vitamin D deficiency at baseline (p = 0.023) (table 1). There was no significant difference for the other clinical characteristics (table 1).
Table 2 shows multiple logistic regressions between NAD, serum vitamin D deficiency and baseline characteristics as explanatory variables with odds ratios (OR) and confidence intervals for each unadjusted, fully adjusted and stepwise backward model. Serum 25OHD deficiency was associated with the onset of NAD (unadjusted OR = 7.50 with p = 0.039; adjusted OR = 19.57 with p = 0.042 for fully adjusted model) and was retained in the stepwise backward model (adjusted OR = 14.95 with p = 0.037). In contrast, serum vitamin D was not associated with the onset of AD (unadjusted OR = 1.06, 95% CI: 0.97; 1.15, p = 0.222).
The main finding of this prospectively followed cohort study of 40 high-functioning older women is that vitamin D deficiency at baseline was associated with the onset of NAD within 7 years.
The term NAD refers particularly to subcortical degenerative dementias – including PD – and to multi-infarct and mixed dementias . In this sense, our results are coherent with previous experimental and cross-sectional cohort studies that highlighted an association of hypovitaminosis D with PD [26,27,28,29] and stroke [7,10]. For instance, vitamin D deficiency was observed more often among PD patients compared to controls whatever their functional status [26,27]. Parallel, chronic hypovitaminosis D may precipitate the occurrence of cardiovascular risk factors that are major determinants of cerebrovascular diseases, including atherosclerosis [30,31] or hypertension [32,33]. In line with this, Buell et al.  recently described among 318 community-dwelling older subjects aged 74 years on average a cross-sectional association between hypovitaminosis D and stroke (OR = 2.29, p = 0.03) and white matter hyperintensities (higher volume: 4.9 vs. 2.9 ml, p = 0.004; and higher grade: 3.2 vs. 2.2, p = 0.02). In contrast, vitamin D repletion may play a neuroprotective role in cerebral ischemia based on trophic induction , as vitamin D3 attenuated cortical infarction induced by middle cerebral arterial ligation in rats .
We found no association of baseline vitamin D concentrations with the onset of AD, although a similar association has already been found in a cross-sectional investigation  and despite the recently proposed involvement of vitamin D in the metabolism of amyloid beta 42 peptide [35,36]. This nonsignificant association may be explained by the lack of power of our study due to the limited number of participants.
In addition to this lack of power with increased beta risk, the main limitations of our study were the use of a nonspecific diagnosis of NAD that encompassed different dementias, and its observational nature that prevented any conclusion on causality [37,38]. Its main strength was the use of a longitudinal design that confirmed the temporal sequence and allowed to assert that baseline hypovitaminosis D predicts the onset of NAD within 7 years. Further studies with greater power are needed to clarify which type of NAD (vascular, degenerative or mixed) occurs in the case of hypovitaminosis D in older adults.
The EPIDOS study participants included coordinators (G. Breart, P. Dargent-Molina, P.J. Meunier, A.M. Schott, D. Hans, and P.D. Delmas) and principal investigators (C. Baudoin and J.L. Sebert, Amiens, M.C. Chapuy and A.M. Schott, Lyon, F. Favier and C. Marcelli, Montpellier, C.J. Menkes, C. Cormier, and E. Hausherr, Paris, and H. Grandjean and C. Ribot, Toulouse). The authors have listed everyone who contributed significantly to the work.
The study was financially supported by the French Ministry of Health. The sponsor had no role in the design and conduct of the study, in the collection, management, analysis, and interpretation of the data, or in the preparation, review, or approval of the article.
Dr. Annweiler serves as an unpaid consultant for Ipsen Pharma Co. He has no relevant financial interest in this article.
Prof. Rolland serves on a board for Cheisi, serves as a consultant for Lilly, Nutricia and Lundbeck, received grants from IDEM and Servier, and received payment for the development of educational presentations from Nutricia. He has no relevant financial interest in this article.
Prof. Schott reports no conflicts of interest. She has no relevant financial interest in this article.
Prof. Blain reports no conflicts of interest. He has no relevant financial interest in this article.
Prof. Vellas serves on the board for the Fondation Plan Alzheimer, serves as a consultant for Astra-Zeneca, Danone/Nutricia, Eisai, Eli Lilly Co., Exhonit, GlaxoSmithKine, Ipsen, Pfizer Inc., Roche and TauRx, and received grants from Avid Radiopharmaceuticals, Bristol Myers Squibb, Danone, Eisai Inc., Eli Lilly Co., Exhonit, GlaxoSmithKline, Ipsen, Lundbeck, Médivation, Merck, Nestlé, Pfizer, Pierre Fabre Laboratories, Roche, Sanofi-Aventis and Servier. He has no relevant financial interest in this article.
Prof. Beauchet serves as an unpaid consultant for Ipsen Pharma Co., and as an unpaid editor for Gériatrie, Psychologie et Neuropsychiatrie du Vieillissement. He has no relevant financial interest in this article.