Low 25-Hydroxyvitamin D Levels and Cognitive Impairment in Hemodialysis Patients

Summary

Background and objectives

25-hydroxyvitamin D (25[OH]D) deficiency and cognitive impairment are both prevalent in hemodialysis patients in the United States. This study tested the hypothesis that 25(OH)D deficiency may be associated with cognitive impairment because of its vasculoprotective, neuroprotective, and immune-modulatory properties.

Design, setting, participants, & measurements

This cross-sectional analysis involved 255 patients enrolled in the Dialysis and Cognition Study between 2004 and 2012. In linear regression models, 25(OH)D was the exposure variable; it was used first as a continuous variable and then stratified as deficient (<12 ng/ml), insufficient (12 to <20 ng/ml), and sufficient (≥20 ng/ml). Principal component analysis was used to obtain the memory and the executive function domains from the individual neurocognitive tests. Scores on individual tests as well as on the memory and executive function domains were the outcome variables. Multivariable models were adjusted for age, sex, race, education, and other potential confounding variables.

Results

Mean serum 25(OH)D ± SD was 17.2±7.4 ng/ml, with 14%, 55%, and 31% of patients in the deficient, insufficient, and sufficient groups, respectively. Patients in the deficient group were more likely to be women, African American, and diabetic and to have longer dialysis vintage. Higher 25(OH)D levels were independently associated with better performance on several tests of executive function (mean difference on component executive score, 0.16 [95% confidence interval, 0.04–0.28; P=0.01] for each SD higher 25[OH]D). No association was seen with tests assessing memory.

Conclusions

25(OH)D deficiency in hemodialysis patients is associated with worse cognitive function, particularly in domains that assess executive function.

Introduction

There is an increasing awareness of the growing burden of cognitive impairment in the aging ESRD population. Dialysis patients are more likely to be cognitively impaired than age-matched controls; as many as two thirds of patients treated with maintenance hemodialysis have some degree of cognitive impairment (1). Cognitive impairment leads to increased morbidity and health care expenditures and may limit dialysis patients’ ability to comprehend complex medical information and make informed decisions about their health (2). In the general population, cerebrovascular disease and Alzheimer’s disease are the most common causes of cognitive impairment, with these conditions not infrequently occurring in conjunction. However, in hemodialysis patients, cerebrovascular disease is likely to play a more important role given the high prevalence of cardiovascular disease at all stages of CKD (3,4). Several traditional vascular risk factors, including older age, hypertension, and dyslipidemia, as well as concurrent coronary or peripheral vascular disease, are associated with worse cognitive performance in dialysis patients (1,3,5). However, few studies have examined the relationship between nontraditional cardiovascular disease risk factors and cognitive function in this population.

Insufficiency or deficiency of 25-hydroxyvitamin D (25[OH]D) is a potential nontraditional risk factor (6–8), and several pathophysiologic reasons may explain why 25(OH)D deficiency may lead to cognitive impairment. 25(OH)D is necessary for production of activated vitamin D (1,25[OH₂]D), which, in turn, has antioxidant, neuroprotective and vasculoprotective properties (8–10). 25(OH)D deficiency is present in a considerable proportion of patients undergoing hemodialysis (11,12), and at least in the general population it is associated with vascular disease (6,7), cognitive impairment, and an increased prevalence of white-matter hyper intensities and large-vessel infarcts on brain magnetic resonance imaging (7). However, few studies have evaluated the relationship of 25(OH)D levels to cognitive function in hemodialysis patients.

Therefore, in this study we evaluated the association of 25(OH)D with detailed measures of cognitive function in patients undergoing maintenance hemodialysis. We hypothesized that low levels of 25(OH)D may be associated with worse cognitive performance, particularly in executive function, because this cognitive domain primarily reflects vascular disease.

Materials and Methods

Participants

Individuals receiving long-term in-center hemodialysis at five Dialysis Clinic Inc. (DCI) units and one hospital-based unit (St. Elizabeth’s Medical Center) in the greater Boston area were screened for the Cognition and Dialysis Study between 2004 and 2012. Eligibility criteria included English fluency, as well as sufficient visual and hearing acuity to complete cognitive testing. Individuals with scores of ≤10 on the Mini-Mental State Examination (MMSE) (13) or those with advanced dementia according to medical record review were excluded. Other temporary exclusion criteria were non–access-related hospitalization within 1 month of enrollment date, receipt of hemodialysis for <1 month, and single pool Kt/V < 1.0. For the current analyses, we also excluded individuals being treated with oral 25(OH)D replacement (n=19). Informed consent was obtained from all the patients and the Tufts Medical Center Institutional Review Board approved the study.

Cognitive Assessment

Participants were administered a detailed neurocognitive battery during a dialysis session by research assistants trained by a neuropsychologist (T.S.). All testing was completed within the first hour of hemodialysis treatment to reduce the effects of fatigue. Reassessment of research assistants by the study neuropsychologist with mock testing sessions or witnessed testing of study participants occurred at 3- to 6-month intervals to assure quality. The neuropsychological battery included well validated and commonly used cognitive tests that possess high inter- and intrarater reliability; many have established age-, sex-, and education-matched normative scores. Tests included the MMSE, the Wechsler Memory Scale-III (WMS-III) Word List Learning Subtest (14), the Wechsler Adult Intelligence Scale-III (WAIS-III) Block Design and Digit Symbol-Coding subtests (14), and Trail Making Tests A and B (15). During the last 2 years of the study, the cognitive panel was expanded to include additional verbal tests assessing memory and executive functions, including Digit Span (forward and backward) (16), the Mental Alternation Test (17), and the Controlled Oral Word Association Test (COWAT) (18). The overall battery assesses a broad range of cognitive functioning, including global ability, verbal intelligence, supraspan learning, auditory retention, visual retention, attention/mental processing speed, visual construction/fluid reasoning, and motor speed (Supplemental Table 1).

Exposure Variable

25(OH)D levels were measured from samples collected at the time of cognitive testing and stored at −80°C. Samples were assayed using a direct enzyme immunoassay method (19). This assay has a correlation of 0.9 with the gold standard Immuno Radio Assay (IRA), with intraassay coefficient of variation <8% and an interassay coefficient of variation of <10%.

Covariates

Demographic and clinical variables were obtained from patient interview and review of medical records at the time of cognitive testing. Level of education was obtained via patient questionnaire. Medical history, including coronary artery disease, peripheral vascular disease, stroke, heart failure, and presence of diabetes, was obtained from medical records and patient report. Coronary artery disease was defined as history of myocardial infarction and/or revascularization. Cause of ESRD, hemodialysis vintage, mean monthly systolic and diastolic BP, body mass index (BMI), and routine predialysis blood tests (including serum calcium, phosphorus, hematocrit, albumin, and single pool Kt/V) were obtained from the electronic medical records at DCI and St. Elizabeth’s hospitals. Activated vitamin D use was also ascertained from medical records and was defined as use of any activated vitamin D product within 2 weeks of cognitive testing. All DCI laboratories were measured in a central laboratory in Nashville, Tennessee.

Statistical Analyses

After excluding 19 individuals taking 25(OH)D, participants were divided into three 25(OH)D groups on the basis of the recent Institute of Medicine guidelines: deficient (<12 ng/ml), insufficient (12 to <20 ng/ml), and sufficient (≥20 ng/ml) (20). Continuous variables are reported as mean ± SD, and categorical variables are reported as proportions. Where necessary, comparisons between groups were made using ANOVA, Kruskal-Wallis test, and chi-squared or Fisher exact test, as appropriate. Multiple linear regression was used to explore the association between 25(OH)D level and performance on individual cognitive tests, with the raw test scores serving as the outcome. We initially modeled 25(OH)D as a continuous variable, with parameter estimates (mean difference) calculated per 1-SD increase. To explore nonlinear relationships, we also modeled 25(OH)D using the three groups defined earlier. A higher score on the tests indicated better cognitive function, except for the Trail Making Tests A and B, for which higher scores indicate worse performance. Analyses for Trail Making Test B used Tobit regression censoring for failure to complete the task within 5 minutes (21).

No one test can assess all domains of cognition; therefore, we developed a composite score from the individual test scores representing the memory and the executive function domains of cognition to account for problems of co-linearity of the test results and multiple hypothesis testing. We used principal component analysis (PCA) with varimax rotation for data reduction (22), a technique that uses shared variance to assign results to overarching domains; this means that an eight-item neurocognitive battery can be reduced to just two variables. In a recent study, we developed two principal components on a larger group of 292 patients (23). We used the same algorithm to obtain PCA scores on the 25(OH)D cohort because vitamin D levels are unlikely to change each test’s contribution to the principal components. The first component was primarily composed of word list learning, recall, and recognition and reflected memory. The second component represented executive functioning, attention and processing speed, with Trail Making Tests A and B, Block Design, and Digit Symbol-Coding tests contributing substantially. PCA domains score had a mean of 0 with SD of 1; a higher score indicated better cognitive performance. We show the results of the individual neurocognitive tests as well as the principal components because the principal components are an artificial construct and their results are not as easy to interpret.

Initially, we performed univariate regression analyses with scores on individual tests and on the domains of PCA as the outcome variables and 25(OH)D levels as the predictor variable. We then fitted parsimonious models adjusted for age only because older patients had higher 25(OH)D levels. We additionally adjusted for sex, race, education, smoking, diabetes status, time of year (May-August, reflecting sunlight availability, versus all other months), BMI, and dialysis vintage given that these were associated with 25(OH)D status at baseline.

We performed several sensitivity analyses. First we adjusted our models for Kt/V, calcium, phosphorus, and serum albumin because these variables may affect cognitive function and 25(OH)D levels. Second, we used different cutoffs for low-sunlight months (November–February) because serum 25(OH)D levels might lag behind seasonal changes (24). Third, we adjusted for activated vitamin D, which may reduce 25 vitamin D levels (25).

Results were considered statistically significant at a P value <0.05. All analyses were performed using SAS software, version 9.2 (SAS Institute, Inc., Cary, NC).

Results

Two hundred seventy-four patients had 25(OH)D levels measured. Nineteen patients were receiving oral 25(OH)D replacement and were therefore excluded. Demographic or clinical characteristics did not significantly differ between patients excluded and included. Of the remaining 255 patients, the mean age was 63 years. Fifty-five percent were male and 77% were white. Mean ± SD 25(OH)D level was 17.2±7.4 ng/ml. Thirty-one percent patients had sufficient, 55% had insufficient, and 14% had deficient 25(OH)D levels (Table 1). Sixty percent of the patients were receiving active vitamin D replacement. There were more women and patients with diabetes in the deficient group than the other groups (P<0.05). Patients who had sufficient 25(OH)D levels were more likely to have levels checked in the high-sunlight months (May–August) than those who had insufficient or deficient 25(OH)D levels (46.3%, 34.5%, and 30.6%, respectively; P=0.06). There were fewer current or past smokers in the deficient and insufficient groups than the sufficient group (P=0.05).

Table 1.

Clinical and demographic variables, stratified by 25-hydroxyvitamin D levels

Variable	25(OH)D Level
	<12 ng/ml	12 to < 20 ng/ml	≥20 ng/ml	Total	Trend P Value
Patients, n (%)	36 (14.1)	139 (54.5)	80 (31.4)	255
Vitamin D (ng/ml)	8.1±1.6	14.6±2.6	25.9±5.9	17.2±7.4
Age (yr)	62.7±17.4	61.9±17.6	64.5±15.9	62.9±16.9	0.52
Women (%)	52.8	48.9	35.0	45.1	0.04
Black (%)	27.8	25.9	16.3	23.1	0.10
Education (%)					0.91
<12th grade	13.9	9.4	10.0	10.2
High school graduate	50.0	54.0	53.8	53.3
> 2 years college	36.1	36.7	36.3	36.5
Stroke (%)	11.1	19.4	13.8	16.5	0.95
Peripheral vascular disease (%)	27.8	20.9	22.5	22.4	0.67
Hypertension (%)	91.7	90.7	90.0	90.6	0.78
Diabetes (%)	63.9	46.0	38.8	46.3	0.02
Heart failure (%)	47.2	30.2	37.5	34.9	0.64
Coronary artery disease (%)	33.3	36.0	35.0	35.3	0.93
Primary cause of ESRD (%)					0.25
Diabetes	44.4	34.5	26.3	33.3
Hypertension	11.1	18.7	23.8	19.2
Other	13.9	18.0	13.8	16.1
Unknown	16.7	10.8	15.0	12.9
GN	13.9	18.0	21.3	18.4
Smoking history (%)					0.05
Never	39.4	44.4	26.9	38.2
Past	51.5	44.4	66.7	52.4
Current	9.1	11.1	6.4	9.4
Vascular access (%)					0.78
Fistula	66.7	66.9	63.8	65.9
Graft	6.1	6.6	5.0	6.0
Catheter	27.3	26.5	31.3	28.1
Timing of 25(OH)D testing (%)					0.06
Low-sunlight months (January–April or September–December)	69.4	65.5	53.8	62.4
High-sunlight months (May–August)	30.6	34.5	46.3	37.7
Monthly mean BP (mmHg)
Pre SBP	140±21	141±22	142±18	141±21	0.48
Pre DBP	72±11	73±13	74±12	73±12	0.33
spKt/V	1.52±0.25	1.52±0.27	1.48±0.20	1.51±0.25	0.27
Body mass index (kg/m2)	29.4±7.4	29.1±7.4	27.5±6.6	28.6±7.2	0.07
Intradialytic weight gain (kg)	2.7±1.1	2.8±1.1	2.5±1.1	2.7±1.1	0.13
Hematocrit (%)	34.8±2.9	35.2±3.6	35.9±3.6	35.4±3.5	0.06
Albumin (g/dl)	3.8±0.4	3.8±0.3	3.8±0.4	3.8±0.4	0.92
Phosphorus (mg/dl)	5.6±1.5	5.7±1.4	5.3±1.6	5.6±1.5	0.07
Dialysis duration (mo)	19 (7, 36)	17 (9, 36)	11 (6, 29)	15 (7, 35)	0.01
Active vitamin D (%)	61.1	59.7	60	60	0.93

Data are presented as percentages, mean ± SD, or median (25th, 75th percentile), as appropriate. 25(OH)D, 25-hydroxyvitamin D; Pre SBP, pre dialysis systolic blood pressure; Pre DBP, pre dialysis diastolic blood pressure.

No significant association was seen between 25(OH)D levels and performance on individual cognitive tests in unadjusted linear regression models, except for the mental alternation test (Table 2). After adjustment for age, there was a statistically significant association between 25(OH)D status and several tests that primarily assess executive function (Table 2). The association remained significant after additional adjustment for race, sex, education, smoking, diabetes, sunlight exposure, dialysis vintage, and BMI. In multivariable analysis, the executive function domain of the PCA was also significantly associated with 25(OH)D levels, with no significant association seen with the memory domain.

Table 2.

Association between 25-hydroxyvitamin D levels modeled as a continuous variable and cognitive performance

Cognitive Test/Component	Patients (n)	Unadjusted		Parsimoniousa		Adjustedb
		Mean Differencec (95% CI)	P Value	Mean Differencec (95% CI)	P Value	Mean Differencec (95% CI)	P Value
Executived	254	0.13 (−0.01 to 0.27)	0.06	0.21 (0.08–0.33)	0.001	0.16 (0.04–0.28)	0.01
Memoryd	254	−0.15 (−0.28 to −0.03)	0.02	−0.09 (−0.21 to 0.03)	0.12	−0.06 (−0.19 to 0.06)	0.32
MMSE	255	0.05 (−0.32 to 0.42)	0.80	0.15 (−0.22 to 0.51)	0.44	0.22 (−0.14 to 0.58)	0.23
Delayed Recall	251	−0.30 (−0.65 to 0.05)	0.09	−0.14 (−0.46 to 0.18)	0.40	−0.05 (−0.38 to 0.27)	0.75
Short Delayed Recall	252	−0.18 (−0.55 to 0.18)	0.33	0.00 (−0.33 to 0.33)	0.99	0.08 (−0.26 to 0.42)	0.63
Recall Total	253	−0.64 (−1.77 to 0.29)	0.18	−0.07 (−0.86 to 0.72)	0.86	0.11 (−0.69 to 0.91)	0.78
Percentage Retention	251	−2.69 (−6.32 to 0.94)	0.15	−1.63 (−5.17 to 1.90)	0.36	−1.06 (−4.79 to 2.66)	0.57
Recognition	251	−0.31 (−0.69 to 0.07)	0.11	−0.16 (−0.52 to 0.20)	0.39	−0.21 (−0.58 to 0.17)	0.28
Block Design	249	0.20 (−1.22 to 1.63)	0.78	0.84 (−0.49 to 2.18)	0.21	0.24 (−1.03 to 1.50)	0.71
Digit Forward	98	0.27 (−0.23 to 0.77)	0.28	0.31 (−0.20 to 0.81)	0.24	0.17 (−0.35 to 0.69)	0.52
Digit Backward	98	0.37 (−0.13 to 0.87)	0.14	0.43 (−0.07 to 0.94)	0.09	0.32 (−0.17 to 0.81)	0.19
Digits Total	98	0.64 (−0.22 to 1.51)	0.14	0.74 (−0.14 to 1.62)	0.10	0.49 (−0.37 to 1.36)	0.26
Digit Symbol Coding	230	0.48 (−1.79 to 2.75)	0.68	2.10 (0.24–3.95)	0.03	2.02 (0.23–3.82)	0.03
Mental Alternation	99	2.23 (0.68, 3.79)	0.0005	2.55 (1.00–4.11)	0.002	2.20 (0.69–3.71)	0.005
COWAT Market	99	1.28 (−0.04 to 2.60)	0.06	1.58 (0.27–2.89)	0.02	1.73 (0.38–3.07)	0.01
COWAT Animal	99	0.87 (−0.37 to 2.11)	0.17	1.33 (0.15–2.51)	0.03	1.20 (0.04–2.36)	0.04
Trail Making Test A	238	−4.04 (−9.60 to 1.52)	0.15	−6.46 (−11.68 to −1.24)	0.02	−5.85 (−11.30 to −0.39)	0.04
Trail Making Test B	234	−7.55 (−22.61 to 7.50)	0.33	−15.85 (−28.85 to −2.85)	0.02	−11.52 (−24.61 to 1.58)	0.08

CI, confidence interval; MMSE, Mini-Mental Status Examination; COWAT, Controlled Oral Word Association Test.

^aAdjusted for age.

^bAdjusted for age, sex, race, education, diabetes, smoking, high May–August versus low (all other months) sunlight months, dialysis vintage, and body mass index.

^cMean difference represent a difference in performance on cognitive tests per 1 SD (7.4 ng/ml) higher 25-hydroxyvitamin D levels. A higher score on the tests indicates better cognitive function, except for Trail Making Tests A and B, for which a higher score indicates worse cognitive function. Analyses for Trail Making Test B test use Tobit regression censoring for failure to complete the task within 5 minutes.

^dPrincipal component analysis was used to calculate scores on executive and memory components, which have a mean of 0 and SD of 1, with positive values representing better performance.

Results showed similar trends when 25(OH)D status was modeled as a categorical variable (Table 3). Patients in the sufficient group had higher levels of cognitive function, particularly in the domains that assess executive function. There was no significant association between 25(OH)D and tests that primarily assessed memory.

Table 3.

Association between 25-hydroxyvitamin D levels modeled as a categorical variable and cognitive performance

Cognitive Test	Unadjusted			Parsimoniousa			Fully Adjustedb
	<12 vs. ≥ 20 ng/mlc	12 to <20 vs. ≥20 ng/mlc	Trend P Value	<12 vs. ≥ 20 ng/mlc	12 to <20 vs. ≥20 ng/mlc	Trend P Value	<12 vs. ≥ 20 ng/mlc	12 to 20 vs. ≥20 ng/mlc	Trend P Value
Executive	−0.44 (−0.86 to −0.02)	−0.07 (−0.36 to 0.23)	0.07	−0.51 (−0.89 to −0.14)	−0.14 (−0.40 to 0.12)	0.01	−0.40 (−0.77 to −0.03)	−0.12 (−0.37 to 0.14)	0.05
Memory	0.12 (−0.08 to 0.52)	0.27 (−0.01 to 0.54)	0.28	0.06 (−0.31 to 0.43)	0.20 (−0.05 to 0.46)	0.44	−0.01 (−0.38 to 0.37)	0.18 (−0.07 to 0.44)	0.67
MMSE	−0.33 (−1.45 to 0.79)	0.24 (−0.54 to 1.03)	0.79	−0.40 (−1.49 to 0.70)	0.14 (−0.62 to 0.91)	0.65	−0.46 (−1.52 to 0.59)	0.09 (−0.63 to 0.82)	0.55
Delayed Recall	0.02 (−1.04 to 1.08)	0.74 (0.00–1.47)	0.50	−0.10 (−1.05 to 0.86)	0.56 (−0.10 to 1.22)	0.71	−0.24 (−1.21 to 0.73)	0.50 (−0.17 to 1.16)	0.92
Short Delayed Recall	−0.57 (−1.66 to 0.52)	0.50 (−0.26 to 1.26)	0.67	−0.71 (−1.67 to 0.26)	0.31 (−0.37 to 0.99)	0.38	−0.95 (−1.94 to 0.04)	0.24 (−0.44 to 0.93)	0.21
Recall Total	0.43 (−2.37 to 3.23)	1.44 (−0.53 to 3.40)	0.48	0.02 (−2.33 to 2.36)	0.84 (−0.80 to 2.49)	0.74	−0.33 (−2.69 to 2.02)	0.66 (−0.97 to 2.28)	0.97
Percentage Retention	0.56 (−10.40 to 11.52)	6.98 (−0.64 to 14.60)	0.50	0.20 (−10.77 to 10.37)	5.85 (1.51–13.21)	0.63	−1.37 (−12.37 to 9.63)	5.62 (−1.94 to 13.17)	0.78
Recognition	−0.05 (−1.21 to 1.11)	0.55 (−0.26 to 1.36)	0.71	−0.16 (−1.24 to 0.92)	0.39 (−0.37 to 1.14)	0.92	−0.03 (−1.14 to 1.07)	0.52 (−0.24 to 1.28)	0.70
Block Design	−0.87 (−5.25 to 3.51)	1.15 (−1.83 to 4.13)	0.96	−1.51 (−5.88 to 2.56)	0.53 (−2.25 to 3.31)	0.65	−0.03 (−3.83 to 3.78)	1.40 (−1.15 to 3.95)	0.71
Digits Forward	−0.92 (−2.72 to 0.87)	−0.26 (−1.29 to 0.78)	0.34	−0.90 (−2.70 to 0.90)	−0.31 (−1.37 to 0.76)	0.32	−0.56 (−2.39 to 1.28)	−0.13 (−1.20 to 0.94)	0.61
Digits Backward	−0.98 (−2.77 to 0.81)	0.05 (−0.99 to 1.08)	0.48	−0.94 (−2.73 to 0.85)	−0.05 (−1.11 to 1.01)	0.44	−0.79 (−2.52 to 0.95)	0.21 (−0.81 to 1.23)	0.70
Digits Total	−1.91 (−5.03 to 1.21)	−0.21 (−2.01 to 1.59)	0.34	−1.84 (−4.96 to 1.28)	−0.36 (−2.21 to 1.49)	0.31	−1.34 (−4.42 to 1.73)	0.08 (−1.72 to 1.88)	0.60
Digit Symbol Coding	−4.46 (−11.66 to 2.74)	0.65 (−4.15 to 5.44)	0.40	−6.66 (−12.51 to −0.81)	−0.59 (−4.48 to 3.31)	0.07	−6.01 (−11.61 to −0.42)	−1.58 (−5.25 to 2.08)	0.06
Mental Alternation	−6.25 (−11.93 to −0.57)	−2.46 (−5.72 to 0.81)	0.02	−5.98 (−11.60 to −0.36)	−3.07 (−6.38 to 0.25)	0.02	−5.09 (−10.55 to 0.36)	−2.26 (−5.45 to 0.92)	0.06
COWAT Market	−4.38 (−9.12 to 0.37)	−1.92 (−4.65 to 0.81)	0.05	−4.09 (−8.57 to 0.57)	−2.57 (−5.32 to 0.18)	0.03	−4.85 (−9.64 to −0.06)	−2.58 (−5.38 to 0.22)	0.03
COWAT Animal	−4.77 (−9.18 to −0.36)	−0.42 (−2.95 to 2.12)	0.11	−4.33 (−8.50 to −0.15)	−1.40 (−3.87 to 1.06)	0.05	−3.28 (−7.42 to 0.86)	−1.05 (−3.47 to 1.37)	0.15
Trail Making Test A	17.30 (0.31–34.30)	−2.29 (−14.08 to 9.49)	0.14	19.55 (3.67–35.44)	0.11 (−10.93 to 11.15)	0.05	17.77 (1.36–34.17)	0.81 (−10.44 to 12.06)	0.09
Trail Making Test B	34.21 (−12.11 to 80.53)	15.12 (−16.99 to 47.22)	0.14	41.55 (1.70–81.40)	23.27 (−4.32 to 50.87)	0.03	33.08 (−6.46 to 72.62)	19.39 (−7.75 to 46.53)	0.07

MMSE, Mini-Mental Status Examination; COWAT, Controlled Oral Word Association Test.

^aAdjusted for age.

^bAdjusted for age, sex, race, education, diabetes, smoking, high-sunlight (May–August) versus low-sunlight (all other months) months, dialysis vintage, and body mass index. Principal component analysis was used to calculate scores on executive and memory components, which have a mean of 0 and SD of 1, with positive values representing better performance.

^cValue are the mean difference in performance on cognitive tests comparing deficient (<12 ng/ml) or insufficient (12 to <20 ng/ml), versus sufficient (≥20 ng/ml) groups. Negative coefficients represent worse performance in those deficient or insufficient, except for Trail Making Tests A and B, for which positive coefficients indicate worse performance versus the sufficient group. Numbers in parentheses are 95% confidence intervals.

The results did not differ after adjustment for serum calcium, phosphorus, Kt/V, hematocrit, albumin, different cutoffs for sunlight months, or activated vitamin D.

Discussion

In this moderate-sized sample of patients undergoing maintenance hemodialysis, we demonstrate that lower 25(OH)D levels are associated with worse performance on tests that assess executive function. Our study also confirms the high prevalence of 25(OH)D deficiency and insufficiency in hemodialysis patients.

Previous studies that have evaluated the relationship of 25(OH)D status with cognitive impairment have yielded conflicting results. In elderly European men, low 25(OH)D levels, defined as <14 ng/ml, were associated with poor performance on tests that assess speed of information processing but not with tests that assess memory (26). In another study, multivariable analyses showed that 25(OH)D levels <20 ng/dl were associated with a greater than twofold increase in the risk of all-cause dementia and stroke (7). Conversely, in a large population-based cross-sectional study, no association was noted between 25(OH)D levels and tests that assessed memory or executive function (27). The latter study may have been limited by the fact that the outcome variable was dichotomous and thus may not have been sensitive enough to detect associations. Furthermore, a comprehensive battery of neuropsychological testing was not performed to assess cognition. In dialysis patients, there are very few data relating 25(OH)D levels to cognitive function. In one large study of incident dialysis patients, lower 25(OH)D levels were associated with lower self-reported physical activity and self-reported mental health in adjusted analyses (28).

There are several mechanisms through which 25(OH)D may be associated with cognitive impairment. Low levels of 25(OH)D have been associated with several vascular risk factors, including hypertension (29), calcification of blood vessels (30), and inflammation (31). In turn, each of these risk factors may promote vascular disease. 25(OH)D may also have neuroprotective properties (10). Higher levels of 25(OH)D have been associated with better cognitive function in patients with Alzheimer’s disease (32). 25(OH)D and its receptors, as well as the enzyme 1α hydroxylase have all been detected in glial and neuronal cells (33), and although activated vitamin D (1,25[OH₂]D) regulates genes associated with neuronal cell proliferation and differentiation, its uptake across the blood-brain barrier is minimal (34). Thus, the majority of the 1,25(OH₂)D in neuronal cells is produced in vivo, and 25(OH)D levels are probably important, independent of repletion with 1,25(OH₂)D analogues or 1,25(OH₂)D levels. Finally, 25(OH)D is known to affect immune modulation (35) and through this mechanism may prevent immune-mediated neuronal injury.

We have previously shown that hemodialysis patients primarily manifest deficits in executive function and that prevalent vascular disease is a risk factor for executive function impairment (3,36). Therefore, we postulate that the association of low 25(OH)D with worse cognitive function may be due to its vasculoprotective properties. Supporting the latter is a recent age- and race-matched case-control study in hemodialysis patients, which showed that patients receiving oral ergocalciferol had lower levels of vascular adhesions molecules than the patients who were not receiving ergocalciferol (31).

Our study has several limitations. Patients with lower 25(OH)D levels are generally more sick than the patients with higher levels, and, although we adjusted analyses for several comorbid conditions, unmeasured or residual confounding remains a possibility. With regard to the former, we did not adjust for parathyroid hormone (PTH) levels in our analyses because the PTH assay changed during the study and PTH levels are known to vary widely. High PTH levels have been associated with vascular disease and mortality in some (37,38), but not all (39–41), studies; this continues to be a controversial topic. Although we excluded patients who were receiving oral 25(OH) vitamin D2 or D3 replacement at the time of the study, it is possible that ascertainment of use was not 100% accurate or that individuals had received it in the past. We believe that our results would most likely have been biased to the null if that were the case because the 25(OH)D status of these individuals would have been misclassified. Sixty percent of the cohort received activated vitamin D, which in turn may be associated with reduced plasma half-life of 25(OH)D (25). Adjustment for activated vitamin D did not, however, change our results. Finally, as in all cross-sectional studies, we cannot establish causation, and in fact reverse causation remains a theoretical possibility given that cognitively impaired patients may be less active and therefore have less sunlight exposure.

Our study also has several strengths. We had a fairly large sample of patients that had adequate representation of both sexes and of African Americans. We used a battery of neuropsychological tests that comprehensively assess all domains of cognition because no single test can adequately assess all domains of cognition. The test administrator was periodically observed and evaluated by a neuropsychologist to ensure inter- and intrarater reliability. Finally, we limited the concerns of multiple hypothesis testing and co-linearity by using PCA.

In conclusion, we demonstrate that there is a high prevalence of 25(OH)D in hemodialysis patients and that vitamin D deficiency may be associated with worse executive function. These results need to be confirmed in large prospective studies. A randomized double-blinded, placebo-controlled pilot study (NCT01229878) (42) is underway to assess the effects of ergocalciferol repletion on physical and cognitive function in 50 maintenance hemodialysis patients. Although small, the results of this trial may provide further insight into this problem in dialysis patients.

Disclosures

None.