Table 3.
Assessment of meta-analytic approach and results of the included systematic reviews
| First author, year | AMSTAR rating | Overall conclusion | Adjustment for potential confounder by the included articles (number of studies) | Analytic approach | Summary estimate (95% CI) | Sensitivity analysis | Subgroup analysis |
|---|---|---|---|---|---|---|---|
| Annweiler, 200911 | Moderate | Inconclusive association | No adjustment (n = 2), partial adjustment: age, gender, ethnicity, solar exposure education, serum vitamins B1, B6 and B12 (n = 3) | N/A | N/A | N/A | N/A |
| Annweiler, 201314 | High | Increased risk of Alzheimer disease | Partial adjustment in a few studies: age, gender, ethnicity, BMI, education, season, medical comorbidities, ApoE genotype, kidney function, plasma homocysteine, multivitamin use, home care centre | Fixed and random effects model with inverse variance method; effect size derived from mean, s.d. and size of each group | Summary effect size 1.40 (0.26–2.54), a large effect size (P = 0.016), I2 = 98% |
N/A | N/A |
| Annweiler, 201315 | High | Decreased executive dysfunction, uncertain effect on episodic memory | Partial adjustment in ten studies: age, gender, BMI, ethnicity, education, income, health status, physical activity, alcohol, tobacco and caffeine consumption, season, centre, serum concentration of PTH, calcium, vitamin E and zinc |
Fixed effect model; effect size derived from mean, s.d. and size of each group | 1. Episodic memory; effect size −0.09 (−0.16 to −0.03), I2 = 86% 2. Executive dysfunction (mean difference of scores); processing speed: 4.01 (1.20–6.83), I2 = 96%; mental shifting: 12.47 (6.78–18.16), I2 = 91%; information updating tests −0.45 (−0.60 to −0.29), I2 = 69% |
N/A | N/A |
| Balion, 201212 | High | Increased risk of Alzheimer disease, decreased cognitive function | Partial adjustment in 12 studies for season, sunlight exposure, site/centre, alcohol, smoking, BMI, baseline cognitive score, comorbidities, physical activities, physical performance, education, energy intake, multivitamin use, vitamin E, race, ethnicity, depression, psychoactive drugs, kidney function, biomedical measure of albumin, ApoE, vitamins B1, B6 and B12, calcium, homocysteine and PTH | Random effects model: using weighted mean difference and Hedge's g-test | 1. Alzheimer disease (n = 888): weighted mean difference: −6.2 (−10.6 to −1.8), I2 = 0% 2. MMSE (n = 2749) effect size 1.2 (0.5–1.9) I2 = 65% |
Two studies were excluded by CPBA assay, change in heterogeneity and overall estimate | A priori subgroup analysis: no change in MMSE point estimate |
| Cao, 201619 | Moderate | Increased risk of cognitive decline | Adjustment on three studies in the meta-analysis (age, gender, race, season, BMI, physical activity, education, income, hypertension, depression, smoking and alcohol consumption) | Random effects model | Risk ratio 1.52 (1.17–1.98) I2 = 45% |
N/A | N/A |
| Etgen, 201213 | Moderate | Increased risk of cognitive impairment | Partial adjustment of all the seven studies for age, gender, race, season, BMI, physical activity, education, income, comorbidities, measure of albumin, ApoE, vitamins B1, B6 and B12, calcium, homocysteine, PTH, zinc, smoking, alcohol consumption and pulse pressure | Random effects model | Odds ratio 2.39 (1.91–3.00) I2 = 56% |
Grouped studies based on study characteristics: gender, cognitive function test, sample size <500, community dwelling, in-patient, no change in overall estimate | |
| Lopes da Silva, 20147 | Moderate | No association between low levels of vitamin D and Alzheimer disease | Adjustment for age in meta-analysis | Random effects model fitted by restricted maximum likelihood | Weighted mean difference −5.77 (−12.11 to 0.58), P = 0.075 | Adjusted based on age, change in point estimate from significant to non-significant | N/A |
| Shen, 201518 | Moderate | Increased risk of Alzheimer disease and dementia | Partial adjustment of all ten studies for age, gender, season, sunlight exposure, alcohol, smoking, BMI, comorbidities, physical activities, physical performance, education, race, ethnicity, depression, ApoE, income, creatinine, cholesterol and plasma homocysteine | Random effects model | Alzheimer disease: odds ratio 1.21 (1.02–1.41) I2 = 0%; Dementia: odds ratio 1.49 (1.09–1.88) I2 = 0% |
N/A | N/A |
| Sommer, 201720 | High | Increased risk of dementia | Adjustment of all six studies | Fixed and random effects model with generic inverse variance model | Point estimate 1.54 (1.19–1.99) I2 = 20% |
Examination of heterogeneity based on different confounding factors, no change in estimates | N/A |
| van der Schaft, 201316 | Moderate | Increased risk of dementia and worsening cognitive function | No adjustment (n = 6), partial adjustment for age, education, BMI and gender (n = 22) |
χ2 exact test to analyse association between study variables and outcome direction | 71% of studies showed positive association between vitamin D level and cognitive function | N/A | N/A |
| Zhao, 201317 | Moderate | No association between low levels of vitamin D and Alzheimer disease | No adjustment | Random effects model | Summary mean difference −1.39 (−2.79 to 0.01), I2 = 98% |
N/A | N/A |
AMSTAR, A Measurement Tool to Assess Systematic Reviews; ApoE, Apolipoprotein E gene; BMI, body mass index; CPBA, competitive protein binding assay; MMSE, Mini-Mental State Examination; N/A, not applicable; PTH, parathyroid hormone.