Salib & Sharp, 1999 
Daily hospital admissions from North Cheshire to a single hospital. Dementia cases were identified by ICD-9 code 290.
189/2070 psychiatric admissions during 1993 were coded as being related to dementia.
UK Meteorological Office data were collected at Manchester airport (30 miles away) every day.
There were no associations found between weather parameters and hospital admissions of people with dementia.
Huss et al., 2009 
Magnetic field exposure (220–380 kV) through power lines
Swiss National Cohort from 2000–5. The study population comprised 4.65 million individuals and 22,821,824 person-years.
29,975 dementia deaths were recorded, including 9,228 AD deaths.
Prospective cohort study: Cox PH models. Exposure was based on distance of place of residence to the nearest power line and duration of exposure (5, 10, or 15 years). Dementia was ascertained from death certification.
Proximity to power lines was associated with an increased risk of dementia but this was not statistically significant at conventional levels (adjusted HR, 95 % CI closest:most distant categories 1.23, 0.96-1.59). Longer duration increased the magnitude of this association (≥15 years at this place of residence 2.00, 1.21-3.33).
Vergara et al., 2013 
Extremely low frequency magnetic fields
Systematic review of studies of occupational exposure to magnetic fields and neurodegenerative disease.
20 AD and 9 dementia studies.
Systematic review and meta-analysis.
There was a small association between occupational magnetic field exposure and AD based on a meta-analysis (RR, 95 % CI 1.27, 1.15-1.40). There was no clear association with dementia (1.05, 0.96-1.14).
Garcia et al., 2008 72]
Extremely low frequency electric and magnetic fields.
Systematic review and meta-analysis
Fourteen studies: 9 case–control, 5 cohort
Pooled cohort risk estimates (OR, 95 % CI 1.62, 1.16-2.27). Pooled case–control risk estimates (2.03, 1.38-3.00, P = 0.004).
Schuez et al., 2009 
All mobile phone subscriptions in Denmark, 1982–1995.
420,095 private mobile phone subscribers of whom 532 were admitted to hospital with a dementia code during follow up.
Prospective cohort study. Mobile phone use was derived from subscription records. Dementia status was identified from hospital admission records.
Mobile phone use was associated with a decreased risk of being hospitalised with AD (Standardized hospitalization ratio, 95 % CI 0.7, 0.6-0.9). Similar results were seen for VaD and “other dementia”.
Afzal et al., 2014 
Danish general population sample recruited to the Copenhagen Heart Study at baseline (1981 to 1983).
10,186 participants, of whom 418 developed AD and 92 developed to VaD.
Prospective cohort study: Cox PH models. Baseline plasma vitamin D levels were related to incident AD and VaD. Dementia status was derived from diagnostic codes recorded on the national Danish Patient Registry.
Lower plasma vitamin D levels were associated with an increased risk of AD (HR, 95 % CI <25th percentile [seasonally-adjusted] vs >50th 1.29, 1.01-1.66; P = 0.03). Similar findings were reported for VaD (1.22, 0.79-1.87; P = 0.42) and all dementia (1.27, 1.01-1.60; P = 0.02).
Littlejohns et al., 2014 
Healthy participants in the US population–based Cardiovascular
1658 adults followed up for mean 5.6 years during which time 171 developed clinically-ascertained dementia (102 AD).
Prospective cohort study: Cox PH models. Serum vitamin D levels were measured at baseline.
Being deficient (25-50nM) or severely deficient (<25nM) in vitamin D was associated with an increased risk of incident AD (multivariable-adjusted HR, 95 % CI: 1.69, 1.06-2.69; 2.22, 1.02-4.83; Ptrend = 0.008). Similar results were seen for all-cause dementia (1.53, 1.06-2.21; 2.25, 1.23-4.13; Ptrend = 0.002).
Annweiler et al., 2011 
Toulouse subset of the EPIDOS study cohort of women aged ≥75 years.
40 participants, of whom 6 developed non-AD (and 4 AD) at clinical assessment over 7 years follow up.
Prospective cohort study: logistic regression. Baseline serum vitamin D was related to incident non-AD dementia.
Vitamin D deficiency at baseline was associated with an increased risk of non-AD dementia (adjusted OR, 95 % CI 19.57, 1.11-343.69; P = 0.042).
Wilkins et al., 2006 
Participants recruited from greater metropolitan St Louis, MO by the Washington University AD Research Center.
40 people with clinically diagnosed mild AD and 40 without dementia.
Cross-sectional study (case–control): logistic regression and general linear model. Serum vitamin D levels were measured.
Vitamin D status was not associated with AD (multivariable-adjusted OR, 95 % CI: deficient vs sufficient 2.80, 0.64-12.28; insufficient vs sufficient 1.78, 0.61-5.19). Vitamin D status was associated with CDR sum of boxes (P = 0.0468); and the Short Blessed Test (P = 0.0077) but not others tests.
Frecker, 1991 
Mortality records in Bonavista Bay, Newfoundland, 1985–86.
191 dementia deaths in 1985, 208 deaths in 1986
Cross-sectional study. Place of birth of all individuals dying with dementia was identified and associated with drinking water samples at those locations from 1986.
The area with the highest dementia mortality (37.5 % in 1985 and 68.8 % in 1986) also had the lowest pH of drinking water (5.2). This association was not assessed for statistical significance, but was argued to not be confounded by age, sex or place of residence stated on death certificate.