Registration Dossier

Data platform availability banner - registered substances factsheets

Please be aware that this old REACH registration data factsheet is no longer maintained; it remains frozen as of 19th May 2023.

The new ECHA CHEM database has been released by ECHA, and it now contains all REACH registration data. There are more details on the transition of ECHA's published data to ECHA CHEM here.

Diss Factsheets

Toxicological information

Epidemiological data

Currently viewing:

Administrative data

epidemiological data
Type of information:
other: Epidemiological observational
Adequacy of study:
supporting study
Study period:
1991 - 2002
2 (reliable with restrictions)
Rationale for reliability incl. deficiencies:
other: Basic data given

Data source

Reference Type:
other: Thesis
Aluminum in Drinking Water and Alzheimer’s Disease: Analysis of the Canadian Study of Health and Aging Prospective Cohort.
Boom, N.
Bibliographic source:
Thesis. 2008. Ottawa

Materials and methods

Study type:
cohort study (prospective)
Endpoint addressed:
Test guideline
no guideline required
GLP compliance:

Test material

Constituent 1
Reference substance name:
aluminium in drinking water
aluminium in drinking water
Details on test material:
- Name of test material (as cited in study report): aluminium in drinking water
- Analytical purity: no data


Type of population:
Ethical approval:
confirmed and informed consent free of coercion received
Details on study design:
HYPOTHESIS TESTED (if cohort or case control study):
The association between local levels of aluminium in drinking water and the risk of developing Alzheimer’s disease (AD)

- Type: Questionnaire / other:
- Details:
Cognitively normal subjects at baseline provided information on all their residencies during their life, on sources of drinking water at each residence (well vs. municipal) and on a number of potentially confounding factors.
Data on water parameters for the period 1980-2000 were received from water treatment plants supplying the subjects’ residences


Canadian Study of Health and Aging: 36 cities and surrounding areas in 5 Canadian regions: British Columbia, the Prairies, Ontario, Quebec and the Atlantic provinces.

- Total population (Total no. of persons in cohort from which the subjects were drawn): A random sample of 15,677 individuals aged 65 years or older
- Participation rate: 10,263 individuals, including 72.1% from the community sample (9,008) and 81.7% of the institutional sample (1,225)
- Selection criteria: Availability of risk factor questionnaires (1,833 subjects excluded based on this criterion) and residential history information (561 subjects excluded); ≥7 years of residential history information during 1980-1991 (157 subjects excluded); no AD at baseline (557 subjects excluded)
- Total number of subjects participating in study: 7,155 subjects; 4,507 of them with successfully assigned Al drinking water concentrations were used in the main analyses.
- Sex/age/race: 65 years or older; both genders
- Smoker/nonsmoker: smokers and non-smokers
- Total number of subjects at end of study: 2,847

- Type: State registry / Regional registry / National registry / Control or reference group / Other comparison group:
- Details:
A category of 0-46.4 µg/L aluminium concentration in drinking water was used as a reference in the categorical analysis.

- Disease(s): Alzheimer’s disease (AD), possible or probable
- ICD No.: ICD-10
- Year of ICD revision: 1993 (see Lindsay et al., Am J Epidemiol. 2002 Sep 1; 156(5):445-53).
- Diagnostic procedure:
Baseline examination: For individuals living in the community, a home interview was conducted, including the Modified Mini-Mental State Examination (3MS) screening test for cognitive impairment. Individuals with a 3MS score below 78/100 were subject to a three-stage medical examination (All participants living in institutions received a clinical examination without the screening interview):
1) The 3MS examination was re-administered by a nurse
2) A physician conducted a standardized physical and neurological examination.
3) A psychometrist administered 15 neuropsychological tests to subjects with 3MS score of 50 or above, i.e. to those whose cognitive status was sufficient for the testing process. The subjects with a 3MS score less than 50 were not given the neuropsychological tests and did not receive a neuropsychological diagnosis.
The physician and neuropsychologist independently made a preliminary diagnosis based on DSM-III-R criteria. The diagnoses were compared at a case consensus conference with participation of the whole clinical team. Those with dementia were classified using NINCDS-ADRDA criteria. AD cases for this study were defined as subjects diagnosed with possible or probable AD.
1996-1997 follow-up examination (CSHA-2): same diagnostic process
2001-2002 follow-up examination (CSHA-3): Same methods as in the previous examinations, but the 3MS cutpoint was changed from <78 to <90. Subjects scoring between 50 and 90 on the 3MS and all subjects who had undergone a previous clinical examination continued to the neuropsychological examination. Subjects classified by the neuropsychologist as cognitively impaired attended the clinical assessment, as well as participants with 3MS scores below 50 and subjects who received a clinical evaluation at CSHA-2.

Only incident cases defined as subjects who were free of dementia at baseline and were diagnosed with probable or possible Alzheimer’s disease at CSHA-2 or CSHA-3 were included in the analysis

- Other health effects:

Enrollment: 1991-1992 (CSHA-1)
Follow-ups: 1996-1997 (CSHA-2) and 2001-2002 (CSHA-3)
Follow-up started on the day of the first screening interview and ended when the subject either was diagnosed with AD, died, was lost to follow-up or completed the CSHA-3. The date of exit from the study for a subject diagnosed with AD at CSHA-2 or CSHA-3 was set at half-way between the last CSHA follow-up when the subject was free of the disease and the follow-up when he/she was diagnosed with AD.
Duration: maximum 11 years.
Completeness: CSHA-2 participation – 4832 subjects out of 7155 participated: 1578 (22%) died and 745 (10%) refused to participate or could not be contacted.
CSHA-3 participation – 2847 subjects participated: 1120 (23% of those participating in CSHA-2) died and 434 (9%) refused to participate or could not be contacted.
Exposure assessment:
Details on exposure:
Oral via drinking water

Measurements of Al in drinking water at the water treatment plants



Al (µg/L) Number of subjects Number of subjects
(model with adjustment (fully adjusted model)
for age and gender only)

0-46.4 1039 798
46.4-103.8 1071 851
103.8-121.5 1478 1189
121.5-443.3 438 356
>443.3 481 390
Statistical methods:
1) Cox proportional hazard regression models.
2) Poisson regression approach (AMFIT Poisson regression) to evaluate if the presence of heavily tied data biased the results

Results and discussion

- Other: Average Al concentration over all municipalities: 134.04 µg/l. Range of municipal averages: 4.5µg/l – 749.7 µg/l.
Average Al concentration for incident AD cases – 150.8 µg/l (95% CI: 131.6, 167.0);
for subjects without AD – 135.5 µg/l (95% CI : 128.2, 142.8)
for subjects lost to follow-up – 131.0 µg/l (95% CI: 110.0, 152,1)

pH average over all municipalities: 7.41, range of municipal averages: 5.7-8.9
Fluoride: average over all municipalities- 0.55 mg/l; range 0.015 mg/l – 1.13 mg/l
Silica: average across all municipalities – 2.70 mg/l; range 0.29 mg/l – 12.45 mg/l
Iron: average across all municipalities - 48,8 µg/l, the municipal averages ranged from 3.67 µg/l to 272.0 µg/l.
Al was positively correlated with fluoride, pH and silica and negatively correlated with iron


- Incidence/ Number of cases for each disease / parameter under consideration:
- Other:
490 incident cases of probable or possible AD

- SMR (Standard mortality ratio):
- RR (Relative risk): HR as an estimate of RR
- OR (Odds ratio):
- Other:

Cox regression, fully adjusted model

Categorical analysis
Al (µg/l) HR (95% CI)
0-46.4 1.00 (Ref.)
46.4-103.8 1.06 (0.72-1.57)
103.8-121.5 1.10 (0.77-1.58)
121.5-443.3 0.94 (0.55-1.61)
>443.3 1.46 (0.96-2.24)

Continuous analysis:
HR associated with an increase in Al concentration of 331.81 µg/l (the difference between the means in the highest and in the lowest quartile of exposure) was 1.28 (95% CI: 1.02-1.60)

Threshold model:
Al concentrations > 100 µg/l vs. < 100 µg/l: HR 1.17 (95% CI: 0.94-1.44)

AMFIT Poisson regression, fully adjusted model, categorical analyses
Al (µg/l) HR (95% CI)
0-46.4 --
46.4-103.8 1.00 (Ref.)
103.8-121.5 1.06 (0.72-1.57)
121.5-443.3 1.12 (0.78-1.61)
>443.3 1.27 (0.86-1.85)
The results were very similar to those obtained with the Cox model.

Cox regression, categorical analysis, fully adjusted model
The overall association assessed by the Wald λ² test was not significant (P=0.4088). No significant increase or decrease in HR in any of the Al categories for pairwise combinations of Al with any other water parameter or for a model including all 5 water parameters together; HRs were reduced in these analyses compared to those listed above.
Adjustment for ApoE status (892 subjects) makes the HR in the top Al category statistically significant: 1.86 (95% CI: 1.05-3.27)
Cox regression, continuous analysis, fully adjusted model
When any of the water parameters are included in the model, the HR is not significant. .
HR adjusted for ApoE status (892 subjects) was 1.44 (95% CI 1.08-1.92)
No evidence of interaction between any of the water parameter variables in either the categorical or continuous model.
Sensitivity analyses (see “Methods”)
1) Accounting for undiagnosed AD for deceased subjects did not appreciably change HRs in either categorical or continuous analyses.
2) Including CIND with the study outcome resulted in a slight change in the HRs but they remained statistically insignificant. The HR in the continuous analysis became non-significant: 0.99 (95% CI: 0.82-1.19)
3) Two-stage regression: RR of 1.22 (95% CI: 0.92-1.63) per 331.81 µg/l increase in Al concentration
Confounding factors:
Fluoride, silica, iron, pH in the drinking water

Gender, age, education, history of stroke, blood pressure

ApoE allele status was determined in a sub-sample of subjects who underwent a clinical examination (892 subjects who also had Al data)

Information on a number of potentially confounding factors was collected at baseline. A variable was retained in the statistical model only if its inclusion changed the HR of at least one exposure quartile by more than 10%.
Strengths and weaknesses:
Large prospective study of a well-defined cohort; detailed information on other potential risk factors for AD including ApoE genotype; availability of information on other water parameters, well-defined health outcome.
The authors point out the following:
- Variations in the information collected from different water treatment plants due to differences in measurement techniques, accuracy and precision, which may lead to exposure misclassification;
- Unavailability of data for some treatment plants for some years resulted in extrapolations in the exposure estimation, which could introduce errors in exposure estimates
- Only total Al in drinking water was available with no speciation
- Values of water parameters at water treatment plants may be different from those at the residential taps due to their changes throughout the distribution system.
- Individual characteristics of exposure (e.g. amount of water consumed) were not available
- Etiologically relevant period of exposure is not well known.

Any other information on results incl. tables

Sensitivity analyses:

1) To assess the effect on risk estimates of failure to account for undiagnosed AD for subjects who died between the CSHA follow-ups: the 82 deceased subjects who were reported by proxy respondents as having been diagnosed with AD before they died were classified as having the outcome of interest.

2) Cognitive impairment not dementia (CIND) often progresses to AD. An analysis was conducted classifying subjects who were diagnosed with CIND at CSHA-2 and did not develop AD by CSHA-3 and subjects diagnosed with CIND at CSHA-3 as having the outcome of interest.

A two-stage regression analysis allowing the baseline hazard function of the standard Cox model to vary at random among the municipalities was conducted to explore the possible effects of spatial auto-correlation (due to assignment of municipal-level data to individuals)

Applicant's summary and conclusion

The results of this study, though suggestive of a higher risk of AD at higher Al levels in drinking water, are equivocal in that inconsistent results were obtained from different analytical approaches. Aluminium in drinking water represents only a small fraction of total oral aluminium exposure, dietary and pharmacological sources being considerably higher and variable. The validity of the exposure measure is unclear.
Executive summary:

This prospective cohort study of potential effects of aluminium in drinking water on the risk of Alzheimer’s disease was based on a sample of 7, 155 subjects from the Canadian Study of Health and Aging. The subjects were recruited in 1991 - 1992, and the incident cases of Alzheimer’s disease (n = 490) were ascertained at follow-up examinations in 1996 - 1997 and in 2001 - 2002.Exposure assessment was based on residential history information collected from the subjects and data on aluminium concentrations and on several other water parameters (fluoride, silica, iron, pH) from the water treatment plants supplying the subjects’ residencies. A number of potentially confounding variables were explored, and only those changing the hazard ratio of at least one exposure category by more than 10% were retained in the final model. Thus, the analyses were adjusted for gender, age, education, history of stroke and blood pressure.ApoE allele status was determined in a sub-sample of subjects (N = 892), and a secondary analysis was conducted with adjustment for this factor.A significant association between the risk of Alzheimer’s disease and Al was found in the analysis applying the continuous standard Cox proportional hazard model, but not in the analysis using the categorical Cox proportional hazard model. After adjusting for autocorrelation in a two stage model no significant association was observed. No interaction between any of the water parameter variables in either the categorical or continuous analysis was observed. The results from this study are equivocal concerning an association between aluminium intake and an increase in the risk of Alzheimer disease.