Table 3 Summary of main methodological aspects of the epidemiological studies on the relationship between Alzheimer's disease and selected occupational exposures.

Authors/Design/Year/Place*	Gender/Ascertainment period/ Data collection methods†	Study population‡	Diagnostic criteria§	Exposure assessment¶	Control of confounding variables
Savitz et al. Retrospective cohort. 1998. USA33	Men. 1950–86. Based on death certificates	Information for 20 068 workers who worked for at least 6 months in electric companies. In 56 deaths, AD was indicated	ICD‐9, code 331.0	Exposure to EMF. Duration of work in exposed jobs was assessed and cumulative exposure at the time of death in five intervals was evaluated	Adjusted for age, calendar year, social class, work status, polychlorinated biphenyl exposure and solvent exposure
Kukull et al. Case–control. 1995. USA‐Seattle27	Both sexes. 1987–92. Personal interview to proxy respondents for cases and controls through structured questionnaire	193 Probable AD cases and 243 control subjects free of dementia and neurological disease, randomly selected from the study base	DSM III‐R and NINCDS‐ADRDA probable	Exposure to solvents through specially designed questions that incorporate duration of exposure. Different groups of solvents are assessed separately	Frequency matching for age and sex. Adjusted for age, sex, education, proxy relationship, alcohol consumption
Tyas et al. Prospective cohort. 2001. Manitoba, Canada38	Both sexes. 1991/2–1996/7. Subjects free of dementia received the questionnaire to complete and return by mail	694 Subjects who screened as cognitively intact were follow up for 5 years. 36 developed AD	Screening with 3MS. NINCDS/ADRDA possible or probable	Methods to measure exposure to specific agents (pesticides, solvents) through questionnaire are not defined. Different groups of solvents are assessed separately	Adjusted for age, sex, education
Baldi et al. Prospective cohort. 2003. Paquid Study‐France40	Both sexes. 1992–8. Personal interview through structured questionnaire to the cohort	1507 Subjects older than 65 years were followed up for 5 years. 96 AD cases were identified	DSM III‐R and NINCDS‐ADRDA. Cases were definitively classified by considering the results of jointly available complementary examinations	Exposure to pesticides: insecticides, herbicides, and fungicides was evaluated with a JEM made for 4 experts through detailed occupational histories. Cumulative exposure was calculated and quartiles were considered	Adjusted for age, tobacco consumption and education
Salib and Hillier. Case–control. 1996. UK28	Both sexes. 1991–3. Proxy respondents of both case and control groups in a direct interview using a structured questionnaire	198 Cases were compared with 164 controls with other dementias and 176 controls free of dementia	NINCDS–ADRDA possible or probable	Exposure to aluminium through occupational history from questionnaire. Subjects were labelled under “Aluminium occupation” category without additional information on criteria applied	Adjusted for age, sex, age of onset, duration of work, duration of condition and family history of dementia
Savitz et al. Case–control. 1998. USA‐25 different states35	Men. 1985–1991. Information was obtained from death certificates, available from the National Center for Health Statistics for selected states	256 Male cases who died of AD in 25 states, were compared with controls (ration 1:3) who died from causes other than leukaemia and brain cancer	ICD‐9, code 331.0	EMF: occupations reported on the death certificate were classified as electrical and non‐electrical occupation according to a previous study. Electrical occupations were additionally classified in 10 different groups	Frequency matching by year of death and age at death. Adjusted for age, calendar year, social class and race
Graves et al. Case–control. 1999. USA‐Seattle36	Both sexes. 1987–92. Direct interview only to the cases and control spouses	From the same study population base as in ref 27. Only people who have spouses as informants who agreed to collaborate, were considered eligible (89 population cases). 89 Population controls free of dementia.	NINCDS‐ADRDA probable	After data collection, IHs scored each job from detailed occupational history for potential exposure to EMF. Exposures were also classified according to duration and intensity	Matching by age, sex and source of information. Adjusted for age and education
Chandra et al. Case–control. 1987. USA‐Denver13	Both sexes. 1975–85. Structured interview through standardised questionnaire applied to the next of kin of both patients and controls	64 Hospital cases and 64 non‐demented hospital controls	NINCDS‐ADRDA probable	Question on “ever exposure” to some metals was included. Specific exposure to lead was collected in the questionnaire	Matching by age, sex, race and type of proxy
Gauthier et al. Case–control. 2001. Canada‐Quebec37	Both sexes. Ascertainment period. not specified Interview structured through standardised questionnaire to the next of kin of both patients and controls	68 Population cases were matched with 68 non‐demented population controls	Screening with 3MS, DSM IV, ICD‐10 and NINCDS‐ADRDA possible or probable	IH assess exposition to pesticides from detailed occupational history. Cumulative exposures were calculated	Matching by age and sex. Adjusted for education level, presence of family cases of AD, and presence of at least one ApoE epsilon4 allele
CSHA** et al. Case–control. 1994. 10 Canadian provinces26	Both sexes. 1991–2. Questionnaire completed by the proxies themselves (usually a close relative) both cases and controls. In seven centres an interviewer administered it	258 Cases with onset of symptoms within 3 years of diagnosis, and 535 population controls confirmed to be cognitively normal	Screening with 3MS. DSM III‐R and NINCDS‐ADRDA probable	Methods to measure exposure to specific agents (pesticides, solvents) through questionnaire are not defined	Frequency matching by study centre residence in community or institution, and age group. Adjusted for age, sex, residence in community or institution, and education
O'Flynn et al. Case–control. 1987. England and Wales24	Men. 1970–9. From death certificates	557 Cases who died of “presenile dementia” were randomised and compared with the same number of controls	Cases were further selected in order to exclude dementias other than AD	The person's most recent full time paid employment as reported to the Registrar at the time the death was registered. Occupations were graded by one of the investigators and an IH into one of three categories according to probable exposure to organic solvents and to lead”	Matching by age and sex
Graves et al. Case–control. 1998. USA‐Seattle32	Both sexes. 1987–92. Direct interview only with the cases and control spouses	From the same study population base as in ref 27. Only people who had spouse informants who agreed to collaborate, were considered eligible (89 population cases). 89 Non‐demented population controls	NINCDS‐ADRDA probable	An IH scored each job from detailed occupational history for potential exposure to aluminium and 5 types of solvent. Exposures were also classified according to duration and intensity	Matching by age, sex and source of information. Adjusted for age and education
Feychting et al. Case–control. 1998. Sweden34	Both sexes. 1989–91. Direct interview of cases' relatives (most often spouse or adult offspring) and of controls through structured questionnaire	From a cohort of twins taken by a register‐based sample of twins, 55 cases were identified (only a case when more than one twin demented). Cases were compared with non‐demented twins controls in two groups of 228 and 238 people	Screening with MMSE. DSM‐III‐R and NINCDS/ADRDA possible and probable	Occupations were linked to a JEM for exposure to EMF. Investigators had account of each subject's primary occupation, the last occupation in the person's work life, and the occupation with the highest magnetic field exposure. Exposures were categorised in three levels	Adjusted for age, sex and education
Noonan et al. Case–control. 2002. USA‐Colorado39	Men. 1987–96. Death certificate data were collected from the Vital Statistics Unit of the Colorado Department of Public Health	1556 Cases older than 60 years, were identified and compared with the same number of controls who died of other causes: leukaemia, brain cancer, or breast cancer	ICD‐9, code 331.0	Exposure to EMF was assessed from primary occupation with three different methods: Electrical/no electrical occupation, according to 4 levels with a different probability of exposure, and according to different exposure levels given by a JEM	Frequency matched by 5‐year age intervals and year of death. Adjusted for age, race and occupational grouping
Sobel et al. 1996. Case–control. USA‐California29	Both sexes. Ascertainment period not specified. Data were collected from the ADDTC at RLAMC	A clinical series of 326 cases who were at least age 65 at the time of their first examination at RLAMC were compared with 152 controls who were cognitively impaired or presented dementia other than vascular dementia	NINCDS‐ADRDA at least probable	Primary occupation was obtained from hospital records. The same method as previous investigation15 was used to measure exposure to EMF in high/low risk	Adjusted in men for sex and age at onset of symptoms. In women adjusted for education too
Shalat et al. 1988. Case–control. USA‐Bedford14	Hombres. 1975–85. Questionnaire completed by the next of kin themselves for both cases and controls	98 Cases obtained from hospitals were compared with 162 population controls	DSM III, NINCDS‐ADRDA (complete description of the diagnostic procedure reported previously)	Exposure to organic solvents and lead was assessed through specific items from the questionnaire and a detailed occupational history. Three IHs assigned likelihood and a semiquantitative level of exposure	Matched for sex, year of birth and town of residence. Adjusted for years of education
Li et al. Case–control. 1992. China25	Both sexes. 1988–9. Direct interview to surrogate informants of both cases and controls through structured questionnaire	70 Cases (54 hospital and 16 population cases) were compared with 140 non‐demented controls (neighbours)	Screening with MMSE. ICD‐10, NINCDS‐ADRDA possible and probable	Method to assess exposure to painting/other organic solvents through questionnaire is not defined	Matched by age and sex. Priority was given to those living closest to the matched patient. Adjusted for solvents is made, but confounding variables are not described
French et al. Case–control. 1985. USA‐Minneapolis12	Men. 1979–82. Direct interview through structured questionnaire or by telephone (6% of completed interviews) to surrogates respondents (usually next of kin) of both cases and controls	78 Hospital cases. 76 hospital controls and 48 neighbourhood controls. Controls with psychiatric disorders, CNS disorders, and alcoholism were excluded	NINCDS‐ADRDA probable. (Histologically confirmed in a subset of study subjects)	Method to assess exposure to pesticides, solvents and lead through questionnaire is not defined	Matching by age, sex and race
Gun et al. Case–control. 1997. Australia30	Both sexes. 1986–9. Direct interview through structured questionnaire to proxy respondents of both cases and controls	170 Hospital cases were compared with 170 population controls	NINCDS‐ADRDA possible and probable	Exposure to solvents, aluminium, lead and pesticides was assessed by a panel of three IHs, using occupational histories and a JEM. Cumulative exposures were calculated	Matching by age and sex. Adjusted for sex, education, family history of AD, early‐ or late‐onset AD cases, and possible versus probable AD cases
Heyman et al. Case–control. 1984. USA‐Durhan11	Both sexes. Ascertainment period unspecified. Direct interview through structured questionnaire to proxy respondents of both cases and controls	46 Hospital cases were compared with 92 population controls free of dementia	Similar to NINCDS‐ADRDA. (Histologically confirmed in a subset of study subjects)	Exposure to solvents and lead was directly assessed through an item from questionnaire which asked about “ever exposure” of at least 10 hours at week at least during 6 months in any occupation of the cases or controls	Matched for sex, race and 5‐year age intervals
Sobel et al. Case–control. 1995. Finland and USA‐California15	Both sexes. Series 1: 1982–5, series 2: 1977–8, series 3: 1984–93. Direct interview through structured questionnaire to proxy respondents of cases, but direct interview to controls	Three clinical series analysed globally and independently with different types of controls: 53, 198, 136 cases were compared with 70, 299, 136 controls, respectively	NINCDS‐ADRDA, or similar to NINCDS‐ADRDA	Exposure to EMF was evaluated by an IH from primary occupations. Exposures were also classified according to intensity in high/low level	Adjusted for age at onset, age at examination, sex, education and social class
Palmer et al. Case–control. 1998. England and Wales31	Men. 36–38 months. Short postal questionnaire either to the patient himself, or if he had died, to the next of kin	From CT records, 204 dementia cases (105 AD) were identified, who were compared with 225 controls with brain cancer and 441 controls with other neurological diseases like cerebrovascular disease, benign tumours, migraine or headache	From CT records of neuroradiology centres. Clinical diagnosis not specified	To assess exposure to solvents only occupations for more than 1 year recalled in the questionnaire were examined. These occupations were linked to a classification of occupations by likely exposure to organic solvents into three levels (high exposure, intermediate or uncertain exposure and low exposure)	Adjusted for age at the CT, neuroradiology centre, and distance of residence at the time diagnosis from the neuroradiology centre

*Papers are listed in decreasing order of their Global Quality Index, see explanation in the text.

†ADDTC, Alzheimer Disease Diagnosis and Treatment Centre; RLAMC, Rancho Los Amigos Medical Centre.

‡AD, Alzheimer's disease; CNS, central nervous system; CT, computed tomography.

§ICD‐9 and ICD‐10: criteria for Alzheimer's disease from the International Classification of Diseases, 9th and 10th revisions; NINCDS‐ADRDA: criteria for Alzheimer's disease from the National Institute of Neurologic and Communicative Disorders and Stroke and the Alzheimer's Disease and Related Disorders Association; DSM III‐R and DSM‐IV: criteria for Alzheimer's disease from the Diagnostic and Statistical Manual for mental disorders, revised 3rd and 4th editions; MMSE, Mini‐Mental State Examination; 3MS, modified Mini‐Mental State Examination.

¶EMF, electromagnetic fields; IH, industrial hygienist; JEM, job exposure matrix.

**CSHA, Canadian Study of Health and Aging investigators.