The influence of smoking on the risk of Alzheimer’s disease

Setting and participants.

Data were analyzed from a sample of Medicare recipients in three contiguous zip codes within the community of Washington Heights in northern Manhattan, New York City.

A total of 2,128 persons participated in the initial phase of the study. A 90-minute, in-person interview of general health and function was completed. This structured interview also included questions about years of formal education. The interview was followed by a standardized clinical assessment including a medical history, physical and neurologic examination, and a brief (approximately 1 hour) neuropsychological battery developed previously for use in this community.12,13 These same clinical assessments were used in the annual follow-up of all participants. The interviews were conducted in either English or Spanish. The Columbia University Institutional Review Board reviewed and approved this project. All participants provided written informed consent.

For this study data were excluded from 290 people (13.6%) who were found demented at the initial interview (using the methods described), 349 (16.4%) who refused subsequent follow-up, 155 (7%) who died after the initial examination, 129 (6.1%) missing smoking information, 25 (1.2%) individuals with PD, 117 (5.5%) with stroke, and 1 individual with both PD and stroke. All data from the remaining 1,062 elderly (730 women and 352 men) without dementia were considered appropriate for the investigation. Annual follow-up examinations were conducted over an average of 2.04 years.

AD diagnosis.

All participants were examined by a neurologist or an appropriately trained internist or psychiatrist, and all underwent neuropsychological testing. The data from the initial and follow-up examinations and interviews, and any existing medical records and imaging studies were reviewed collectively at a consensus conference of physicians and neuropsychologists to establish diagnoses. Clinicians were unaware of the APOE genotype during the diagnostic process. The diagnosis of dementia or the specific clinical diagnosis of AD was based on specified research criteria14,15 and required evidence of cognitive deficit on clinical evaluation in addition to objective neuropsychological findings, as well as evidence of impairment in social or occupational function. Patients meeting criteria14 for probable or possible AD with a Clinical Dementia Rating16 score of 1.0 or higher were considered to have the clinical diagnosis of AD. To exclude the diagnosis of mixed dementia (i.e., AD and vascular dementia), cerebrovascular disease was identified by a clinical history of stroke or TIA, or by evidence of cerebrovascular lesions on brain imaging. All those with any history of cerebrovascular disease or mixed dementia were excluded from the study.

Ethnic group.

For ethnic group classification we used the format suggested by the 1990 United States Census Bureau.17 In the first question, participants were asked to which ethnic group they belonged. The possible responses included white, black (African American or Caribbean American), American Indian, Asian, or other. Regardless of the response, a second question asked whether or not the person being interviewed considered him or herself Hispanic. Several choices were then provided to indicate specific Hispanic origin (e.g., Puerto Rico, Dominican Republic, Cuba). Each individual was also asked to identify the country of their birthplace. For this study, we used the categories of black (non-Hispanic), white (non-Hispanic), and Hispanic consistent with the census data.

Risk factor interview.

A structured risk factor questionnaire was developed for the assessment of exposures to putative risk factors related to dementia.18 The interviews were given in English and Spanish according to the preference of the patient.

For smoking there were three sets of questions. A trigger question asked whether or not the individual ever smoked at least one cigarette per day for a period of 1 year or more. If the answer was no, no further questions were asked. If the subject answered yes, the individual was asked at what age they began smoking, whether they were still smoking, at what age they had stopped if no longer smoking, and how many cigarettes on average they had smoked or still smoked per day.

The two other sets of smoking questions asked about cigar and pipe use. Both began with a trigger question asking whether the individual had ever smoked a cigar or pipe. If the subject answered no, no further questions were asked. If the subject answered yes, follow-up questions similar to those for cigarettes were asked but referred to the number of cigars or pipe-fulls smoked daily.

A composite vascular “heart” disease variable was formulated to assess whether cardiac disease may act as a possible confounder of AD. The vascular disease variable included anyone reporting a history of myocardial infarction, congestive heart failure, angina, and anyone taking cardiac medication.

APOE genotyping.

The APOE genotype for each participant was determined after isolating DNA from white blood cells and digesting the DNA with HhaI.19 Leukocyte DNA was amplified by PCR using specifically synthesized oligonucleotide primers and Taq polymerase modified from those described by Hixson and others.20,21 The digested PCR fragments were then separated by electrophoresis.

Data analysis.

The frequencies for each demographic variable (age, gender, education, and ethnic group) and for smoking were compared among those at-risk participants and those with AD using chi-square analyses and Fisher’s exact tests.22

Risk ratio estimates were calculated to assess the relationship between AD and smoking, in addition to other demographic variables. Univariate measures of association were estimated from chi-square analysis. Multivariate risk ratios were estimated from Cox proportional hazard models for each risk factor.23 As recommended by Korn et al.,23 the time-to-event variable was the age at onset of AD; therefore, no further age adjustment was required. The first stage of assessment included only smoking history. We then included education, and the final model included education in addition to ethnicity, and APOE genotype. This information was used to identify differences with respect to education and ethnic group identity represented in the community.

Demographics.

Table 1 shows the demographic characteristics of the study participants by AD and nondemented control subjects. Obvious group differences were apparent for age at interview and education. Patients with AD were significantly older than the control subjects (χ² = 25.0, p < 0.001) and had the least amount of formal education (χ² = 26.4, p < 0.001), whereas patients with AD were very similar to control subjects in ethnic group representation and smoking status. ⇓

Reliability.

The test–retest reliability18 of the risk factor interview was in the good to excellent range for the questions regarding smoking (κ = 0.72).

Smoking.

The frequency of smoking varied significantly by age and gender, but not by ethnic group or education (table 2). Smoking decreased with age, and varied by sex in that more women reported that they never smoked. There was no difference in tobacco use across ethnic groups. The majority of smokers were cigarette smokers. Only four patients with AD and five nondemented subjects restricted their smoking to cigars or a pipe. A few in each group smoked cigars or a pipe as well as cigarettes according to our definition. We did not perform separate analyses for those who use only cigars or pipes because these were relatively infrequent as exposures. However, we did include cigar and pipe smokers in comparing smokers with nonsmokers. Among those who quit smoking, the mean time since cessation was 21.4 years. Because it was possible that smoking might have delayed or hastened the onset of AD, we compared the age at onset of symptoms between smokers and nonsmokers with AD. The mean age at onset of symptoms for smokers with AD was significantly lower than for nonsmokers (nonsmokers, 80.9 ± 5.9 versus smokers, 76.5 ± 4.9;p < 0.001).

Alzheimer’s disease.

Although there were no statistically significant differences in the risk of developing AD between current smokers, past smokers, and those who reported never smoking on univariate analysis (χ² = 5.5, p < 0.06), there were slightly more current smokers among the subjects (see table 1). Multivariate analysis showed that a past history of smoking was not associated significantly with AD (relative risk [RR] = 0.7; 95% CI, 0.5 to 1.1). However, those who were current smokers had an increased risk of developing AD as indicated by the unadjusted RRs and CIs in table 3 (RR = 1.9; 95% CI, 1.2 to 3.0). Adjustment for education and ethnicity did not alter the measure of association between smoking and AD (RR = 1.7; 95% CI, 1.1 to 2.8). The association between smoking and AD was slightly lowered in the presence of APOE-ε4 (RR = 1.4; 95% CI, 0.6 to 3.3; table 4), but not for current smokers without APOE-ε4 (RR = 2.1; 95% CI, 1.2 to 3.7). Table 5 lists the distribution of smoking history by AD status, APOE status, and age group. ⇓

After reviewing these results additional questions arose concerning vascular disease acting as a possible confounder. Smoking is a risk factor for vascular disease. Cardiac disease increases the risk of vascular dementia. Therefore, we looked at two additional multivariate models (multiplicative and additive) to determine the likelihood of interaction. The multiplicative proportional hazard model included ethnic group, smoking history, and a composite vascular “heart” disease variable. In the multiplicative model there was no evidence of interaction among those who were both current smokers and had cardiac disease (current smokers: RR = 1.9; 95% CI, 1.1 to 3.3; heart disease: RR = 1.1; 95% CI, 0.72 to 1.78; current smokers × heart disease: RR = 1.1; 95% CI, 0.4 to 3.2). In the additive model, which included all of the variables just listed, there was the suggestion of interaction among those who were both current smokers and had cardiac disease (current smokers: RR = 1.9; 95% CI, 1.1 to 3.3; heart disease: RR = 1.1; 95% CI, 0.7 to 1.8; current smokers and heart disease: RR = 2.4; 95% CI, 0.97 to 5.7). However, the number of individuals in this group was small (n = 14), thus limiting the interpretation. Finally, stratification by heart disease did not alter the relationship between smoking and AD in those with and without APOE-ε4.