Heavy Smoking in Midlife and Long-term Risk of Alzheimer Disease and Vascular Dementia

Although it is well established that smoking increases the risk of several diseases, there has unfortunately been a recent increase in smoking among young adults in some developed countries.¹ Current estimates suggest there are several million deaths attributable to smoking and markedly increased cardiovascular- and cancer-associated mortality rates.² Although smoking increases risk of most diseases and death, some studies suggest that it is associated with a lower risk of certain neurodegenerative diseases such as Parkinson disease.³ The link between smoking and risk of Alzheimer disease (AD), the most common subtype of dementia, has been somewhat controversial, with some studies suggesting that smoking reduces the risk of cognitive impairment.⁴

Because of the long preclinical phase of AD, it is important to identify exposure to putative risk factors in middle age, before pathologic features of AD have started to develop in the brain.^5-7 However, there are relatively few large population-based cohort studies that have studied the relationship between smoking in early life and development of dementia later in life. Most smoking and dementia studies were conducted in elderly cohorts with a relatively short follow-up time (2 to 7 years).^8-14 The few studies that have examined midlife smoking have been conducted in relatively homogenous populations. One study that specifically focused on the amount of midlife smoking found that the risk of AD increased along with the level of smoking,¹⁵ while another study found that smoking status in midlife was associated with increased risk of dementia and AD especially among apolipoprotein E (APOE) ε4 carriers.¹⁶ A third study found a nonsignificant positive association between smoking in midlife and risk of hospitalization for dementia.¹⁷ Yet, both smoking and AD rates vary by race, and some studies suggest that the prevalence of dementia, as well as smoking, is higher in African American populations.^18,19 Thus, further evaluation of smoking and risk of dementia is warranted in large, ethnically diverse populations.

The association between smoking and different subtypes of dementia beyond AD is also a less investigated issue. Smoking is known to increase the risk of stroke,²⁰ and consequently, may also predispose to VaD. However, few previous longitudinal studies have investigated this association. To date, only 1 study, to our knowledge, has evaluated the long-term association of midlife smoking on risk of VaD.¹⁵ Other studies that have shown a positive association have had elderly cohorts with short follow-up times (approximately 2 years).^9,14

The goal of this study was to investigate the association between the amount of midlife smoking and long-term risk of all-cause dementia, AD, and VaD 2 to 3 decades later in a large multiethnic cohort of health plan members.

The study cohort includes members of the Kaiser Permanente Medical Care Program of Northern California, who participated in a voluntary health examination called the Multiphasic Health Checkup (MHC) in San Francisco and Oakland during 1978 through 1985, when they were 50 to 60 years old (N = 33 108). Kaiser Permanente of Northern California is a nonprofit, integrated health maintenance organization that covers more than one-fourth of the population in the geographic areas served. The members are representative of the sociodemographics of the local population.²¹ The MHC was given as part of routine medical care at all San Francisco and Oakland medical clinics. Our analytic cohort comprises 21 123 people who were still alive and members of the health plan in 1994, when electronic medical record diagnoses of dementia became available. People who had missing data on the smoking questionnaire (n = 1045) were excluded.

The purpose of the MHC was to collect a large amount of data on health habits and medical conditions of the participants, and it was given as part of routine medical care. It included a detailed interview on health behavior (including smoking), health status, and medical as well as family history. Participants were asked if they were never, former, or current smokers. For current smokers, amount smoked per day was asked with the following categories: smoking less than 0.5 pack per day, 0.5 to 1 pack per day, 1 to 2 packs per day, and more than 2 packs per day. In addition, several clinical measurements, including height, weight, and systolic and diastolic blood pressure (BP) of the participants, were carried out, and body mass index (BMI) was calculated as weight in kilograms divided by height in meters squared. A blood sample was drawn for determining the serum cholesterol level. High cholesterol was defined as a total serum cholesterol level of 240 mg/dL or greater (to convert cholesterol to millimoles per liter, multiply by 0.0259). The participants were considered to have hypertension if they had 1 of the following: self-report of physician-diagnosed hypertension, use of antihypertensive medication, systolic BP of 140 mm Hg or higher, or diastolic BP of 90 mm Hg or higher. Diabetes was defined by self-report of physician-diagnosed diabetes, use of insulin or oral hypoglycemic agents, a fasting (last food eaten in ≥8 hours) glucose level of 140 mg/dL or higher, or a nonfasting (last food eaten in ≤4 hours) glucose level of 200 mg/dL or higher (to convert glucose to millimoles per liter, multiply by 0.0555). Stroke and cardiovascular disease were recorded from hospital discharge diagnoses (International Classification of Diseases, Ninth Revision [ICD-9 ] codes for ischemic stroke, 433-438; hemorrhagic stroke, 430-432; and cardiovascular disease 410, 411, 413, 414, 428, 440, 443, and V717) from 1978 through the end of the study. Final multivariate models were adjusted for both stroke during dementia follow-up (stroke from 1994 to 2008) or intercurrent stroke (stroke between baseline and the start of dementia ascertainment from 1978 to end of 1993). The study was approved by the internal review board of Kaiser Permanente.

The dementia diagnoses were derived from electronic health records. The diagnoses considered in this study include both AD (n = 1136) (ICD-9, Clinical Modification [ICD-9-CM ] code 331.0) and VaD (n = 416) (ICD-9-CM code 290.4), which were made by a neurologist or neuropsychologist. General dementia diagnoses (n = 5367), which additionally includes diagnosis of unspecified dementia (ICD-9-CM code 290.0), were made by an internal medicine physician. The diagnoses were ascertained from January 1, 1994, to July 31, 2008. The mean (SD) age in 1994, at the onset of dementia follow-up was 71.6 (5.8) years.

All statistical analyses were carried out using SAS version 9.1 software (SAS Institute, Inc, Cary, North Carolina). The sociodemographic and clinical characteristics of the participants were compared using the χ² test for categorical variables and analysis of variance for continuous variables. The Cox proportional hazards model with age as a time scale was used to investigate the relationship between midlife smoking and dementia, AD, and VaD. Participants were censored according to age at dementia diagnosis, age at date of death, age at date of end of Kaiser membership (as defined by a lag in health plan membership of ≥90 days), or age at end of follow-up (July 31, 2008). Individuals with a dementia diagnosis other than the one under investigation were excluded from the analyses regarding AD or VaD separately (AD models n = 17 360 and VaD models n = 16 294). For cases, the calculated person-years in this study included age (as a time scale) in January 1, 1994, until the age at diagnosis of dementia, AD, or VaD and, for controls, until age at death, age at lag in health plan membership, or age at end of follow-up (July 31, 2008). Incidence rates were determined specifically for smoking categories. Age-adjusted incidence rates were calculated using the whole cohort as the standard population with 4 age groups including younger than 77 years, 77 to 81 years, 81 to 85 years, and older than 85 years.

For the Cox proportional hazard model analyses, smoking was defined as a categorical variable with the following 6 levels: never smokers, former smokers, and current smokers who smoke less than 0.5 pack per day, 0.5 to 1 pack per day, 1 to 2 packs per day, and more than 2 packs per day. In the Cox models, the never smokers were used as a reference group. The models were first adjusted for the potential sociodemographic confounders age (as a time scale), sex, education (categorized as high school, trade school, college 1-2 years, college 3-4 years, and postgraduate, with grade school as a reference), race (entered as white, Asian, and other, with African American as a reference group), and marital status (classified as never married and divorced/widowed/separated, with married as a reference group). Second, the models were also adjusted for midlife BMI, hyperlipidemia, diabetes, hypertension, heart disease, and stroke (yes/no) during the follow-up. Finally, the analyses were additionally adjusted for midlife alcohol drinking (classified as former, occasionally, 1-2 drinks per day, and >3 drinks per day, with never drinkers as a reference group). To examine whether the association between smoking and dementia, AD, or VaD was modified by sex or race, interaction terms were added to the Cox models.

A total of 5367 people (25.4%) were diagnosed as having dementia (including 1136 AD cases and 416 VaD cases) during a mean follow-up period of 23 years. Sociodemographic and clinical characteristics of the study population are given in Table 1. As expected, those who were diagnosed as having dementia were older, had fewer years of education, and were more likely to be women than the individuals without dementia. There were differences in the occurrence of dementia between race groups: compared with whites, African Americans were more likely to have a dementia diagnosis, while Asians were less likely. In addition, a higher percentage of divorced, widowed, or separated persons and a higher percentage of never drinkers were found among those with a dementia diagnosis. Dementia was also associated with a greater likelihood of a higher mean midlife BMI and all comorbidities.

Although both the incidence rates and age-adjusted incidence rates did not increase in a linear fashion by smoking levels, a dramatic increase in the incidence of dementia was found for those individuals who reported smoking 2 or more packs per day at midlife (Table 2). The calculated incidence rate of dementia for those persons was 312.21 (95% confidence interval [CI], 235.11-389.30), and the age-adjusted incidence rate was 786.42 (95% CI, 481.23-1091.61) per 10 000 person-years.

Smoking more than 2 packs daily in midlife increased the later risk of dementia (hazard ratio [HR], 2.01; 95% CI, 1.57-2.58) in the unadjusted Cox proportional hazards model (adjusted for age as a time scale, Table 2). In the fully adjusted multivariate model (adjusted for age as a time scale, sex, education, race, marital status, hypertension, high cholesterol level, BMI, diabetes, heart disease, stroke, and alcohol drinking), using the nonsmokers as a reference group, the HR (95% CI) was 2.14 (1.65-2.78) for those smoking more than 2 packs per day, 1.44 (1.26-1.64) for 1 to 2 packs per day, and 1.37 (1.23-1.52) for 0.5 to 1 packs per day. Former smoking (HR, 1.00; 95% CI, 0.94-1.07) or smoking less than 0.5 pack (HR, 1.04; 95% CI, 0.91-1.20) per day did not have an effect on dementia risk (Figure, A). Those who previously smoked or who smoked less than 0.5 pack per day had a dementia risk similar to nonsmokers. An additional fully adjusted multivariate model was conducted to control only for stroke that occurred between baseline and the start of dementia assessment (intercurrent stroke) to consider the potential confounding or mediating effects of stroke, which occurred prior to dementia, on the smoking-dementia association. These results were not markedly different than the prior model: compared with never smokers the HR (95% CI) was 2.12 (1.64-2.75) for those smoking more than 2 packs per day, 1.43 (1.26-1.63) for 1 to 2 packs per day, and 1.36 (1.22-1.51) for 0.5 to 1 pack per day. Former smoking (HR, 1.00; 95% CI, 0.94-1.07) or smoking less than 0.5 pack (HR, 1.05; 95% CI, 0.91-1.20) was not associated with risk of dementia.

There was an increased risk of AD also observed for those smoking more than 2 packs daily in midlife (crude model HR, 2.36; 95% CI, 1.54-3.61;Table 2) compared with those not smoking. In the fully adjusted Cox model, the association remained the same (HR, 2.57; 95% CI, 1.63-4.03; Figure, B). For those smoking 1 to 2 packs (HR, 1.18; 95% CI, 0.92-1.52) or 0.5 to 1 pack (HR, 1.11; 95% CI, 0.90-1.36) per day, there was an increase in risk, although nonsignificant, whereas former smoking (HR, 1.00; 95% CI, 0.89-1.13) or smoking less than 0.5 pack (HR, 0.80; 95% CI, 0.61-1.06) per day was not associated with a risk of AD compared with nonsmokers.

Despite the smaller number of cases, an increased risk of VaD later in life was also detected among midlife smokers (crude model for smoking >2 packs daily, HR, 2.02; 95% CI, 0.90-4.55;Table 2). Even after controlling for various potential vascular confounding factors, in the fully adjusted Cox model, those smoking more than 2 packs per day in midlife were almost 3 times (HR, 2.72; 95% CI, 1.20-6.18) more likely to develop VaD later in life than the nonsmoking individuals (Figure, C). The observed risk increase was borderline significant for those smoking 1 to 2 packs (HR, 1.42; 95% CI, 0.95-2.13) per day. Former smoking (HR, 0.99; 95% CI, 0.80-1.22), smoking less than 0.5 (HR, 1.05; 95% CI, 0.69-1.61) or 0.5 to 1 packs (HR, 1.20; 95% CI, 0.84-1.70) per day in midlife did not increase the later risk of VaD compared with nonsmokers.

There were no significant interaction terms of smoking × sex or smoking × race (P > .05). Post hoc stratified analyses did not reveal any trends that showed differences in the association between smoking and dementia by race or sex.

Stroke was both highly associated with smoking (χ² test, P < .001) and dementia risk (P < .001), although it did not significantly interact with smoking on risk of dementia, AD, or VaD (P > .05 for all interaction term models of smoking × stroke on risk). Although stroke (whether intercurrent or during follow-up) did not confound or attenuate the association between smoking and dementia risk, we wanted to examine the association between smoking and dementia risk separately for those with and without a stroke because stroke is a robust predictor of dementia and highly associated with smoking. Heavy smoking is a risk factor for dementia, AD, and VaD even among those without a stroke or without an intercurrent stroke (Table 3). These post hoc subgroup analyses should be interpreted with caution, since there were no statistically significant interaction terms of stroke times smoking on dementia risk.

To our knowledge, this is the first study to investigate the long-term association between the amount of smoking in midlife and the risk of dementia, AD, and VaD later in life in a large, multiethnic cohort. The risk of dementia, AD, and VaD was dose dependent in that the risk increased with the increasing amount of smoked cigarettes. Indeed, very heavy smokers (those who reported smoking >2 packs per day in middle age) were at the greatest risk of dementia even decades later in life.

The association between smoking specifically in middle age and the long-term subsequent risk of dementia later in life has been addressed previously in only 3 cohort studies with smaller sample sizes of predominantly white participants.^15-17

The majority of previous studies on smoking and dementia have evaluated the association between smoking and short-term risk of dementia, though most of these studies calculated pack-years for smoking.^9,13,14 Our findings are consistent with the studies indicating that heavier smoking is associated with a greater dementia risk,^9,13-15 although some studies found no association.¹²

Evaluation of smoking in middle age is important, since it is possible that subclinical dementia or cognitive impairment can affect behavior or induce bias owing to falsely recalled information. This can be of particular importance if smoking information is collected among an elderly population. In the present study, the participants had a mean age of 58 years at the time of the baseline examination, so it is highly unlikely that subclinical dementia would have biased our results. In addition, the baseline examination conducted in midlife and the long follow-up time enables us to evaluate smoking habits before neuropathologic changes have likely commenced in the brain, and thereby, to investigate smoking more specifically as a true long-term “risk factor,” rather than a “risk predictor,” for dementia.

In this study, midlife smoking was associated with an increased risk of both AD and VaD. The relationship between smoking and VaD has not been as widely investigated as the relationship with AD. Smoking is a well-established risk factor for stroke,²⁰ and consequently, it can also predispose to multi-infarct dementia. Interestingly, in the present study the association between smoking and VaD remained significant after controlling for various potential vascular confounding factors (including stroke); thus, smoking seems to also have some independent effect on VaD, beyond acceleration of cerebrovascular disease. These mechanisms, as well as the mechanisms through which smoking makes one susceptible to AD, need to be further clarified in future studies. It is well recognized that smoking augments oxidative stress and inflammation, which are also believed to be important pathophysiologic mechanisms in AD.²² It is possible that smoking affects the development of dementia via vascular and neurodegenerative pathways.

In this large and diverse cohort with several ethnic groups, as well as both sexes, no interaction between smoking and sex or between smoking and race was found. This is an important finding because, despite that the incidence of dementia is suggested to vary by race,¹⁸ previous studies of putative risk factors for dementia have mainly been conducted within white cohorts. The present study is generalizable to elderly populations at risk for dementia. Based on the present results, we can postulate that the deleterious effects of smoking on risk of dementia seem to be the same for both sexes and across different ethnic groups.

One obvious limitation of this study is the definition of dementia diagnoses, which were obtained from medical records. Although the diagnoses were not systematically and clinically determined according to a strict protocol, the dementia subtypes (AD and VaD) were diagnosed by a neurologist or neuropsychologist in a memory clinic according to common clinical practice. The diagnoses of VaD in this study refers to multi-infarct dementia, which is more straightforward to diagnose than AD or mixed dementia. Therefore, the diagnoses of VaD may be considered reliable, while the AD group may also include people with mixed dementia, since AD is more likely to be diagnosed in a person having memory impairment as a leading symptom. Furthermore, there might be some subjects with undiagnosed dementia in the cohort and, in addition, some AD and VaD cases might have been missed in participants who had died prior to the onset of the ascertainment in 1994. Thus, some selective survival effect cannot be ruled out regarding the present results. Yet, smoking is associated with increased mortality,²³ and accordingly, if it is assumed that among the deceased there were more smokers and that the deceased were also more likely to have dementia, then the present results would not overestimate but rather underestimate the true effects of smoking on the risk of dementia later in life. Another limitation is that the smoking data were collected at midlife only, and thus the present study cannot evaluate whether smoking cessation is associated with reduced dementia risk.

To our knowledge, this is the first study evaluating the amount of midlife smoking on long-term risk of dementia and dementia subtypes in a large multiethnic cohort. Our study suggests that heavy smoking in middle age increases the risk of both AD and VaD for men and women across different race groups. The large detrimental impact that smoking already has on public health has the potential to become even greater as the population worldwide ages and dementia prevalence increases.

Correspondence: Rachel A. Whitmer, PhD, Division of Research, Kaiser Permanente, 2000 Broadway, Oakland, CA 94612 (Rachel.Whitmer@kp.org).

Accepted for Publication: August 19, 2010.

Published Online: October 25, 2010. doi:10.1001/archinternmed.2010.393

Author Contributions:Study concept and design: Quesenberry and Whitmer. Acquisition of data: Quesenberry, Zhou, and Whitmer. Analysis and interpretation of data: Rusanen, Kivipelto, Quesenberry, Zhou, and Whitmer. Drafting of the manuscript: Rusanen, Zhou, and Whitmer. Critical revision of the manuscript for important intellectual content: Rusanen, Kivipelto, Quesenberry, and Whitmer. Statistical analysis: Quesenberry and Zhou. Obtained funding: Rusanen and Whitmer. Administrative, technical, and material support: Quesenberry and Whitmer. Study supervision: Kivipelto and Whitmer.

Financial Disclosure: Dr Kivipelto received honoraria for serving on the scientific advisory board of Elan and Pfizer and serves as a speaker on scientific meetings organized by Janssen, Novartis, and Pfizer.

Funding/Support: This study was supported by The National Graduate School of Clinical Investigation, EVO grants from Kuopio University Hospital, and grants from the Juho Vainio Foundation and Maire Taponen Foundation. This study was also supported by a Kaiser Permanente Community Benefits Grant and National Institute of Health and Academy of Finland Grant 120676.

Role of the Sponsors: The funders had no role in study design or in the collection, analysis, and interpretation of data; in the writing of the report; or in the decision to submit the article for publication.