A Prospective Study of Chronic Obstructive Pulmonary Disease and the Risk for Mild Cognitive Impairment

Importance  Previous studies suggest cross-sectional associations between a diagnosis of chronic obstructive pulmonary disease (COPD) and mild cognitive impairment (MCI). However, few studies have assessed whether COPD, a potentially modifiable factor, is associated with an increased risk for MCI and whether the relation is specific to the type of MCI.

Objective  To investigate whether a diagnosis of COPD and duration of COPD are associated with an increased risk for incident MCI and MCI subtypes (amnestic MCI [A-MCI] and nonamnestic MCI [NA-MCI]).

Design, Setting, and Participants  A prospective population-based cohort from the Mayo Clinic Study on Aging. We included 1425 cognitively normal individuals aged 70 to 89 years who were randomly selected from Olmsted County, Minnesota, on October 1, 2004, using the medical records linkage system. At baseline and every 15 months thereafter, participants underwent assessment with a nurse interview, neurologic examination, and neuropsychological testing. A diagnosis of COPD was confirmed via medical record review. A baseline diagnosis of COPD and duration of COPD were examined as risk factors for MCI and MCI subtypes using Cox proportional hazards models and adjusting for demographic variables and medical comorbidities, with age as the time scale.

Exposure  A baseline diagnosis of COPD and duration of COPD.

Main Outcomes and Measures  Incident MCI, A-MCI, and NA-MCI.

Results  Of the 1425 participants with normal cognition at baseline, 370 developed incident MCI. The median duration of follow-up was 5.1 years (interquartile range, 3.8-5.4 years). A diagnosis of COPD significantly increased the risk for NA-MCI by 83% (hazard ratio, 1.83 [95% CI, 1.04-3.23]), but not of any MCI or A-MCI in multivariate analyses. We found a dose-response relationship such that individuals with COPD duration of longer than 5 years at baseline had the greatest risk for any MCI (hazard ratio, 1.58 [95% CI, 1.04-2.40]) and NA-MCI (2.58 [1.32-5.06]).

Conclusions and Relevance  A diagnosis of COPD is associated with an increased risk for MCI, particularly NA-MCI. We have found a dose-response relationship between COPD duration and risk for MCI. These findings highlight the importance of COPD as a risk factor for MCI and may provide a substrate for early intervention to prevent or delay the onset and progression of MCI, particularly NA-MCI.

Chronic obstructive pulmonary disease (COPD) is a progressive but potentially treatable and preventable disease characterized by chronic airflow limitation and associated with an abnormal inflammatory response of the lungs to noxious particles or gases.¹ Current surveillance data show that more than 13.5 million adults 25 years and older in the United States have COPD.² Chronic airflow limitation can result in hypoxemia and hypercapnia, which may predispose these patients to an increased risk for cognitive dysfunction.^3,4

Previous studies^4-6 suggest that COPD is associated with hypoxemia and cognitive impairment. Cross-sectional studies^7,8 have also reported a higher frequency of mild cognitive impairment (MCI) in individuals with compared with those without a diagnosis of COPD. In the absence of a curative therapy for dementia, the early identification of modifiable risk factors for MCI, the earliest symptomatic phase of dementia, is important for preventing or delaying the onset of cognitive impairment. One recent longitudinal study⁹ reported that persons with COPD and asthma were at increased risk for MCI/dementia. However, that study used self-reported diagnoses and did not examine MCI subtypes (amnestic MCI [A-MCI] and nonamnestic MCI [NA-MCI]), which may provide insights into the underlying causes of MCI. In the present study, we examined the association between medical record–confirmed diagnoses of COPD, duration of COPD, and risk for any MCI and MCI subtypes in individuals with normal cognition enrolled and followed up in the Mayo Clinic Study of Aging (MCSA).

The MCSA is a prospective, population-based study designed to identify the prevalence, incidence, and risk factors of MCI. The details of the study design have been published elsewhere.¹⁰ In brief, from an enumeration of Olmsted County, Minnesota, residents aged 70 to 89 years on October 1, 2004 (n = 9953), and identified using the Rochester Epidemiology Project (REP) medical records linkage system,¹¹ a sample of 5233 individuals was randomly selected by age and sex stratification. The study cohort underwent evaluation for eligibility using the following inclusion characteristics: residency in Olmsted County, absence of dementia (determined through medical record review by a behavioral neurologist), and not being terminally ill or in hospice. Of the 4398 eligible individuals, 2719 agreed to participate (2050 underwent evaluation in person and 669 via telephone interview). After excluding prevalent cases of MCI and individuals who died or dropped out before any follow-up, the present analysis included 1425 individuals with normal cognition at baseline (Supplement [eFigure]). The study was approved by the institutional review boards of the Mayo Clinic and the Olmsted Medical Center, Rochester, Minnesota. Written informed consent was obtained from all participants.

All participants were interviewed by a nurse or study coordinator, underwent a neurologic evaluation by a physician (B.S., A.K.P., Y.E.G., and R.C.P.), and completed neuropsychological testing administered by a psychometrist, as previously reported.¹⁰ The nurse interview included questions about memory to the participant and an informant using the Clinical Dementia Rating Scale.¹² The physician examination included a medical history review, a complete neurologic examination, and administration of the Short Test of Mental Status¹³ and the Unified Parkinson’s Disease Rating Scale.¹⁴ The neuropsychological battery consisted of 9 cognitive tests used to assess function in 4 domains (memory, language, executive function, and visuospatial skills). For the purpose of determining impairment for an MCI diagnosis, the raw scores on each test were adjusted for age and education and scaled using normative data from the Mayo’s Older American Normative Studies.¹⁵ Within each domain, the scaled test scores were summed and scaled to obtain the final global and domain-specific z scores.¹⁰ A domain-specific score less than 1.0 SD below the age-specific mean among the general population was considered possible cognitive impairment. A decision made about impairment in a cognitive domain took occupation into consideration. Normal cognition, MCI, or dementia was diagnosed according to published criteria and was based on a consensus agreement among the interviewing nurse, examining physician, and the neuropsychologist, taking into account all the information collected.^10,16 At follow-up visits, examiners were blinded to cognitive diagnoses from previous evaluations. The MCI cases were further classified into A-MCI or NA-MCI, depending on whether the memory domain was impaired.

We identified potential cases of COPD using 2 sources of information. We used automated digital algorithms¹⁷ and ascertained diagnoses through the REP medical records linkage system.¹¹

The automated digital algorithm is a highly sensitive automatic method of extracting comorbidities, including COPD, from the electronic medical records using Boolean combinations of clinical variables and natural language-processing data feeds.¹⁷ The implementation of an automatic note search strategy to extract COPD from the electronic medical records is advantageous in that it facilitates fast recognition of COPD cases with high sensitivity (>98%) and specificity (>99%).^17,18 However, some health care providers in Olmsted County do not have medical records that can be accessed by the digital algorithms. As a result, we also used codes from the International Classification of Diseases, Ninth Revision (ICD-9), and the Hospital International Classification of Diseases Adapted (HICDA) from the REP and manual medical record data abstraction to identify individuals with COPD.

The REP compiles the medical records and residency status of each person who visited any health care provider in Olmsted County since January 1, 1966; health care providers include the Mayo Clinic, the Olmsted Medical Center, and their affiliated clinics.¹¹ Using the REP records linkage system, we identified all MCSA enrollees with any of the following primary ICD-9 or HICDA codes indicative of possible COPD: 491.xx (chronic bronchitis), 492.xx (emphysema), or 496.xx (chronic airway obstruction, not elsewhere classified).¹⁹

After the identification of potential participants with COPD, their clinical records were reviewed to make a confirmatory diagnosis. Patients were determined to have COPD if the medical record included a physician diagnosis of COPD (eg, documented diagnosis of chronic bronchitis, emphysema, or COPD in the admission note, progress note, or discharge summary from the index hospitalization) and/or use of medication therapy for COPD (eg, β₂-agonists, anticholinergics, or methylxanthines). Of the 1425 individuals with normal cognition at baseline, 171 were confirmed to have a diagnosis of COPD. The κ values for the automated digital algorithms and identification of COPD cases using the REP codes were excellent (κ = 0.8, with 94% agreement).

To assess the reliability of abstracting information for a COPD diagnosis, 2 reviewers (B.S. and A.K.P.) independently abstracted the electronic medical records of 50 randomly selected participants (10 COPD and 40 non-COPD cases). The interreviewer agreement was also excellent (κ = 0.9, with 94% agreement).

Demographics (age, sex, and education) were assessed by interview at the baseline visit. Participants were asked to bring all medications to the study visit, and the names of the medications were recorded. Information related to smoking (former, current, or never), cardiovascular comorbidities, hypertension, coronary artery disease (angina, myocardial infarction, coronary revascularization, or coronary artery bypass grafting), and diabetes mellitus were abstracted from a review of the electronic medical records. Depressive symptoms were assessed using the Beck Depression Inventory II depression score.²⁰ Evidence of a stroke was assessed at baseline by the study physician and further validated using the medical records. The study coordinator measured height and weight for computing body mass index (calculated as weight in kilograms divided by height in meters squared). Genotyping of apolipoprotein E (APOE) alleles was performed using standard methods.

Continuous variables were reported as medians with the interquartile range, and categorical variables as counts with percentages. We examined differences in the baseline demographic and health-related characteristics between participants with and without MCI and with and without COPD using χ² tests for categorical variables and Wilcoxon rank sum tests for continuous variables. The date of onset of MCI was defined as the midpoint between the last evaluation when the subject had normal cognition and the first evaluation when the participant received a diagnosis of MCI. For the 16 subjects who progressed directly from normal cognition at one visit to dementia at the next visit, the date of onset of MCI was defined as the midpoint between the last evaluation when the participant had normal cognition and the first evaluation when the participant was diagnosed as having dementia. Participants who refused further participation, were lost to or unavailable for follow-up, or died were censored at their last evaluation.

We used multivariate Cox proportional hazards models to investigate the association between COPD and incident MCI and MCI subtypes, with age as a time scale, directly standardized by age and sex to the Olmsted County population on October 1, 2004, and adjusted for nonparticipation at baseline using reciprocal probability weighting in Poisson regression models. Results are reported as hazard ratios (HR) with 95% CIs. We evaluated 3 models, each building on the previous model. Model 1 adjusted for age, educational level, and sex. Model 2 controlled for the variables in model 1 plus Beck Depression Inventory II depression scores (as a categorical variable, <13 and ≥13) and history of stroke. Model 3 controlled for the variables in model 2 and included APOE genotype (any ε4 vs no ε4 allele), smoking (ever vs never), diabetes mellitus, hypertension, coronary artery disease, body mass index, and baseline global z score. The proportional hazards assumption for COPD at enrollment, COPD duration at baseline, and COPD duration levels defined at baseline (none, ≤5 years, and >5 years) were valid. Duration of COPD was dichotomized at the median of 5 years, and the same models were run. No significant interaction was found between age and COPD or between sex and COPD. All the statistical tests were performed at the conventional 2-tailed α level of .05 using commercially available software (JMP, version 9.0.1, and SAS; SAS Institute, Inc).

Of the 1425 cognitively normal subjects with at least 1 follow-up visit, 370 developed incident MCI. Of these, 230 (62.2%) had A-MCI, 97 (26.2%) had NA-MCI, 27 (7.3%) had MCI of unknown type, and 16 (4.3%) progressed from normal cognition at one visit to having dementia at the next (Figure). The median duration of follow-up was 5.1 (interquartile range, 3.8-5.4) years. Compared with participants included in the subsequent analyses, those who were lost to follow-up had lower educational levels (median of 12 vs 13 years; P = .02) but were similar in sex and frequency of stroke, cardiovascular comorbidities, diabetes mellitus, and COPD. At baseline, participants who developed incident MCI were significantly (P < .05; Table 1) older and less educated and had a higher frequency of stroke, diabetes mellitus, depression, and APOE ε4 genotype compared with those who retained normal cognition.

At baseline, 171 individuals had a COPD diagnosis (Table 2). Compared with individuals without a COPD diagnosis, those with the diagnosis were more often men and older, had a higher frequency of coronary artery disease, hypertension, and stroke, and were more likely to report former or current smoking (P < .05).

A diagnosis of COPD was associated with an increased risk for MCI (HR, 1.43 [95% CI, 1.07-1.91]) after adjusting for covariates in model 2, but the result was attenuated and of borderline nonsignificance after additional adjustment for variables in model 3 (1.33 [0.96-1.84]) (Table 3). When examining MCI subtypes, COPD was associated with almost a 2-fold risk for NA-MCI (HR, 1.83 [95% CI, 1.04-3.23]) in the fully adjusted model 3. We found no association of COPD with an increased risk for A-MCI.

We found a dose-response relationship between the duration of COPD and risks for any MCI and NA-MCI (Table 4). The overall HR for any MCI increased from 1.11 (95% CI, 0.70-1.74) in participants with a COPD duration of no longer than 5 years at baseline to 1.58 (1.04-2.40) in participants with a COPD duration of longer than 5 years. Similarly, the overall HR for NA-MCI increased from 1.14 (95% CI, 0.46-2.79) in participants with a COPD duration at baseline of no longer than 5 years to 2.58 (1.32-5.06) in participants with COPD for longer than 5 years. We did not observe a dose-response effect when examining duration of COPD and A-MCI.

In this population-based, prospective study of individuals 70 years and older, COPD was associated with an increased risk for NA-MCI and a risk for incident MCI of borderline nonsignificance. Further, we found a dose-response relationship such that the greatest risk for MCI and NA-MCI was among individuals with duration of COPD of longer than 5 years. These findings highlight the importance of COPD as a risk factor for MCI and may provide a substrate for early intervention or to treat COPD more aggressively at earlier points to prevent or delay the onset and progression of MCI, particularly NA-MCI.

Previous studies have shown that COPD is associated with poor performance in tests of attention, memory, and executive function, especially in patients with severe COPD.^5,21,22 However, only 2 cross-sectional studies^7,8 have investigated the association between COPD and MCI using standard criteria. One study⁸ compared the frequency of MCI among 45 patients with moderate to severe COPD and 50 healthy control individuals referred from an outpatient pulmonary clinic. The group with moderate to severe COPD had a higher frequency of MCI, especially NA-MCI, than the healthy controls. In the MCSA, a recent report⁷ also found a higher frequency of COPD in persons with MCI compared with individuals with normal cognition. However, only 1 previous longitudinal study from Finland⁹ examined whether COPD was associated with the risk for MCI. The authors reported that a diagnosis of COPD in midlife (ages 39-64 years) but not in late life (ages 65-80 years) was associated with an increased risk for MCI (HR, 1.85 [95% CI, 1.05-3.28]). In fact, they found a trend for individuals with COPD in late life to have a reduced risk for MCI. In contrast to that study, we found that a diagnosis of COPD was associated with an increased risk for MCI and NA-MCI and that the risk increased with COPD duration. Several potential explanations for the incongruent results exist. First, the prior study incorporated self-reported COPD, whereas we reviewed the medical records to confirm a diagnosis of COPD. Second, in the Finland study,⁹ investigators used a multistage evaluation such that only subjects with positive screening results underwent full evaluation for a diagnosis of MCI, which may have caused some individuals to be missed. In the MCSA, we performed a comprehensive evaluation for every subject, and the consensus-based approach provided diagnoses of MCI and MCI subtypes.²³ Last, the cognitive status of subjects in the Finland study⁹ was evaluated several years after baseline (>10 years), whereas in the MCSA, the study participants underwent evaluations at 15-month intervals. Chronic obstructive pulmonary disease is associated with an increased risk for mortality, especially in the elderly. Therefore, a long interval between the baseline and follow-up visits in the Finland study could have led to a survival bias and resulted in an inverse association between COPD and MCI.

The increased risk for NA-MCI in patients with COPD indicates a potential role of inflammation and vascular disease in the pathogenesis of NA-MCI. Patients with COPD have increased levels of systemic inflammatory markers, including interleukin 6, C-reactive protein, leukotriene B₄, tumor necrosis factor, and interleukin 8.^24-26 These inflammatory markers have been associated with cognitive impairment and NA-MCI.^27-29 Chronic obstructive pulmonary disease is also associated with an increased risk for cardiovascular disease,³⁰ and cardiovascular disease is a risk factor for MCI.^31-33 Even after adjustment for vascular diseases and factors, the association between COPD and MCI persisted, suggesting that COPD is an independent predictor of MCI risk.

The association between COPD and MCI risk could also be mediated by the increased hypoxic insults to the brain, which may lead to the generation of free radicals, inflammation, neuronal damage, and glial activation.³⁴ Furthermore, hypoxia in COPD could lead to impairment in executive tasks requiring attention allocation,³⁵ possibly explaining the association between COPD and NA-MCI in our study. Patients with COPD have altered⁶ and depressed cerebral perfusion³⁶ that, along with decreased oxygen saturation, could have a significant effect on cognitive function. Indeed, the deterioration in cerebral perfusion and cognitive performance has been shown to be greater in the patients with hypoxemic COPD compared with nonhypoxemic COPD.⁶ Plasma clusterin (ie, apolipoprotein J) may be another mediator of the COPD-MCI relationship. Clusterin levels have been shown to be elevated in patients with COPD³ and associated with cognitive decline.³⁷

This study has several strengths. First, the MCSA is a population-based, prospective cohort study, designed to investigate the association between COPD and MCI. The participants constituted an elderly cohort of individuals aged 70 to 89 years who were randomly selected using the REP medical records linkage system. Second, the diagnosis of MCI was made by a comprehensive clinical evaluation and by a consensus decision among the examining physician and nurse and a neuropsychologist, thus enhancing its reliability. Third, the investigators assessing the medical records for the identification of COPD cases were blinded to the diagnosis of MCI, thus avoiding diagnostic suspicion bias.

Limitations also warrant consideration. First, we defined COPD as physician diagnosed and did not base the definition on results of spirometry, which is the recommended diagnostic test for COPD but was not routinely available in our population. In the absence of spirometry, previous studies have shown that COPD, particularly mild COPD, is underdiagnosed in the population.^2,38 However, physician-diagnosed COPD may also overdiagnose COPD if not confirmed by spirometry findings.^39,40 Second, the Olmsted County population 70 years and older is primarily white and of European ancestry, potentially reducing the generalizability of the study results to other populations. However, the findings from the previous studies conducted in Olmsted County have been shown to be generalizable to the Upper Midwest population,⁴¹ thereby providing important information regarding various diseases, and are consistent with national data. Third, we did not have adequate data for smoking duration and could not adjust for years of smoking. Instead, we characterized smokers as those who ever vs those who never smoked at baseline. Controlling for smoking status had little effect on the models, suggesting that the association between COPD and MCI is independent of smoking status.

The diagnosis of COPD is associated with an increased risk for MCI, particularly NA-MCI. We have found a dose-response relationship between the duration of COPD duration and risk for MCI. Our findings highlight the importance of COPD as a risk factor for MCI and may provide a substrate for early intervention to prevent or delay the onset and progression of MCI, particularly the NA-MCI subtype.

Accepted for Publication: January 17, 2014.

Corresponding Author: Michelle M. Mielke, PhD, Division of Epidemiology, Department of Health Sciences Research, Mayo Clinic, 200 First St SW, Rochester, MN 55905 (mielke.michelle@mayo.edu).

Published Online: March 17, 2014. doi:10.1001/jamaneurol.2014.94.

Author Contributions: Drs Singh and Mielke had full access to all the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis.

Study concept and design: Singh, Mielke, Roberts, Geda, Petersen.

Acquisition, analysis, or interpretation of data: All authors.

Drafting of the manuscript: Singh, Mielke, Parsaik, Roberts, Petersen.

Critical revision of the manuscript for important intellectual content: All authors.

Statistical analysis: Singh, Mielke, Cha, Christianson, Pankratz.

Obtained funding: Petersen.

Administrative, technical, or material support: Petersen.

Study supervision: Mielke, Roberts, Scanlon, Petersen

Conflict of Interest Disclosures: Dr Mielke receives research support from the National Institute on Aging, National Institutes of Health (NIA/NIH) and has served as a consultant for Eli Lilly. Dr Roberts receives research support from the NIA/NIH and from Abbvie Health Economics and Outcomes Research. Dr Scanlon has participated as an investigator in clinical trials funded by Boehringer Ingelheim, GlaxoSmithKline, Forest, Novartis, and Pearl Therapeutics and has served as a consultant to GlaxoSmithKline and Merck. Dr Petersen serves on scientific advisory boards for Pfizer, Inc, Janssen Alzheimer Immunotherapy, Elan Pharmaceuticals, and GE Healthcare, receives royalties from the publication of Mild Cognitive Impairment (Petersen RC, ed. New York, NY: Oxford University Press; 2003), and receives research support from the NIA/NIH. No other disclosures were reported.

Funding/Support: This study was supported by grants P50 AG016574, U01 AG006786 (under which study participants were enrolled), K01 MH068351, and K01 AG028573 from the NIH; by the Robert H. and Clarice Smith and Abigail van Buren Alzheimer’s Disease Research Program; by Clinical and Translational Science Award UL1 TR000135 from the National Center for Advancing Translational Sciences, a component of the NIH; and by grant R01 AG034676 from the REP.

Role of the Sponsor: The funding sources had no role in the design and conduct of the study; collection, management, analysis, and interpretation of the data; preparation, review, or approval of the manuscript; and decision to submit the manuscript for publication.