Study design and participants
The Good Aging in Skåne (GÅS) is a prospective, longitudinal population-based study aiming to investigate different aspects of aging27. Briefly, 60- to 93-year-old subjects living in the south of Sweden, Skåne, are randomly selected from the Swedish population register. Participants are offered a thorough physical, medical, and psychological examination and are invited to follow-up examinations at regular intervals every three (≥78 years of age) to six (<78 years of age) years until death. To encourage the participation of frail adults, the study team performs home visits following the same study protocol. However, patients living at nursing homes or with limited mobility may receive a shorter examination if they are not able or willing to perform a full examination. Currently, three waves have been fully recruited, with an initial participation rate of approximately 60%. As shown in Fig. 1, data from 5804 participants examined between 13th Feb 2001 and 21st Jun 2021 were retrieved. In order to study early-onset progression and to minimize bias, participants diagnosed with COPD and/or dementia at the first study visit were excluded from this analysis. The GÅS study is conducted according to the Declaration of Helsinki and Good Clinical Practice guidelines. The study is approved by the Lund University Ethics Review Board (LU 744-00). All participants provided written informed consent.
Cognition in the GÅS study was assessed using a neuropsychological test battery including 12 instruments administrated by experienced test administrators following a standardized protocol. In this work, we focused only on cognitive functions that were most frequently found to be affected in COPD patients in previous studies (global cognition, memory, executive function, and word fluency2,10,14,16,18,20,28). Global cognitive status was assessed using the mini-mental state examination (MMSE). Episodic memory was assessed using a 16-word free recall test. Executive function was assessed using a modified version of the WAIS-III digit span backward test. The language domain was assessed using a categorical verbal fluency test (animals). A description of the GÅS test battery has been presented elsewhere29.
Socio-demographics and morbidities
Socio-demographic variables included age, sex, and years of formal education. Alcohol consumption was self-reported. Morbidities were identified by several methods: self-reports, medical examination, reviewing medical records, and linking to the Skåne Healthcare Registry (SHR). In the SHR, diseases are classified according to the International Classification of Diseases System version 10 (ICD-10). Cerebrovascular disease included cerebral infarction, cerebral hemorrhage, occlusion and stenosis of precerebral or cerebral arteries, and/or transient cerebral ischemia. Heart disease included acute myocardial infarction, ischemic heart disease, presence of cardiac and vascular implants, heart failure, and atrial fibrillation or other arrhythmias. COPD also includes emphysema and chronic bronchitis. During the study visit, symptoms of depression were assessed using the Montgomery–Åsberg depression rating scale (MADRS). Dementia diagnosis was primarily extracted from medical records and included Alzheimer’s disease, mixed dementia, vascular dementia, Lewy body dementia, and frontotemporal dementia. Complementary diagnosis of all-cause dementia was classified by the study physician during thorough examination according to the Diagnostic and Statistical Manual of Mental Disorders 4th edition (DSM-IV). The date of death was obtained by linkage to the Swedish people and address register.
Participants with COPD were diagnosed in standard clinical practice. Clinically, diagnosis of COPD in Skåne county is based on three criteria: spirometry verified obstructive after bronchodilation, current airway symptoms, and a history of risk factors for COPD.
During the study visits, a spirometry assessment was performed using a Vitalograph 2120 spirometer (Vitalograph Ltd., Buckingham, UK) according to the American Thoracic Society guidelines30. Bronchodilators were not administrated during the first wave baseline visit. Subjects received 1.0 mg of β2-receptor agonist terbutaline 10 min prior to the spirometry in the majority of visits. The spirometry results were used in a sensitivity analysis as an alternative way to identify COPD cases.
Mixed models for repeated measures were implemented to estimate the mean neuropsychological test scores over time. The models had a random intercept for each participant. The rate of change of neuropsychological test score was calculated as the difference between the score at the follow-up visits and the baseline score, divided by the time between the visits. Using a directed acyclic graph, we identified which factors were required in the model to estimate the direct effect of COPD on cognitive performance. Age, sex, education, alcohol consumption, heart disease, cerebrovascular disease, depression score, and diabetes at baseline were then included to control for confounding. Since our model is optimized for the interpretation of the coefficients related to COPD, no conclusion is drawn regarding the effect of the other covariates on cognition31. In order to improve precision, we also included the test score at baseline in the models. Confidence intervals were calculated using robust standard errors. The goodness of the fit was assessed visually using residual plots. In this study, there were several statistical hypotheses of interest (one for each cognitive domain). To reduce the burden of multiplicity, we implemented a fixed sequence testing procedure32. Briefly, we ordered the cognitive domain hypotheses according to our a priori expectations of observing a difference between COPD and non-COPD participants. In this procedure, the first hypothesis is tested, and if rejected, the second is tested. The testing procedure continues until a hypothesis cannot be rejected. This hierarchical testing approach controls the type I error at 5% in the strong sense. The implemented testing order was (1) immediate recall test, (2) fluency test, (3) digit span test, and (4) MMSE. An extended Cox model using baseline characteristics (age, sex, education, heart disease, cerebrovascular disease, and diabetes) was implemented to estimate the risk of dementia for participants with and without COPD. The presence of COPD diagnosis was modeled as a time-dependent covariate. Violations of the proportional hazard assumption were assessed visually using log-log plots, as well as with a formal test, in order to investigate whether the log hazard-ratio function was constant over time. The overall prevalence of dementia and COPD in the GÅS study are 6.5%33 and 5.5%23, respectively. Therefore, the number of incident dementia cases in COPD participants in this study was expected to be low. Dementia is thus presented as a secondary outcome and excluded from the testing hierarchy.
Potential sources of bias were COPD misclassification, cohort effects, and attrition. Assigning a correct COPD diagnosis in older adults with comorbidities is difficult, mainly due to overlapping symptoms and difficulties performing a correct spirometry34. To overcome the uncertainty in COPD diagnosis in clinical practice, sensitivity analyses were performed using spirometry measurements collected during the study visits to identify COPD cases. Participants whose ratio of forced expiratory volume in the first second (FEV1) and forced vital capacity (FVC) was <0.7 were classified as having COPD. The data were collected over the course of 20 years, and therefore cohort effects are expected. The primary models were stratified by study wave to investigate this issue. Attrition is inevitable in cohort studies involving older adults. Thus, we conducted sensitivity analyses considering only the first follow-up visit. The statistical analyses were performed using Stata IC 17.0 software (StataCorp LLC, TX, USA) and Python 3.8.5 (Python Software Foundation).
Further information on research design is available in the Nature Research Reporting Summary linked to this article.