Atrial Fibrillation and Cognitive Decline: the Cardiovascular Health Study

Supplemental Material

Thacker EL, McKnight B, Psaty BM, Longstreth WT Jr, Sitlani CM, Dublin S, et al. Atrial fibrillation and cognitive decline: a longitudinal cohort study

Appendix e-1. Supplemental methods and references

Figure e-1. Flowchart of participants and cognitive test scores for analysis

Table e-1. Model-predicted mean 3MSE scores for selected current ages and ages at incident AF

Table e-2. Model-predicted mean DSST scores for selected current ages and ages at incident AF

Appendix e-1

Supplemental methods

Assessment of covariates.Age, birth year, sex, race, education, cigarette smoking, and alcohol use were reported by participants at baseline. Baseline diabetes was defined as use of insulin or oral hypoglycemic drugs; or fasting serum glucose ≥ 7.0 mmol/L (126 mg/dL); or if participants were not fasting ≥ eight hours, random serum glucose ≥ 11.1 mmol/l (200 mg/dL).e1,e2 Systolic and diastolic blood pressures were the average of two measures. Baseline hypertension was defined as use of antihypertensive medication plus history of hypertension; or systolic blood pressure ≥ 140 mmHg; or diastolic blood pressure ≥ 90 mmHg.2 Baseline history of coronary heart disease, heart failure, and stroke were determined by self-report and medical record review.e3Incident diabetes and hypertension were identified through medication inventory (annually), serum glucose (1992/92 and 1996/97), and blood pressure measures (annually except 1995/96), using the same definitions as at baseline. Incident coronary heart disease, heart failure, and clinical stroke were ascertained through semi-annual clinic or phone follow-up contacts with participants or proxies and confirmed with information from clinical care including medical records, test results, and for stroke, brain images.e4-e6

Statistical analysis. We considered cognitive scores to be in two groups: (a) scores obtained from participants with no prior history of atrial fibrillation, and (b) scores obtained from participants after they developed incident atrial fibrillation. Age was a time-varying continuous variable centered at 77 years. Incident AF was a time-varying binary variable, and years since incident AF was a time-varying continuous variable.

The mixed effects linear models had an unstructured covariance matrix for the random effects. The model for 3MSE included random effects for the intercept, age, and agee2; and fixed effects for agee3, agee4, agee5, incident AF, years since incident AF, age × years since incident AF, and agee2 × years since incident AF. The model for DSST included random effects for the intercept and age; and fixed effects for agee2, agee3, agee4, incident AF, and years since incident AF.

The random effects were chosen a priori, however the random effect for agee2 in the DSST model was omitted because its coefficient was inestimable. The fixed effects for higher orders of age above agee2 were included because they minimized the Bayesian Information Criterion (BIC), a measure of model fit used to compare multiple potential models of varying complexity.e7 The interaction terms for age variables with AF variablesin the 3MSE models were included because they had Wald test p values < 0.05. In the DSST models, interactions for age with AF variables had Wald test p values > 0.05 and were therefore omitted.

e-References

e1.Cushman M, Cornell ES, Howard PR, Bovill EG, Tracy RP. Laboratory methods and quality assurance in the Cardiovascular Health Study. Clin Chem. 1995;41:264-270.

e2.Psaty BM, Lee M, Savage PJ, Rutan GH, German PS, Lyles M. Assessing the use of medications in the elderly: methods and initial experience in the Cardiovascular Health Study. J Clin Epidemiol. 1992;45:683-692.

e3.Psaty BM, Kuller LH, Bild D, et al. Methods of assessing prevalent cardiovascular disease in the Cardiovascular Health Study. Ann Epidemiol. 1995;5:270-277.

e4.Ives DG, Fitzpatrick AL, Bild DE, et al. Surveillance and ascertainment of cardiovascular events: the Cardiovascular Health Study.AnnEpidemiol.1995;5:278-285.

e5.Schellenbaum GD, Heckbert SR, Smith NL, et al. Congestive heart failure incidence and prognosis: case identification using central adjudication versus hospital discharge diagnoses. AnnEpidemiol.2006;16:115-122.

e6.Longstreth WT Jr, Bernick C, Fitzpatrick A, et al. Frequency and predictors of stroke death in 5,888 participants in the Cardiovascular Health Study. Neurology. 2001;56:368-375.

e7.Ramsey FL, Schafer DW. The Statistical Sleuth: A Course in Methods of Data Analysis, 3rd ed. Brooks/Cole; 2012.

1

Figure e-1. Flowchart of participants and cognitive test scores for analysis

Green boxes indicate numbers of participants. Orange boxes indicate numbers of 3MSE and DSST scores. Abbreviations: CHS, Cardiovascular Health Study; 3MSE, Modified Mini-Mental State Examination; DSST, Digit Symbol Substitution Test; AF, atrial fibrillation.

1

Table e-1. Model-predicted mean 3MSE scores for selected current ages and ages at incident AF*

Abbreviations: 3MSE, Modified Mini-Mental State Examination; AF, atrial fibrillation.

* Adjusted for birth year, sex, race, education, cigarette smoking, alcohol use, diabetes, hypertension, systolic blood pressure, coronary heart disease, heart failure, and the interaction of age with each covariate except birth year.

†Mean 3MSE scores and 95% CIs shown in the table correspond to those plotted in Figure 1 and Model 1 in Table 2. Scores on 3MSE range from 0 (worst cognitive performance) to 100 (best cognitive performance).

1

Table e-2. Model-predicted mean DSST scores for selected current ages and ages at incident AF*

Abbreviations: DSST, Digit Symbol Substitution Test; AF, atrial fibrillation.

* Adjusted for birth year, sex, race, education, cigarette smoking, alcohol use, diabetes, hypertension, systolic blood pressure, coronary heart disease, heart failure, and the interaction of age with each covariate except birth year.

†Mean DSST scores and 95% CIs shown in the table correspond to those plotted in Figure 2 and Model 1 in Table 3. Scores on DSST range from 0 (worst cognitive performance) to 90 (best cognitive performance).

1