COPD and cognitive impairment: the role of hypoxemia and oxygen therapy

Figures & data

Table 1 Baseline characteristics of FLOW cohort of COPD vs referent subjects

Characteristic	COPD (n = 1,202)	Referents (n = 302)	P-value
Age in years, mean (SD)	58.2 (6.2)	58.5 (6.2)	0.50
Female, n (%)	691 (57.4%)	185 (61%)	0.23
Race/ethnicity, n (%)			0.96
White, non-Hispanic	810 (67%)	200 (66%)
African-American	206 (17%)	57 (19%)
Asian/Pacific Islander	35 (3%)	8 (3%)
Hispanic	111 (9%)	28 (9%)
Other	40 (3%)	9 (3%)
Educational attainment, n (%)			<0.0001
High school or less	352 (29%)	42 (14%)
Some college	524 (44%)	95 (31%)
College/graduate degree	326 (27%)	165 (55%)
Smoking status, n (%)			<0.0001
Never smoked	165 (13%)	158 (52%)
Current smoker	393 (33%)	12 (4%)
Ex-smoker	644 (54%)	132 (44%)

Abbreviations: COPD, chronic obstructive pulmonary disease; FLOW, Function, Living, Outcomes and Work; SD, standard deviation.

Table 2 COPD and the risk of cognitive impairment

Comparison	Unadjusted OR (95% CI)	Controlling for age, sex, race, educational attainment, and smoking history OR (95% CI)
Referent Group	1.0	1.0
COPD	2.86 (1.34–7.46)	2.42 (1.043–6.64)

Abbreviations: CI, confidence interval; OR, odds ratio.

Table 3 Relationship between COPD severity and cognitive impairment

Disease severity measure	Risk of cognitive impairment
	Unadjusted OR (95% CI)	Multivariable OR (95% CI)†
Forced expiratory volume in 1 second*	0.97 (0.85–1.11) P = 0.63	0.93 (0.79–1.09) P = 0.35
Baseline oxygen saturation ≤88%‡	4.41 (0.92–21) P = 0.064	5.45 (1.014–29.2) P = 0.048
BODE Score*	1.13 (1.002–1.27) P = 0.046	1.08 (0.95–1.23) P = 0.25
COPD Severity Score*	1.08 (0.94–1.23) P = 0.28	1.03 (0.90–1.17) P = 0.66

* odds ratio per ½ standard deviation decrement.

† Separate multivariable logistic regression analysis for each severity measure plus age, sex, race, educational attainment, and smoking history.

‡ 61 subjects had oxygen saturation measured on supplemental oxygen (range 1–4 liters/minute). When oxygen use was added to the multivariable model, low oxygen saturation remained associated with a greater risk of cognitive impairment (OR 5.46; 95% CI 1.007–30; P = 0.049). In additional analysis, oxygen flow rate (no oxygen, flow = 0 liters/minute) was added to the multivariable model with minimal impact on the results (OR 5.44; 95% CI 1.008–29; P = 0.049).

Abbreviations: BODE, BMI (Body Mass Index), Obstruction, Dyspnea, Exercise Capacity; CI, confidence interval; COPD, chronic obstructive pulmonary disease; OR, odds ratio.

Figure 1 Relationship between resting oxygen saturation and the risk of cognitive impairment. Oxygen saturation was measured using pulse oximetry. The LOWESS (locally weighted regression scatter plot smoother) procedure was used to graphically depict the relationship between oxygen saturation and the logit (ie, log odds) of cognitive impairment. This method fitted a flexible smoothed curve that did not impose a linear relationship.

Table 4 Cardiovascular disease and the risk of cognitive impairment in COPD

Cardiovascular disease	Risk of cognitive impairment
	Unadjusted OR (95% CI)	Multivariable OR (95% CI)†
Stroke (n = 68; 4.5%)	0.81 (0.25–2.66) P = 0.73	0.78 (0.23–2.61) P = 0.68
Other cardiovascular disease (n = 841; 56%)*	1.57 (0.92–2.69) P = 0.10	1.30 (0.74–2.31) P = 0.36

* hypertension, coronary artery disease, or congestive heart failure.

† Multivariable logistic regression analysis including age, sex, race, educational attainment, and smoking history.

Abbreviations: CI, confidence interval; OR, odds ratio.

Table 5 Home oxygen use and the risk of cognitive impairment in COPD

Statistical model	Risk of cognitive impairment OR (95% CI)
Unadjusted	0.74 (0.26–1.71) P = 0.53
Multivariable model*	0.40 (0.11–1.17) P = 0.11
Multivariable model (propensity score adjusted)†	0.14 (0.07–0.27) P < 0.0001

* Multivariable logistic regression analysis including home oxygen use plus age, sex, race, educational attainment, smoking history, baseline oxygen saturation, BODE Score, COPD Severity Score, FEV₁, and height.

† Multivariable model including inverse probability-weighted estimators, which were derived from propensity scores. Propensity scores were developed from a separate multivariable logistic regression model to predict home oxygen use. Variables included were age, sex, race, smoking history, BODE Score, FEV₁, and baseline oxygen saturation (≤88%). Area under receiver operator characteristic curve for model was 0.92. See Methods for more details.

Abbreviations: BODE, BMI (Body Mass Index), Obstruction, Dyspnea, Exercise Capacity; CI, confidence interval; CI, confidence interval; COPD, chronic obstructive pulmonary disease; FEV₁, Forced Expiratory Volume in one second; OR, odds ratio.