INTRODUCTION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Lung function declines slowly throughout adult life, even in healthy persons. Cross-sectional analyses have suggested that the decline may accelerate after age 70 (1, 2). Many predictors of lung function decline have been identified in middle-aged persons (3), and the cross-sectional correlates of lung function in population samples of elderly persons have been previously published (7). However, the predictors of longitudinal decline in lung function have not been described in elderly persons and may differ from the cross-sectional correlates. The Cardiovascular Health Study (CHS) was started in 1990 in the United States as a population-based observational study of cardiovascular disease (CVD) in elderly persons. Spirometry testing was included in the CHS because lung function was previously demonstrated to be a strong independent risk factor for cardiovascular mortality in middle-aged persons (8- 11). Spirometry was repeated during two of the annual CHS follow-up examinations of the cohort, giving us the opportunity to determine the independent baseline predictors of change in lung function (FEV₁ and FVC) in this elderly cohort. The present study uses longitudinal statistical methods to determine significant baseline predictors of longitudinal changes in spirometry data.

	METHODS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Recruitment

The original CHS cohort of 5,201 men and women was selected using Medicare eligibility lists provided by the U.S. Health Care Financing Administration for four communities: Forsyth County, North Carolina; Pittsburgh, Pennsylvania; Sacramento County, California; and Washington County, Maryland during the period from May 1989 to May 1990. Potential participants were randomly sampled from the four communities to fill eight subgroups, stratified by age and sex. Recruitment was designed to achieve a 60:40 female:male ratio in each of the age groups, with 35% age 65 to 69 yr, 25% age 70 to 74 yr, 20% age 75 to 79 yr, and 20% age 80 yr and older. All persons of at least 65 yr of age who were residents of sampled households were asked to participate. Persons were excluded from participating in the study if they were institutionalized; terminally ill; unable to walk; unable to give informed consent; or likely to move from the area within the next 3 yr. Of those contacted, 9.5% were ineligible and 34.4% refused to participate. Between June 1992 and June 1993, an additional 687 African-American elderly were added to the study, with recruitment similar to that of the original cohort. Study participants self-reported their race as white, black, American Indian/Alaskan Native, Asian/Pacific Islander, or as specified in other. No steps were taken to verify the subjects' self-reports, and only the majority categories, white and black, were analyzed in this report.

Enrolled CHS participants were younger, more educated, and more likely to be married than those who refused or who were ineligible. The design and recruitment of the CHS cohort are described in detail elsewhere (12, 13). The research protocol was reviewed and approved by the institutional review board for human studies of each clinical center. Complete informed consent was obtained from all participants.

Interviews and Clinical Examination

Baseline clinical examinations took place during the 12 mo starting in June 1989. Along with other questionnaires, a subset of the standardized American Thoracic Society (ATS) DLD-78 Respiratory Questionnaire (14) was administered by a centrally trained interviewer. Participants were asked to fast for 12 h before the morning interview and examination. Examination components included spirometry, random-zero seated blood pressure, blood chemistry and lipid analysis, and a physical examination. Follow-up examinations occurred annually after baseline, with repeated spirometry testing being conducted only at Years 4 and 7. A detailed account of the interviews and the clinical examinations has been published elsewhere (15, 16).

Spirometry Testing

A standard spirometry system, validated by a third party to meet ATS spirometer recommendations (17), was purchased for all four clinical centers. A water-sealed spirometer connected to a personal computer was used. Software assisted the technician with quality control (QC) of maneuvers, calculated the pulmonary function test (PFT) variables, suggested interpretations, printed reports, and compressed graphical data for transmission and archival storage. Leak checks and spirometer volume checks using a 3-L syringe were repeated every morning. Internal spirometer temperature was measured for each maneuver, and was used for BTPS corrections.

The FVC maneuver was both explained and demonstrated to each participant. Participants were sitting unless they were severely overweight, as determined by a body mass index (BMI) of 35 kg/m² or more. After every FVC maneuver, acceptability and reproducibility checks were applied, quality control messages displayed, and the best three previous flow-volume curves from the test session were displayed. The FVC maneuver was repeated up to eight times or until at least three acceptable and two reproducible FVC maneuvers were obtained, in accordance with ATS recommendations (18, 19). The largest FEV₁ and the largest FVC from an acceptable maneuver were reported. Identical spirometers, software, procedures, and reading center personnel were used for all baseline and follow-up examinations. The majority of the technicians at each of the four clinical sites were also the same. There were no survey biases seen during the analysis.

The quality of the spirometry testing conducted by the technicians was monitored centrally throughout the testing (20). All test sessions were reviewed at the Pulmonary Function (PF) Reading Center by a single QC supervisor. A test session that just met the 1987 ATS recommendations received a grade of B. Reliable FVC maneuvers received a supervisor testing grade of A, B, or C. Results from test sessions graded D or F were excluded from the analysis. We elected to include sessions graded "C" because they included two acceptable maneuvers; however, the FEV₁ values did not match within 100 ml, thus not fulfilling the ATS 1987 recommendations (the most current when the baseline examination was completed, beginning in June 1989). However, the ATS PF committee recognized that a goal of < 100 ml reproducibility was excessively stringent, and relaxed this criterion to < 200 ml in 1994 (our QC grade of C). However, in both documents, ATS states that results from test sessions with suboptimal quality should not be discarded, because the inability to perform good spirometry tests is a predictor of morbidity and mortality in some studies.

Statistical Analysis

The longitudinal spirometry data were analyzed using a random effects model (REM) (21, 22). The form of the model is identical to that used in ordinary multiple regression, but the methods used to estimate the regression coefficients are modified to account for the correlation between repeated measures on the same subject. The model yields estimates of the mean PFT results related to age and the additional covariables, similar to multiple regression, and in addition gives subject-specific intercepts and slopes (i.e., the random effects portion of the model). This method of analysis allows subjects to have an unequal number of observations that can be at different times for different subjects. The covariables entered into the model were as follows: sex, weight, waist circumference, height, race, diagnosed myocardial infarction, diagnosed congestive heart failure (CHF), diagnosed angina, use of beta-blocker medications, systolic blood pressure, smoking status, asthma status, diagnosed pneumonia, diagnosed emphysema, diagnosed chronic bronchitis, dyspnea grade, test session grade, and clinic site (Table 1). All covariables were taken from the baseline spirometry survey, which included the ATS-DLD-78 questions. Self-reports of doctor diagnosis of asthma, pneumonia, emphysema, and chronic bronchitis, as well as smoking status and level of dyspnea defined subjects' status (see Appendix ). Covariables were included as fixed covariates in the models with the exception of age and test session grade. At each spirometry survey the test session grade and age were calculated and offered to the models as the only time-dependent covariates. Interaction terms with age, sex, and race were tested for significance for each of the covariates. The REM analyses were done using algorithms described by Jones and Boadi-Boateng (22) which permit the testing of several types of within-subject error structures, including a first-order autoregressive error structure. Selection of the best fitting model was done using maximum likelihood ratio tests for nested models and Akaike's Information Criterion (AIC) for non-nested models. Comparisons between the complete CHS cohort and those included in the analyses were made using Student's unpaired t test and chi-square tests. All hypotheses were tested at the  = 0.05 significance level.

	RESULTS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

There were 5,242 and 5,281 subjects who met the inclusion criteria of having at least one PFT measurement with quality control grade of C or better, corresponding to 89.0% and 89.7% of the baseline cohort for FEV₁ and FVC, respectively. The basic demographic statistics for those selected for analyses of their FEV₁ data are listed in Table 1 by sex. The sample population consisted of 57.6% females, who had a mean age of 72.7 yr at baseline, and was 86.3% white and 13.7% African-American. The male population had a mean age of 73.5 yr and was significantly more white (88.8%) and less African-American (11.2%) than females. Males were statistically more likely to have had a history of myocardial infarction, CHF, angina, beta-blocker use, and past history of smoking than females. However, females were more likely to be current smokers. Cohort participants included in the analyzed sample for FEV₁ and FVC were younger on average and reported fewer cardiovascular risk factors than those who did not meet inclusion criteria.

The mean time period between baseline spirometry and the first follow-up, and the first and second follow-up tests for the original cohort (acceptable spirometry maneuvers at each test) was 3.9 ± 0.2 and 3.0 ± 0.1 yr (mean ± SD), respectively. The supplemental African-American cohort recruited before the first follow-up examination for the original cohort contributed a maximum of two spirometry tests differing in time by 3.0 ± 0.1 yr. Table 2 reports the number and percentage of the baseline cohort that had acceptable spirometry tests for FEV₁ and FVC, stratified by age. Participants who were younger when they enrolled in the study (age 65 to 69 yr at their baseline examination) were more likely than the older participants to have performed acceptable quality test sessions during all three exams (baseline, Year 4 and Year 7 follow-up). For example, 43% of those in the 65 to 69 age group, 35% in the 70 to 74 age group, 27% in the 75 to 79 age group, and only 15% in the age 80+ group gave acceptable quality FEV₁ at all three examinations. The less frequent follow-up data in the participants who were older when they enrolled in the study are largely due to their higher morbidity and mortality rates (preventing their participation in follow-up clinic visits).

The unadjusted mean annual changes in FEV₁ for CHS participants stratified by sex, baseline smoking status, and ethnicity are given in Table 3 for comparison purposes with previous studies that reported only simple change in FEV₁ by sex and smoking status (23, 24). Annual changes were calculated by computing delta FEV₁ values between the furthest acceptable spirometry maneuvers divided by the time between testing. Thus, annual change for the original cohort represents at the maximum a 7-yr mean, but a 3-yr mean for the supplemental African-American cohort. Currently smoking elderly white men, women, and African-American men experienced greater declines in FEV₁ than never and former smokers. Elderly African-American women recruited into the CHS experienced a relatively low rate of decline in FEV₁, regardless of smoking status.

The results of fitting the REM model to the spirometry data are listed in Tables 4 and 5 and plotted in Figures 1-6. For both data sets (FEV₁ and FVC) the best-fitting REM model included the first-order autoregressive within-subject error structure. These results are presented in three major categories: anthropometric, cardiovascular, and pulmonary factors. The estimated coefficients are separated into two columns indicating either a constant difference from the reference group (i.e., change in intercept) or a difference in slope with age (differences in slope with age are indicated by "Age × variable" interactions). If the intercept is significant and the interaction with age is not, this implies that the slopes with age for both groups are the same. For example, the FEV₁ coefficient for male sex equals 1.076 (Table 4), which indicates a 1.076-L constant difference in the intercept between males and females, with males being higher. The Age × Male sex coefficient (0.0053) indicates that males had a slightly steeper slope with age than females (about a 5 ml/yr more rapid decline in FEV₁).

View larger version (10K): [in this window] [in a new window]   Figure 1.   Mean predicted FEV1 by age for elderly CHS participants, stratified by sex and race. C = white (solid lines); AA = African-American (dashed lines).

View larger version (11K): [in this window] [in a new window]   Figure 2.   Mean predicted FEV1 by age for elderly men, stratified by baseline reports of dyspnea on exertion, use of beta-blocker medication, and a diagnosis of emphysema, compared with other men (Reference).

View larger version (10K): [in this window] [in a new window]   Figure 3.   Mean predicted FEV1 by age for elderly men, stratified by baseline smoking status: current smokers, former smokers, and never smokers (Reference).

View larger version (10K): [in this window] [in a new window]   Figure 4.   Mean predicted FVC by age for elderly CHS participants, stratified by sex and race. C = white (solid lines); AA = African-American (dashed line).

View larger version (11K): [in this window] [in a new window]   Figure 5.   Mean predicted FVC by age for elderly men, stratified by baseline reports of dyspnea on exertion, use of beta-blocker medication, and a diagnosis of emphysema, compared with other men (Reference).

View larger version (10K): [in this window] [in a new window]   Figure 6.   Mean predicted FVC by age for elderly men, stratified by baseline smoking status.

Among the anthropometric factors, weight and height were positively associated with FEV₁, whereas age and waist size were negatively related. The rates of FEV₁ decline for African-Americans with age were not as steep as those for whites (+12 ml/yr) but were still negative (see Figure 1). Figure 1 shows these sex and racial differences; where the lower rate of decline in FEV₁ for African-Americans over time leads to insignificant racial differences at older ages, in particular for women.

Among the cardiovascular variables, CHF (156 ml), systolic hypertension (485 ml), and taking beta-blocker medication (461 ml) were all associated with lower levels of FEV₁. The association of CHF was independent of age whereas the effects of systolic hypertension and the use of beta-blockers diminished with age (i.e., the slopes were 6 ml/yr less negative). These age-related declines in the steepness of the rate of fall in FEV₁ make the negative association of taking beta-blockers on FEV₁ negligible in subjects over 84 yr of age, and for systolic hypertension over age 90.

Subjects reporting respiratory diseases (emphysema and current asthma), smoking, and dyspnea on exertion had lower mean FEV₁ values, and among men the effect of respiratory diseases on FEV₁ was greater than among women. The association of emphysema and dyspnea with FEV₁ lessens with age for both males and females (Figure 2). Subjects reporting asthma (former or current) had lower levels of FEV₁ than subjects who never had asthma. Asthma also has a significantly larger effect on FEV₁ levels for males compared with females (decreasing FEV₁ by 246 ml). Chronic bronchitis (144 ml) and pneumonia (81 ml) were associated with significantly lower levels of FEV₁. Among cigarette smokers, former smokers had significantly lower levels of FEV₁ than never smokers, and current smokers had significantly steeper slopes (Figure 3). African-American current smokers had a significantly higher mean FEV₁ than did white current smokers (192 ml higher).

The same anthropometric variables that were significant for FEV₁ were significant for FVC (Table 5). However, there were some differences: the coefficient for age was larger, suggesting a more rapid decline of FVC than FEV₁ with age (66 ml/yr versus 47 ml/yr, respectively) among the subjects in the reference group. Also the weight coefficient for FVC was negative, but was positive for FEV₁. African-Americans had significantly lower mean levels of FVC (1.017 L and 1.245 L for females and males respectively) but slightly more positive slopes (10 ml/yr) than whites. The significant interactions between race and age suggest that racial differences diminish with increasing age (Figure 4).

The cardiovascular variablesCHF, systolic hypertension, and the use of beta-blockerswere associated with even lower mean FVC (when compared with FEV₁). The association of CHF did not change with age, whereas the effects of beta-blockers and systolic hypertension again diminished with age (+7 and +11 ml/yr, respectively) (Figure 5).

The same pulmonary variables that were associated with FEV₁ were also associated with FVC with similar magnitude and direction (see Table 4). Men who were current smokers at the baseline examination had excessive declines in FVC with increasing age (13 ml/yr) (Figure 6).

	DISCUSSION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

In the current study, we have demonstrated several notable independent baseline predictors of change in lung function, FEV₁ and FVC in an elderly population. Both spirometry measures, FEV₁ and FVC, decrease with age for whites and African-Americans, with African-Americans (especially women) having slower rates of decline with age. Subjects reporting CHF, high systolic blood pressure (> 160 mm Hg), or taking beta-blockers had significantly lower spirometry levels; however, the effects of high blood pressure and taking beta-blockers diminished with increasing age. Chronic bronchitis, pneumonia, emphysema, and asthma were associated with reduced spirometry levels. Participants with current asthma had a mean 0.5 L lower FEV₁ at their baseline examination, but did not subsequently experience more rapid declines in FEV₁ compared with those without asthma. The most notable findings of these analyses were associated with current smoking status. Current smoking (especially for men) was associated with more rapid rates of decline in FVC and FEV₁ compared with never smokers. African-American current smokers had a significantly higher mean FEV₁ than did white current smokers.

Previous investigators, when analyzing spirometry changes over several years from middle-aged, population-based cohorts, found several independent predictors of excessive decline in FEV₁: medium to heavy current cigarette smoking (25), occupational exposures (26), airway hyperresponsiveness (27, 28), chronic phlegm (29), and malnutrition (32). Other analyses showed that changes in other factors during the testing intervals, such as smoking cessation (33), weight gain (33), and a reduction in fruit consumption (36), were also associated with changes in lung function over several year. In the current report we attempted to demonstrate baseline predictors influencing longitudinal spirometry change; hence we did not include in our models changes in factors that occurred subsequent to the baseline CHS examination.

In contrast to the aforementioned analyses, using a REM we found no baseline factors other than current smoking that were independent predictors of accelerated decline in FEV₁ or FVC over several years in a population sample of elderly persons. Three baseline factors surprisingly predicted less decline in FEV₁ and FVC: a history of dyspnea on exertion, the presence of emphysema, and the use of beta-blocker medications. We believe that these positive associations are the result of a survivor effect; participants with these factors, which indicate clinical heart or lung disease at the baseline examination, had significantly lower mean lung function at the baseline examination and were, therefore, more likely to experience events that prevented them from attending follow-up clinic examinations (or from providing acceptable spirometry tests). The mean follow-up values of the survivors were not as low as those of the group tested at baseline. This hypothesis is strengthened by the positive association with the age-squared term (results not shown), indicating that decreases in spirometry with age moderate as age increases.

The factors that were found by the REM to be independently associated with lower FEV₁ at all three examinations of this cohort (but not an accelerated decline in lung function, because their age interaction terms were not significant) were previously described using simple linear regression models, using only the baseline data (37, 38). Height, weight, waist size, and race are all factors that affect lung size (both FEV₁ and FVC), whereas smoking, systolic hypertension, and chronic bronchitis are markers of clinical disease, and pneumonia, asthma, emphysema, and CHF all indicate clinical heart or lung diseases. All of these factors, with the notable exception of smoking, however, appear to have exerted their influence in reducing lung function before to the baseline examination, and their presence at the baseline examination did not lead to excessive subsequent declines in lung function during the follow-up years. Current smokers, however, had low lung function at the baseline examination, as well as accelerated losses of FEV₁ (men and women) and FVC (men only) with age during the study.

Results of a study by Ulrik and Lange (39) showed an excessive decline in FEV₁ in patients with asthma, which we did not find in this cohort; albeit that this cohort included approximately 300 subjects with asthma (past or current), with evidence that only 30% of those with current asthma were taking inhaled corticosteroids (40).

The magnitude of the independent association of age on lung function in the current analysis, a mean loss of FEV₁ of 52 ml/yr for men and 47 ml/yr for women, is substantially higher than that estimated from the cross-sectional analysis of the healthy subset of the original cohort at their baseline examination, where the age coefficients were 27.1 ml/yr for men and 32.5 ml/yr for women (38). The significant sex × age interaction term shows that healthy elderly men do lose lung function at a faster rate (in absolute terms) when compared with healthy elderly women. The opposite was suggested by the previous cross-sectional analysis (38). The lack of a negative age-squared term shows that accelerated decline in lung function in healthy elderly women does not occur, as was suggested by previous investigators using cross-sectional data from Tucson, Arizona (1).

A similar study was conducted by Sherrill and coworkers (41) on elderly participants of the Tucson Epidemiology Study of Obstructive Airways Disease. This analyses included 1,524 white (non-Hispanic) subjects all over 55 yr of age, who had up to six PFT spanning a period of up to 14 yr. Their FEV₁ and FVC longitudinal data were analyzed using the same REM procedure. Their results also showed that males had steeper absolute rates of decline with age for FEV₁ and FVC than females; however, they did not observe significant differences in the rates of decline between FEV₁ and FVC, as reported herein. They also reported that current smokers and ex-smokers had significantly lower levels of FEV₁ than never smokers. Male current smokers had significantly steeper rates of decline with age, and the increased rate of decline was similar to that found in the current study (7 versus 5 ml/yr). In contrast to the current findings, the only significant association of smoking on FVC they found was among ex-smoking females. They also found reduced levels of FEV₁ and FVC related to dyspnea, but independent of age (whereas the current study found the association with dyspnea to diminish with increasing age).

Elderly African-American women in the CHS appear to be relatively resistant to the effects of aging and of cigarette smoking on lung function. Their mean rate of decline of FEV₁

(approximately 35 ml/yr, regardless of smoking status) is significantly less than that of elderly white women (47 ml/yr for never smokers). The mechanisms that account for this stability of their lung function should be studied further. A recent meta-analysis of several National Heart, Lung, and Blood Institute (NHLBI)-funded observational studies also showed that African-American women exhibited lesser smoking-related declines in FEV₁ than did white women (5 versus 9 ml/yr excess decline in those smoking > 10 cigarettes/d).

Potentially modifiable risk factors associated with low lung function include systolic hypertension, CHF, cigarette smoking, and asthma. Many elderly patients with CHF are not treated optimally and most elderly patients with asthma are not taking anti-inflammatory medications (42).

We expected that the use of beta-blocker medications would be associated with lower FEV₁ because they can cause bronchoconstriction in patients with asthma, and because the CVD that are treated with beta-blockers may also cause reduced lung function (10). However, we were surprised to note that declines in FEV₁ and FVC were less in patients using beta-blockers. This effect should be studied prospectively in other cohorts and cardiovascular intervention trials. One of the limitations of the current study is the shorter duration of follow-up for the supplemental African-American cohort. However, given the number of African-Americans included, we determined that we had greater than 80% power to detect a 3% change in the annual decline of lung function (assuming  = 0.05) (43). Other limitations of this study include the exclusion of institutionalized elderly persons, the marginally higher socioeconomic status of the cohort, and the survivorship effect of those who completed the baseline examination and follow-up exams.

In summary, the results of this study show that elderly African-Americans had significantly lower mean FEV₁ and FVC values than whites but that their rate of decline with age was significantly less. Elderly persons with CHF, systolic hypertension, and taking beta-blocker medications had significantly lower mean FVC (and FEV₁) values, but this association diminished with increasing age. Several pulmonary diseases were also associated with lower values, including chronic bronchitis, pneumonia, emphysema, and asthma, but only those with emphysema had rates of decline that differed from the reference group. Finally, past smoking was associated with lower FEV₁ and FVC values; however, only current smokers (males only for FVC) had steeper rates of decline with age. The most notable finding of these analyses was that current smoking (especially for men) was associated with more rapid rates of decline in FVC and FEV₁. African-Americans (especially women) had slower rates of decline in FEV₁ than did whites. Although participants with current asthma had a mean 0.5 L lower FEV₁ at their baseline examination, they did not subsequently experience more rapid declines in FEV₁.