Brief Communication Effects of Cigarette Smoking on Rate of Loss of Pulmonary Function in Adults: A Longitudinal Assessment':'
XIPING
xu,
DOUGLAS W. DOCKERY, JAMES H. WARE, FRANK E. SPEIZER, and BENJAMIN G. FERRIS, JR.
As a part of the Six Cities Study of Air Pollution and Health, we selected a random sample of adult residents of Six Cities in the eastern United States and examined them at 3-yr intervals over a 6-yr period (1). In a previous report, we described analyses of the baseline spirometric data showing that levels of FVC and FEV 1 in both current and former smokers depended linearly on total pack-years of cigarette smoking. Current cigarette smokers had an additional deficit relative to former smokers with the same smoking history as measured in pack -years, This report uses data obtained during a 6-yr follow-up of this cohort to obtain longitudinal estimates of the effects of current smoking behavior on the rate of loss of pulmonary function between examinations. The sampling design of the original study has been described in detail elsewhere (2-4). Briefly, a random sample of adults 25 to 74 yr of age was drawn in each of six U.S. cities (Watertown, MA; KingstonHarriman, TN; Steubenville, OH; a geographically defined portion of St. Louis, MO; Topeka, KA; and Portage, WI) beginning in September 1974. Participants completed a standardized respiratory disease questionnaire and performed a forced expiratory maneuver on a water-filled 8-L recording survey spirometer (Warren E. Collins, Braintree, MA) while in the sitting position, without noseclip (2, 3). The means of the best three values of FEV 1 and FVC, corrected to BTPS, were determined. Each participant was invited back for a follow-up examination at the same time of year 3 and 6 yr after the initial examination. The data for each subject were expressed as the initial pulmonary function measurements and the changes in pulmonary function level during the 3-yr intervals between follow-up examinations. Thus, each subject contributed one observation to the cross-sectional analysis of initial measurements and up to two observations (intervals) to the longitudinal analysis of changes in pulmonary function during the 6 yr of follow-up. Both smoking history and smoking during intervals between examinations were measured in pack-years. Never smokers reported lifetime smoking of less than 25 packs and no cigarette smoking during an interval. Former smokers reported a smoking history but no smoking during an interval. Continuing smokers reported smoking throughout the interval, quitters were smoking at the beginning but stopped in the interval, and starters did not smoke at the beginning but started during the interval. The cross-sectional effects of age and smoking on FEV 1 and FVC at the initial visit, and effects of these variables on change in FEV 1 and FVC be-
SUMMARY Data from a random sample of 8,191 men and women selected In six U.S. cities and examined on three occasions over a 6-yr follow-up period were analyzed by longitudinal methods to describe the effects of smoking history and current smoking behavior on rate of loss of pulmonary function during adult life. Former smokers had age- and height-adjusted rates of decline (34.3 ml/yr for men and 29.6 ml/yr for women) comparable with those of never smokers (37.8mllyr for men and 29.0 ml/yr for women) but much smaller than those of continuing smokers (52.9 ml/yr for men and 38.0 ml/yr for women). The accelerated rate of loss of FEV1 among smokers depended linearly on the number of cigarettes smoked per day during the interval between examinations. The estimated Increase in rate of loss associated with smoking was 12.6 ml/yr per pack/day for men and 7.2 ml/yr per pack/day for women. These longitudinal estimates of the effects of smoking were approximately 50% larger than estimates obtained from cross-sectional analysis of the initial pulmonary function examination. Men who started smoking had accelerated rates of loss (55.9 ml/yr) as did women (43.1 ml/yr). ,Smokers who stopped smoking between examinations had reduced declines (41.2 ml/yr for men and 28.7 ml/yr for women) compared with continuing smokers. The age-specific rates of loss suggest that the benefits of cessation may be greatest among the youngest smokers. The changes associated with starting or cessation of smoking were much smaller than those found In cross-sectional analyses of these data. In summary, cigarette smoking causes an accelerated annual loss of pulmonary function that may be underestimated by cross-sectional studies. Those who stop smoking will experience only a small recovery in pulmonary function level, but they will cease to lose pulmonary function at an accelerated rate. AM REV RESPIR DIS 1992; 146:1345-1348
tween examinations, were estimated simultaneously using methods described previously (4). The crosssectional model for the initial measurement for ith subject, can be written as PFit = b, + b 1*(Agecit) + b 2*(Agecit)2 + b 3*(Pack-yearit) + b 4*(Current smokingj.) + eit, where PFit = height-adjusted FEV 1 or FVC for subject i at visit 1, Agej, = age of the ith subject at visit 1 (Agecj, = Agej, - 50), Pack-yean 1 = lifetime pack-years of cigarette smoking at visit 1, Current smokingi, = 1 (smoker at visit 1), 0 (otherwise), and eit = error in the regression equation. Two longitudinal models were considered in this study. The more general model can be expressed as Dit = b'l*(AgeCi,t+1 - Agecjj) + b'2*(Ageci,t+12 - Agecjr') + b'3*(Packyean,t+1 - Packyearjj) + b'4*(Start smokingjj) + b's*(Stop smokingjj) + eit, where Dit = PFi,t+l - PFit, t = 1,2, Start smokingjj, is an indicator variable equal to 1 if the participant started smoking in the interval and o otherwise, and Stop smoking is defined analogously. This longitudinal model provides estimates of the effects of age and change in pack-years of smoking that are directly comparable to the coefficients of the cross-sectional model. The two indicator variables for change in smoking status allow separate estimation of the effects of initiation and cessation of smoking. In this regard, the longitudinal model is more general than the cross-sectional model. In a separate analysis, intervals during which participants reported a change in smoking status were deleted from the analysis, and the indicator
variables for change in smoking status were dropped from the model. The regression coefficients and variances of the estimated coefficients of the crosssectional and longitudinal models were estimated simultaneously by the method of generalized estimating equations (5). The data for men and women were analyzed separately. Effects of cigarette smoking are expressed as effects per pack/day. Age-adjusted rates were based on the age distri-
(Receivedin originalform December30, 1991 and in revised form April 6, 1992) 1 From the Environmental Epidemiology Program, Department of Environmental Health, and the Department of Biostatistics, Harvard School of Public Health, and the Channing Laboratory, Brigham and Women's Hospital, Harvard Medical School, Boston, Massachusetts. 2 Supported in part by Grants ES-Oll08 and ES-OOOO2 from the National Institute of Environmental Health Sciences, by Cooperative Agreement CR-811650 from the Environmental Protection Agency, and by Contract RP-l00l with the Electric Power Research Institute. 3 Correspondence and requests for reprints should be addressed to Douglas W. Dockery, ScD, Environmental Epidemiology Program, Department of Environmental Health, Harvard School of Public Health, 655 Huntington Avenue, Boston, MA 02115.
1345
1346
BRIEF COMMUNICATION
TABLE 1 AVERAGE RATE OF CHANGE OF FEV1 (ml/yr) BY SEX, AGE, AND SMOKING GROUPS DURING THE INTERVAL
Age (yr) Men 25-34 35-44 45-54 55-64 65-78 n Women 25-34 35-44 45-54 55-64 65-78 n
Continuous Smokers (cig/day)
(n)
Never Smoker
Former Smoker
Total
< 15
15-24
25+
Starter
Quitter
878 977 1,496 1,247 839 5,437
-15.2 -24.9 -41.9 -45.9 -55.6 1,475
-6.4 -23.9 -39.4 -43.2 -52.3 1,842
-34.2 -34.5 -63.8 -60.4 -63.3 1,605
-36.5 30.4 -55.0 -49.0 -66.8 218
-20.6 -40.3 -57.4 -53.8 -54.1 519
-41.7 -39.0 -68.8 -68.5 -73.4 868
-19.4 -73.1 -56.3 -65.4 -58.6 349
6.2 -47.4 -32.2 -66.7 -60.2 127
1,039 1,158 1,823 1,557 1,066 6,643
-13.8 -25.0 -30.4 -34.1 -39.2 3,449
-13.9 -29.8 -26.0 -40.6 -34.9 977
-22.2 -34.9 -40.1 -48.3 -38.2 1,837
-16.5 -28.9 -32.5 -42.7 -29.4 486
-21.9 -35.8 -45.7 -52.0 -47.9 721
-28.0 -36.9 -37.8 -50.4 -37.4 630
-29.1 -61.3 -43.0 -52.8 -22.4 241
7.4 -6.2 -31.7 -38.9 -70.1 115
TABLE 2 AGE-ADJUSTED AVERAGE RATE OF CHANGE OF FEV1 AND FVC (mllyr) BY SEX AND SMOKING GROUPS DURING THE INTERVAL*
FEV1 Men Women FVC Men Women
Continuous Smokers (cig/day)
Never Smoker
Former Smoker
Total
< 15
15-24
25+
Starter
Quitter
-37.8 (2.0)
-34.3 (1.8)
-52.9 (2.0)
-37.4 (6.0)
-47.2 (3.6)
-59.9 (3.5)
-55.9 (9.6)
-41.2 (5.0)
-29.0 (0.9)
-29.6 (1.6)
-38.0 (1.2)
-31.2 (2.4)
-42.0 (1.8)
-38.9 (2.5)
-43.1 (5.7)
-28.7 (4.3)
-43.8 (2.1)
-43.4 (2.1)
-59.8 (2.5)
-58.6 (7.2)
-51.9 (4.3)
-63.8 (4.4)
-70.4 (11.8)
-49.3 (6.1)
-32.9 (1.0)
-34.0 (1.8)
-40.1 (1.4)
-35.2 (2.7)
-42.9 (2.2)
-40.6 (2.9)
-39.5 (6.9)
-39.9 (5.2)
• Standard errors, shown in parentheses, are calculated by robust methods (5).
TABLE 3 CROSS-SECTIONAL (CS) AND LONGITUDINAL (LN) ESTIMATES OF THE EFFECTS OF AGE AND CIGARETIE SMOKING ON HEIGHT-ADJUSTED FEV1 AND FVC FOR MEN AND WOMEN* (Age - 50) (mllyr)
(Age - 50)2 (mllyr 2 )
Pack-Years (ml/yr/pack/day)
CS LN
3,502 (16)
- 34.6 (0.8)* - 34.1 (1.2)
- 0.16 (0.05) - 0.47 (0.04)
-7.4 (0.5) -12.6 (1.5)
-232 (22)
CS LN
2,518 (9)
-27.8 (0.4) -28.8 (0.7)
- 0.10 (0.03) - 0.26 (0.03)
-5.2 (0.5) -7.2 (1.2)
-114(16)
CS LN CS LN
4,398 (18)
-34.1 -40.5 -27.9 -31.3
-
Model FEVl Men Women FVC Men Women
Current Smoking (ml)
Intercept (ml)
3,118 (11)
(0.9) (1.4) (0.5) (0.8)
0.17 0.70 0.22 0.44
(0.06) (0.05) (0.04) (0.03)
-6.0 -10.9 -4.1 - 6.1
(0.6) (1.7) (0.6) (1.4)
-159 (24) -62 (18)
• Standard errors are shown in parentheses.
bution of the sample in decades. All pulmonary function values were corrected for body size by dividing by the square of the subject's height, measured at the initial visit (2). Regression coefficients and fitted values were converted to volume units by multiplying by the square of the mean height, 1.73 m for men and 1.61 m for women.
The initial sample consisted of 8,191 white adults, 6,508 (79070) of whom were reexamined 3 yr later, and 5,572 (68%) after 6 yr. Loss-to-follow-up was 38% among men and 31% among women. Current smokers had slightly higher loss-to-follow-up rates (35%
with incomplete follow-up) than either never smokers (29%) or former smokers (31%). Womenweremore likelyto report neversmoking (52%) than were men (27%), whereas a higher percentage of men (34%) than women (15%) reported being former smokers. During follow-up, 6.4% of the men and 3.6% of the women reported that they had quit smoking between examinations, 2.3% of men and 1.7% of women reported starting smoking. Average annual rates of decline in FEV 1 increased continuously with age among never smokers and younger smokers for both men and women (table 1).The "flattened" declines observed in the oldest cigarette smokers may be explained by small sample size, survival effects, and loss-to-follow-up. Sex-specific age-standardized rates of loss of FEV 1 increased with the average number of cigarettes smoked per day during the interval (table 2). Former smokers had rates of loss comparable to those of never smokers. Smokers who quit had average rates of loss slightly greater than those of never smokers and former smokers, but still significantly smaller than the rates for smokers (table 2). Nonsmokers who started had rates of loss greater than never smokers, former smokers, and all but the heaviest smokers. Sex-specific longitudinal regression analyses of the FEV 1 changes for intervals in which participants did not change smoking status showed that current cigarette smoking was associated with an increase in the rate of loss of FEV 1 of 12.6 ml/yr per pack/day in men and 7.2 mllyr per pack/day in women (table 3). Corresponding estimates of the effects of cigarette smoking history obtained from cross-sectional analysis of the pulmonary function measurements at the initial examination were 7.4 ml per pack-year in men and 5.2 ml per pack-year in women. Thus, the cross-sectional estimates of the effect of smoking are 59% and 72% of the longitudinal estimates for men and women, respectively. Among men, the difference between the cross-sectional and longitudinal estimates was highly significant (p < 0.001), but the difference did not reach statistical significance among women (p = 0.13). The observed sex-specific rates of change of FEV 1 versus packs/day of cigarette smoking among ever-smokers (0 for former smokers and tertiles of packs/day for continued smokers) along with the predicted means for each group, which were calculated from the above fitted longitudinal model using the age and smoking variables described above are shown in figure 1. The curvature of the predicted values is due to the differences in age between the smoking groups. A similar analysis was performed for the initial measurements and changes between examinations in height-adjusted FVC (table 3). For men, the longitudinal and cross-sectional estimates of the effect of cigarette smoking on rate of loss (10.9 versus 6.0 ml/yr per pack/day, respectively) were significantly different (p = 0.01). For women, the differ-
1347
BRIEF COMMUNICATION
smaller change in pulmonary function associated with either smoking cessation or initiation than the estimated effect of current smoking than in the cross-sectional model would suggest.
-25
-30
....
»
-35
~
-40
rLI
c... ~
U v: ~ Q
-45
-50
,
~ rLI
o
;Z;
.,;
-55
Men
::r::
u
~
rLI
-60
;::;1
-65
-70
MEAN PACKS / PER DAY
Fig. 1. Predicted and observed annual rates of change of height-adjusted FEV, for men and women. The circles and error bars represent the mean ± 95% confidence interval of the observed annual rate of loss of height-adjusted FEV, for subjects grouped by former smokers (mean packs/day = 0) and continued smokers divided into three subgroups by tertiles of mean packs/day.The dashed line connects the means predicted from the longitudinal model.
ence (6.1 versus 4.1 mllyr per pack/day) was not significant (p = 0.18). As reported previously, the mean rate of loss in height-adjusted FEV 1 was somewhat greater in the second than in the first interval (4). Inclusion of an examination effect in the model had no discernable effect « 0.1 mllyr per pack/day) on the estimates of longitudinal coefficients for pack-years of smoking in the longitudinal models. Further analyses were performed to investigate the effects of changes in smoking status between examinations on the change in level of pulmonary function. In the crosssectional model (table 3), current smokers had an additional deficit in FEV 1 of 232 ml in men and 114ml in women relative to former smokers with the same smoking history in pack-years. In the longitudinal analysis, subjects who started smoking in an interval (127 men and 115 women) experienced an additional loss in FEV 1 of 34 (SE = 25) ml for men and 49 (SE = 16) ml for women above that predicted from their cumulative smoking in the interval. Similarly, losses of FVC associated with the start of smoking were estimated to be 47 (SE = 30) ml in men and 39 (SE = 18) ml in women. Smokers who stopped in the interval (349 men and 241 women) regained small amounts of FEV 1 (13 ± 15 ml for men and 21 ± 13 ml for women), but no apparent recovery of FVC ( - 3 ± 17ml for men and 0 ± 15for women). Thus, the longitudinal data indicated a much
The longitudinal analyses reported here show that, in an initial cohort of 8,191 adults followed for as long as 6 yr, current cigarette smokers lose pulmonary function at a rate that depends linearly on the number of cigarettes smoked each day during the interval. Former smokers, however, lose function at the same rate as never smokers. Although the deficits associated with cigarette smoking are, in the main, irretrievable, smokers can limit the effects of smoking by cessation. These results confirm the findings of previous studies as summarized in the 1990 Surgeon General's Report on the Health Effects of Smoking Cessation (6), which concluded from available longitudinal data that those who stopped smoking ceased to lose FEV 1 at an accelerated rate. Cigarette smokers experienced a substantial acceleration in rate of loss of FEV 1 and FVC relative to never and former smokers. The rate depended linearly on the number of cigarettes smoked per day during the interval between examinations (table 2). Very similar values were reported by Fletcher and colleagues (7), in an 8-yr longitudinal study of 792 employed men. In that study, the annual loss of FEV 1 was 36 mllyr for never smokers, 44 mllyr for smokers of < 5 cigarettes/day; 46 mllyr for 5 to 15 cigarettes/day; 54 ml/yr for 15 to 25 cigarettes/day; and 54 mllyr for > 25 cigarettes/day. In the present study, the rate of decline of FEV 1 was found to increase linearly with number of cigarettes smoked, increasing 12.6mllyr per pack/day for a man, and 7.2 mllyr per pack/day for a woman. The longitudinal estimates of the cumulative effects of smoking are somewhat larger than estimates obtained from cross-sectional analysis of the initial pulmonary function examination (7.4 mllyr per pack/day for men and 5.2 mllyr per pack/day for women). These differences are unlikely to be due to model misspecification, but rather they can be explained by two effects. First, it seems likely that random misclassification of smoking history will occur more frequently in the crosssectional analysis than in the longitudinal analysis, which requires recall only for the past 3 yr. The increased misclassification will lead to a more severe underestimate of the effect of smoking in the cross-sectional analyses. Second, the cross-sectional and longitudinal models actually address slightly different questions. The cross-sectional data describe the change in the population mean value of a pulmonary function variable as a function of age and pack-years of cigarette smoking. The longitudinal data, however, describe the effects of changes in age and pack-years of cigarette smoking on pulmonary function level in those who completed a follow-up interval. Because initial pulmonary function
level is a risk factor for mortality, morbidity, and failure to provide an acceptable spirometric examination (8, 9), there will be selective attrition of subjects with low pulmonary function. The change in pulmonary function level for those who are alive and able to perform the spirometric examination at both ages will be larger than the difference in population mean pulmonary function levels at two ages or values of pack-years. Consequently, it would be expected that longitudinal estimates of average rate of loss associated with both aging and cumulative pack-years of cigarette smoking will be somewhat larger than cross-sectional estimates of rate of change in the population mean pulmonary function level. Clinicians who wish to evaluate the rate of pulmonary function loss of an individual patient should use longitudinal information as the source of norms. Although cohort effects resulting from changing patterns and types of cigarette consumption may be also important, the available evidence is not sufficient to address the issue. At the initial examination, current smokers had lower levels of FE VI than did ex-smokers with the same lifetime smoking burden, 232 mllower among men and 114 mllower among women. These differences would appear to reflect a reversible component of the deficit caused by cigarette smoking. Men and women who started smoking between examinations experienced declines 34 ± 25 and 49 ± 16 ml larger, respectively, than did continuing smokers with the same increment in lifetime pack-years. Those who quit smoking between examinations experienced declines 13 ± 15 and 21 ± 13 ml smaller for men and women, respectively, than did continuing smokers with the same number of packs between examinations. Thus, the longitudinal analysis supports the earlier finding of a reversible component of the cigarette smoking effect, but it suggests that such effect is substantially smaller than the earlier cross-sectional estimate. Buist and colleagues (10) have shown that pulmonary function improved after cessation of smoking for 6 to 8 months, with no further improvement observed at 30 months. Given that there are only 3 yr between examinations in this study and the time since cessation was not known, the reduced longitudinal estimate of the acute benefit of cessation may be due to the time since cessation being too short. Camilli and coworkers (11) examined rates of decline of FEV 1 in a population sample of 1,705 adults in Tucson, AZ, with a mean follow-up of 9.4 yr. The age-specific rates of loss suggested that the benefits of cessation may be greatest among the youngest smokers who have the shortest smoking histories, suggesting that the early effects of smoking are relativelyreversible and could represent in part a bronchoconstrictive effect. Similar results were found in this study; 72 men and 48 women who quit smoking between 25 and 34 yr of age experienced mean increases of FEV 1 between examinations, whereas all other smoking groups irrthe same age category experienced declines in FEV 1.
1348
BRIEF COMMUNICATION
References 1. Dockery DW, Speizer FE, Ferris BG Jr, Ware JH, Louis TA, Spiro A. Cumulative and reversible effects of lifetime smoking on simple tests of lung function in adults. Am Rev Respir Dis 1988; 137:286-92. 2. Dockery DW, Ware JH, Ferris BG Jr, et al. Distribution of FEV 1 and FVC in healthy, white adult never-smokers in six U.S. cities. Am Rev Respir Dis 1985; 131:511-20. 3. Ferris BG Jr, Speizer FE, Spengler JD, et al. Effects of sulfur oxides and respirable particles on human health: methodology and demography of population in study. Am Rev Respir Dis 1979; 120:767-79.
4. Ware JH, Dockery DW, Louis TA, Xu X, Ferris BG Jr, Speizer FE. Longitudinal and cross-sectional estimates of pulmonary function decline in neversmoking adults. Am J Epidemiol 1990; 132: 685-700. 5. Liang KY, Zeger SL. Longitudinal data analysis using generalized linear models. Biometrika 1986; 73:13-22. 6. U.S. Department of Health and Human Services. The health benefits of smoking cessation. Bethesda, MD: USDHHS, 1990. (Publication no. DHHS [CDC] 90-8416). 7. Fletcher C, Peto R, Tinker C, Speizer FE. The natural history of chronic bronchitis and emphysema. London: Oxford University Press, 1976. 8. Beaty TH, Newill CA, Cohen BH, et al. Ef-
fects of pulmonary function on mortality. J Chronic Dis 1985; 38:703-10. 9. Eisen EA, Wegman DE, Louis TA. Effects of selection in a prospective study of forced expiratory volume in Vermont granite workers. Am Rev Respir Dis 1983; 128:587-91. 10. Buist AS, Nagy JM, Sexton GT. The effect of smoking cessation on pulmonary function: a 30-month follow-up of two smoking cessation clinics. Am Rev Respir Dis 1979; 120:953-7. 11. Camilli AB, Burrows B, Knudson RJ, LyleSK, Lebowitz MD. Longitudinal changes in forced expiratory volume in one second in adults: effects of smoking and smoking cessation. Am Rev Respir Dis 1987; 135:794-9.
XIPING
xu,
DOUGLAS W. DOCKERY, JAMES H. WARE, FRANK E. SPEIZER, and BENJAMIN G. FERRIS, JR.
As a part of the Six Cities Study of Air Pollution and Health, we selected a random sample of adult residents of Six Cities in the eastern United States and examined them at 3-yr intervals over a 6-yr period (1). In a previous report, we described analyses of the baseline spirometric data showing that levels of FVC and FEV 1 in both current and former smokers depended linearly on total pack-years of cigarette smoking. Current cigarette smokers had an additional deficit relative to former smokers with the same smoking history as measured in pack -years, This report uses data obtained during a 6-yr follow-up of this cohort to obtain longitudinal estimates of the effects of current smoking behavior on the rate of loss of pulmonary function between examinations. The sampling design of the original study has been described in detail elsewhere (2-4). Briefly, a random sample of adults 25 to 74 yr of age was drawn in each of six U.S. cities (Watertown, MA; KingstonHarriman, TN; Steubenville, OH; a geographically defined portion of St. Louis, MO; Topeka, KA; and Portage, WI) beginning in September 1974. Participants completed a standardized respiratory disease questionnaire and performed a forced expiratory maneuver on a water-filled 8-L recording survey spirometer (Warren E. Collins, Braintree, MA) while in the sitting position, without noseclip (2, 3). The means of the best three values of FEV 1 and FVC, corrected to BTPS, were determined. Each participant was invited back for a follow-up examination at the same time of year 3 and 6 yr after the initial examination. The data for each subject were expressed as the initial pulmonary function measurements and the changes in pulmonary function level during the 3-yr intervals between follow-up examinations. Thus, each subject contributed one observation to the cross-sectional analysis of initial measurements and up to two observations (intervals) to the longitudinal analysis of changes in pulmonary function during the 6 yr of follow-up. Both smoking history and smoking during intervals between examinations were measured in pack-years. Never smokers reported lifetime smoking of less than 25 packs and no cigarette smoking during an interval. Former smokers reported a smoking history but no smoking during an interval. Continuing smokers reported smoking throughout the interval, quitters were smoking at the beginning but stopped in the interval, and starters did not smoke at the beginning but started during the interval. The cross-sectional effects of age and smoking on FEV 1 and FVC at the initial visit, and effects of these variables on change in FEV 1 and FVC be-
SUMMARY Data from a random sample of 8,191 men and women selected In six U.S. cities and examined on three occasions over a 6-yr follow-up period were analyzed by longitudinal methods to describe the effects of smoking history and current smoking behavior on rate of loss of pulmonary function during adult life. Former smokers had age- and height-adjusted rates of decline (34.3 ml/yr for men and 29.6 ml/yr for women) comparable with those of never smokers (37.8mllyr for men and 29.0 ml/yr for women) but much smaller than those of continuing smokers (52.9 ml/yr for men and 38.0 ml/yr for women). The accelerated rate of loss of FEV1 among smokers depended linearly on the number of cigarettes smoked per day during the interval between examinations. The estimated Increase in rate of loss associated with smoking was 12.6 ml/yr per pack/day for men and 7.2 ml/yr per pack/day for women. These longitudinal estimates of the effects of smoking were approximately 50% larger than estimates obtained from cross-sectional analysis of the initial pulmonary function examination. Men who started smoking had accelerated rates of loss (55.9 ml/yr) as did women (43.1 ml/yr). ,Smokers who stopped smoking between examinations had reduced declines (41.2 ml/yr for men and 28.7 ml/yr for women) compared with continuing smokers. The age-specific rates of loss suggest that the benefits of cessation may be greatest among the youngest smokers. The changes associated with starting or cessation of smoking were much smaller than those found In cross-sectional analyses of these data. In summary, cigarette smoking causes an accelerated annual loss of pulmonary function that may be underestimated by cross-sectional studies. Those who stop smoking will experience only a small recovery in pulmonary function level, but they will cease to lose pulmonary function at an accelerated rate. AM REV RESPIR DIS 1992; 146:1345-1348
tween examinations, were estimated simultaneously using methods described previously (4). The crosssectional model for the initial measurement for ith subject, can be written as PFit = b, + b 1*(Agecit) + b 2*(Agecit)2 + b 3*(Pack-yearit) + b 4*(Current smokingj.) + eit, where PFit = height-adjusted FEV 1 or FVC for subject i at visit 1, Agej, = age of the ith subject at visit 1 (Agecj, = Agej, - 50), Pack-yean 1 = lifetime pack-years of cigarette smoking at visit 1, Current smokingi, = 1 (smoker at visit 1), 0 (otherwise), and eit = error in the regression equation. Two longitudinal models were considered in this study. The more general model can be expressed as Dit = b'l*(AgeCi,t+1 - Agecjj) + b'2*(Ageci,t+12 - Agecjr') + b'3*(Packyean,t+1 - Packyearjj) + b'4*(Start smokingjj) + b's*(Stop smokingjj) + eit, where Dit = PFi,t+l - PFit, t = 1,2, Start smokingjj, is an indicator variable equal to 1 if the participant started smoking in the interval and o otherwise, and Stop smoking is defined analogously. This longitudinal model provides estimates of the effects of age and change in pack-years of smoking that are directly comparable to the coefficients of the cross-sectional model. The two indicator variables for change in smoking status allow separate estimation of the effects of initiation and cessation of smoking. In this regard, the longitudinal model is more general than the cross-sectional model. In a separate analysis, intervals during which participants reported a change in smoking status were deleted from the analysis, and the indicator
variables for change in smoking status were dropped from the model. The regression coefficients and variances of the estimated coefficients of the crosssectional and longitudinal models were estimated simultaneously by the method of generalized estimating equations (5). The data for men and women were analyzed separately. Effects of cigarette smoking are expressed as effects per pack/day. Age-adjusted rates were based on the age distri-
(Receivedin originalform December30, 1991 and in revised form April 6, 1992) 1 From the Environmental Epidemiology Program, Department of Environmental Health, and the Department of Biostatistics, Harvard School of Public Health, and the Channing Laboratory, Brigham and Women's Hospital, Harvard Medical School, Boston, Massachusetts. 2 Supported in part by Grants ES-Oll08 and ES-OOOO2 from the National Institute of Environmental Health Sciences, by Cooperative Agreement CR-811650 from the Environmental Protection Agency, and by Contract RP-l00l with the Electric Power Research Institute. 3 Correspondence and requests for reprints should be addressed to Douglas W. Dockery, ScD, Environmental Epidemiology Program, Department of Environmental Health, Harvard School of Public Health, 655 Huntington Avenue, Boston, MA 02115.
1345
1346
BRIEF COMMUNICATION
TABLE 1 AVERAGE RATE OF CHANGE OF FEV1 (ml/yr) BY SEX, AGE, AND SMOKING GROUPS DURING THE INTERVAL
Age (yr) Men 25-34 35-44 45-54 55-64 65-78 n Women 25-34 35-44 45-54 55-64 65-78 n
Continuous Smokers (cig/day)
(n)
Never Smoker
Former Smoker
Total
< 15
15-24
25+
Starter
Quitter
878 977 1,496 1,247 839 5,437
-15.2 -24.9 -41.9 -45.9 -55.6 1,475
-6.4 -23.9 -39.4 -43.2 -52.3 1,842
-34.2 -34.5 -63.8 -60.4 -63.3 1,605
-36.5 30.4 -55.0 -49.0 -66.8 218
-20.6 -40.3 -57.4 -53.8 -54.1 519
-41.7 -39.0 -68.8 -68.5 -73.4 868
-19.4 -73.1 -56.3 -65.4 -58.6 349
6.2 -47.4 -32.2 -66.7 -60.2 127
1,039 1,158 1,823 1,557 1,066 6,643
-13.8 -25.0 -30.4 -34.1 -39.2 3,449
-13.9 -29.8 -26.0 -40.6 -34.9 977
-22.2 -34.9 -40.1 -48.3 -38.2 1,837
-16.5 -28.9 -32.5 -42.7 -29.4 486
-21.9 -35.8 -45.7 -52.0 -47.9 721
-28.0 -36.9 -37.8 -50.4 -37.4 630
-29.1 -61.3 -43.0 -52.8 -22.4 241
7.4 -6.2 -31.7 -38.9 -70.1 115
TABLE 2 AGE-ADJUSTED AVERAGE RATE OF CHANGE OF FEV1 AND FVC (mllyr) BY SEX AND SMOKING GROUPS DURING THE INTERVAL*
FEV1 Men Women FVC Men Women
Continuous Smokers (cig/day)
Never Smoker
Former Smoker
Total
< 15
15-24
25+
Starter
Quitter
-37.8 (2.0)
-34.3 (1.8)
-52.9 (2.0)
-37.4 (6.0)
-47.2 (3.6)
-59.9 (3.5)
-55.9 (9.6)
-41.2 (5.0)
-29.0 (0.9)
-29.6 (1.6)
-38.0 (1.2)
-31.2 (2.4)
-42.0 (1.8)
-38.9 (2.5)
-43.1 (5.7)
-28.7 (4.3)
-43.8 (2.1)
-43.4 (2.1)
-59.8 (2.5)
-58.6 (7.2)
-51.9 (4.3)
-63.8 (4.4)
-70.4 (11.8)
-49.3 (6.1)
-32.9 (1.0)
-34.0 (1.8)
-40.1 (1.4)
-35.2 (2.7)
-42.9 (2.2)
-40.6 (2.9)
-39.5 (6.9)
-39.9 (5.2)
• Standard errors, shown in parentheses, are calculated by robust methods (5).
TABLE 3 CROSS-SECTIONAL (CS) AND LONGITUDINAL (LN) ESTIMATES OF THE EFFECTS OF AGE AND CIGARETIE SMOKING ON HEIGHT-ADJUSTED FEV1 AND FVC FOR MEN AND WOMEN* (Age - 50) (mllyr)
(Age - 50)2 (mllyr 2 )
Pack-Years (ml/yr/pack/day)
CS LN
3,502 (16)
- 34.6 (0.8)* - 34.1 (1.2)
- 0.16 (0.05) - 0.47 (0.04)
-7.4 (0.5) -12.6 (1.5)
-232 (22)
CS LN
2,518 (9)
-27.8 (0.4) -28.8 (0.7)
- 0.10 (0.03) - 0.26 (0.03)
-5.2 (0.5) -7.2 (1.2)
-114(16)
CS LN CS LN
4,398 (18)
-34.1 -40.5 -27.9 -31.3
-
Model FEVl Men Women FVC Men Women
Current Smoking (ml)
Intercept (ml)
3,118 (11)
(0.9) (1.4) (0.5) (0.8)
0.17 0.70 0.22 0.44
(0.06) (0.05) (0.04) (0.03)
-6.0 -10.9 -4.1 - 6.1
(0.6) (1.7) (0.6) (1.4)
-159 (24) -62 (18)
• Standard errors are shown in parentheses.
bution of the sample in decades. All pulmonary function values were corrected for body size by dividing by the square of the subject's height, measured at the initial visit (2). Regression coefficients and fitted values were converted to volume units by multiplying by the square of the mean height, 1.73 m for men and 1.61 m for women.
The initial sample consisted of 8,191 white adults, 6,508 (79070) of whom were reexamined 3 yr later, and 5,572 (68%) after 6 yr. Loss-to-follow-up was 38% among men and 31% among women. Current smokers had slightly higher loss-to-follow-up rates (35%
with incomplete follow-up) than either never smokers (29%) or former smokers (31%). Womenweremore likelyto report neversmoking (52%) than were men (27%), whereas a higher percentage of men (34%) than women (15%) reported being former smokers. During follow-up, 6.4% of the men and 3.6% of the women reported that they had quit smoking between examinations, 2.3% of men and 1.7% of women reported starting smoking. Average annual rates of decline in FEV 1 increased continuously with age among never smokers and younger smokers for both men and women (table 1).The "flattened" declines observed in the oldest cigarette smokers may be explained by small sample size, survival effects, and loss-to-follow-up. Sex-specific age-standardized rates of loss of FEV 1 increased with the average number of cigarettes smoked per day during the interval (table 2). Former smokers had rates of loss comparable to those of never smokers. Smokers who quit had average rates of loss slightly greater than those of never smokers and former smokers, but still significantly smaller than the rates for smokers (table 2). Nonsmokers who started had rates of loss greater than never smokers, former smokers, and all but the heaviest smokers. Sex-specific longitudinal regression analyses of the FEV 1 changes for intervals in which participants did not change smoking status showed that current cigarette smoking was associated with an increase in the rate of loss of FEV 1 of 12.6 ml/yr per pack/day in men and 7.2 mllyr per pack/day in women (table 3). Corresponding estimates of the effects of cigarette smoking history obtained from cross-sectional analysis of the pulmonary function measurements at the initial examination were 7.4 ml per pack-year in men and 5.2 ml per pack-year in women. Thus, the cross-sectional estimates of the effect of smoking are 59% and 72% of the longitudinal estimates for men and women, respectively. Among men, the difference between the cross-sectional and longitudinal estimates was highly significant (p < 0.001), but the difference did not reach statistical significance among women (p = 0.13). The observed sex-specific rates of change of FEV 1 versus packs/day of cigarette smoking among ever-smokers (0 for former smokers and tertiles of packs/day for continued smokers) along with the predicted means for each group, which were calculated from the above fitted longitudinal model using the age and smoking variables described above are shown in figure 1. The curvature of the predicted values is due to the differences in age between the smoking groups. A similar analysis was performed for the initial measurements and changes between examinations in height-adjusted FVC (table 3). For men, the longitudinal and cross-sectional estimates of the effect of cigarette smoking on rate of loss (10.9 versus 6.0 ml/yr per pack/day, respectively) were significantly different (p = 0.01). For women, the differ-
1347
BRIEF COMMUNICATION
smaller change in pulmonary function associated with either smoking cessation or initiation than the estimated effect of current smoking than in the cross-sectional model would suggest.
-25
-30
....
»
-35
~
-40
rLI
c... ~
U v: ~ Q
-45
-50
,
~ rLI
o
;Z;
.,;
-55
Men
::r::
u
~
rLI
-60
;::;1
-65
-70
MEAN PACKS / PER DAY
Fig. 1. Predicted and observed annual rates of change of height-adjusted FEV, for men and women. The circles and error bars represent the mean ± 95% confidence interval of the observed annual rate of loss of height-adjusted FEV, for subjects grouped by former smokers (mean packs/day = 0) and continued smokers divided into three subgroups by tertiles of mean packs/day.The dashed line connects the means predicted from the longitudinal model.
ence (6.1 versus 4.1 mllyr per pack/day) was not significant (p = 0.18). As reported previously, the mean rate of loss in height-adjusted FEV 1 was somewhat greater in the second than in the first interval (4). Inclusion of an examination effect in the model had no discernable effect « 0.1 mllyr per pack/day) on the estimates of longitudinal coefficients for pack-years of smoking in the longitudinal models. Further analyses were performed to investigate the effects of changes in smoking status between examinations on the change in level of pulmonary function. In the crosssectional model (table 3), current smokers had an additional deficit in FEV 1 of 232 ml in men and 114ml in women relative to former smokers with the same smoking history in pack-years. In the longitudinal analysis, subjects who started smoking in an interval (127 men and 115 women) experienced an additional loss in FEV 1 of 34 (SE = 25) ml for men and 49 (SE = 16) ml for women above that predicted from their cumulative smoking in the interval. Similarly, losses of FVC associated with the start of smoking were estimated to be 47 (SE = 30) ml in men and 39 (SE = 18) ml in women. Smokers who stopped in the interval (349 men and 241 women) regained small amounts of FEV 1 (13 ± 15 ml for men and 21 ± 13 ml for women), but no apparent recovery of FVC ( - 3 ± 17ml for men and 0 ± 15for women). Thus, the longitudinal data indicated a much
The longitudinal analyses reported here show that, in an initial cohort of 8,191 adults followed for as long as 6 yr, current cigarette smokers lose pulmonary function at a rate that depends linearly on the number of cigarettes smoked each day during the interval. Former smokers, however, lose function at the same rate as never smokers. Although the deficits associated with cigarette smoking are, in the main, irretrievable, smokers can limit the effects of smoking by cessation. These results confirm the findings of previous studies as summarized in the 1990 Surgeon General's Report on the Health Effects of Smoking Cessation (6), which concluded from available longitudinal data that those who stopped smoking ceased to lose FEV 1 at an accelerated rate. Cigarette smokers experienced a substantial acceleration in rate of loss of FEV 1 and FVC relative to never and former smokers. The rate depended linearly on the number of cigarettes smoked per day during the interval between examinations (table 2). Very similar values were reported by Fletcher and colleagues (7), in an 8-yr longitudinal study of 792 employed men. In that study, the annual loss of FEV 1 was 36 mllyr for never smokers, 44 mllyr for smokers of < 5 cigarettes/day; 46 mllyr for 5 to 15 cigarettes/day; 54 ml/yr for 15 to 25 cigarettes/day; and 54 mllyr for > 25 cigarettes/day. In the present study, the rate of decline of FEV 1 was found to increase linearly with number of cigarettes smoked, increasing 12.6mllyr per pack/day for a man, and 7.2 mllyr per pack/day for a woman. The longitudinal estimates of the cumulative effects of smoking are somewhat larger than estimates obtained from cross-sectional analysis of the initial pulmonary function examination (7.4 mllyr per pack/day for men and 5.2 mllyr per pack/day for women). These differences are unlikely to be due to model misspecification, but rather they can be explained by two effects. First, it seems likely that random misclassification of smoking history will occur more frequently in the crosssectional analysis than in the longitudinal analysis, which requires recall only for the past 3 yr. The increased misclassification will lead to a more severe underestimate of the effect of smoking in the cross-sectional analyses. Second, the cross-sectional and longitudinal models actually address slightly different questions. The cross-sectional data describe the change in the population mean value of a pulmonary function variable as a function of age and pack-years of cigarette smoking. The longitudinal data, however, describe the effects of changes in age and pack-years of cigarette smoking on pulmonary function level in those who completed a follow-up interval. Because initial pulmonary function
level is a risk factor for mortality, morbidity, and failure to provide an acceptable spirometric examination (8, 9), there will be selective attrition of subjects with low pulmonary function. The change in pulmonary function level for those who are alive and able to perform the spirometric examination at both ages will be larger than the difference in population mean pulmonary function levels at two ages or values of pack-years. Consequently, it would be expected that longitudinal estimates of average rate of loss associated with both aging and cumulative pack-years of cigarette smoking will be somewhat larger than cross-sectional estimates of rate of change in the population mean pulmonary function level. Clinicians who wish to evaluate the rate of pulmonary function loss of an individual patient should use longitudinal information as the source of norms. Although cohort effects resulting from changing patterns and types of cigarette consumption may be also important, the available evidence is not sufficient to address the issue. At the initial examination, current smokers had lower levels of FE VI than did ex-smokers with the same lifetime smoking burden, 232 mllower among men and 114 mllower among women. These differences would appear to reflect a reversible component of the deficit caused by cigarette smoking. Men and women who started smoking between examinations experienced declines 34 ± 25 and 49 ± 16 ml larger, respectively, than did continuing smokers with the same increment in lifetime pack-years. Those who quit smoking between examinations experienced declines 13 ± 15 and 21 ± 13 ml smaller for men and women, respectively, than did continuing smokers with the same number of packs between examinations. Thus, the longitudinal analysis supports the earlier finding of a reversible component of the cigarette smoking effect, but it suggests that such effect is substantially smaller than the earlier cross-sectional estimate. Buist and colleagues (10) have shown that pulmonary function improved after cessation of smoking for 6 to 8 months, with no further improvement observed at 30 months. Given that there are only 3 yr between examinations in this study and the time since cessation was not known, the reduced longitudinal estimate of the acute benefit of cessation may be due to the time since cessation being too short. Camilli and coworkers (11) examined rates of decline of FEV 1 in a population sample of 1,705 adults in Tucson, AZ, with a mean follow-up of 9.4 yr. The age-specific rates of loss suggested that the benefits of cessation may be greatest among the youngest smokers who have the shortest smoking histories, suggesting that the early effects of smoking are relativelyreversible and could represent in part a bronchoconstrictive effect. Similar results were found in this study; 72 men and 48 women who quit smoking between 25 and 34 yr of age experienced mean increases of FEV 1 between examinations, whereas all other smoking groups irrthe same age category experienced declines in FEV 1.
1348
BRIEF COMMUNICATION
References 1. Dockery DW, Speizer FE, Ferris BG Jr, Ware JH, Louis TA, Spiro A. Cumulative and reversible effects of lifetime smoking on simple tests of lung function in adults. Am Rev Respir Dis 1988; 137:286-92. 2. Dockery DW, Ware JH, Ferris BG Jr, et al. Distribution of FEV 1 and FVC in healthy, white adult never-smokers in six U.S. cities. Am Rev Respir Dis 1985; 131:511-20. 3. Ferris BG Jr, Speizer FE, Spengler JD, et al. Effects of sulfur oxides and respirable particles on human health: methodology and demography of population in study. Am Rev Respir Dis 1979; 120:767-79.
4. Ware JH, Dockery DW, Louis TA, Xu X, Ferris BG Jr, Speizer FE. Longitudinal and cross-sectional estimates of pulmonary function decline in neversmoking adults. Am J Epidemiol 1990; 132: 685-700. 5. Liang KY, Zeger SL. Longitudinal data analysis using generalized linear models. Biometrika 1986; 73:13-22. 6. U.S. Department of Health and Human Services. The health benefits of smoking cessation. Bethesda, MD: USDHHS, 1990. (Publication no. DHHS [CDC] 90-8416). 7. Fletcher C, Peto R, Tinker C, Speizer FE. The natural history of chronic bronchitis and emphysema. London: Oxford University Press, 1976. 8. Beaty TH, Newill CA, Cohen BH, et al. Ef-
fects of pulmonary function on mortality. J Chronic Dis 1985; 38:703-10. 9. Eisen EA, Wegman DE, Louis TA. Effects of selection in a prospective study of forced expiratory volume in Vermont granite workers. Am Rev Respir Dis 1983; 128:587-91. 10. Buist AS, Nagy JM, Sexton GT. The effect of smoking cessation on pulmonary function: a 30-month follow-up of two smoking cessation clinics. Am Rev Respir Dis 1979; 120:953-7. 11. Camilli AB, Burrows B, Knudson RJ, LyleSK, Lebowitz MD. Longitudinal changes in forced expiratory volume in one second in adults: effects of smoking and smoking cessation. Am Rev Respir Dis 1987; 135:794-9.
