PREVENTIVE
MEDICINE
Pulmonary
20, 621-637 (1991)
Function
in Relation to Cigarette Smoking Smoking Cessation’
and
MARY C. TOWNSEND, DR.PH.,* ALAIN G. DUCHENE,~ JULIET MORGAN, PH.D.,+ AND WARREN S. BROWNER, M.D., M.P.H.§
For the MRFIT
Research Group*
*Department of Epidemiology, Graduate School of Public Health, University of Pittsburgh, Pennsylvania 15260; tDivision of Biostatistics, School of Public Health, Minnesota, Minneapolis, Minnesota 554S5; $Department of Medicine, University Chicago, Illinois 60637; and JGeneral Internal Medicine Section, VA Medical San Francisco, California 94121
Pittsburgh, University of of Chicago, Center,
Background. More than half of the subjects in the MRFIT smoked at baseline and 10% of the subjects stopped smoking permanently during the first year of the trial. In this report, rates of decline in forced expiratory volume in 1 set (FEV,) are compared for early permanent quitters and smokers who continued to smoke throughout the trial. Methods. Since pulmonary function testing was not standardized across all centers until the third annual visit cycle, change in FEV, is examined over the latter half of the trial; the level of FEV, is analyzed cross-sectionally at the midpoint of the trial. Analyses are limited to 4,926 subjects who never used beta-blockers or smoked cigars, cigarillos, or pipes during the trial and who had annual FEV,s measured over 24 years in the latter half of the trial. Results. Quitters during the first 12 months experienced smaller declines in FEV, over the latter half of the trial than continuing smokers, with - 50.7 ml/year versus - 59.0 ml/year, respectively, adjusted for the level of FEV, (P = 0.05). Cross-sectionally, those who had never smoked, former smokers, quitters, and continuing smokers showed a gradient of decreasing FEV,, and all four smoking groups were significantly different from each other (P < 0.05). Conclusions. These data suggest that if a middle-aged, healthy smoker stopped smoking permanently, he could expect his FEV, to deteriorate at a more gradual rate 34 years after stopping smoking than a similar smoker who continued to smoke. No information was available for the complete MRFIT cohort on the pulmonary function effects immediately following smoking cessation. 0 1991 Academic PXSS, 1~.
INTRODUCTION
The effects on pulmonary function of never having smoked, quitting smoking, and continued smoking have been investigated in cross-sectional and longitudinal epidemiological studies (l-7). Cross-sectionally, those who had never smoked and former and continuing smokers exhibit a gradient of decreasing levels of pulmonary function (14). Longitudinally, pulmonary function declines at a more gradual rate in those who had never smoked and former smokers than in continuing smokers (3-7). This difference in rates of decline is more pronounced for those with airway obstruction than for those with normal levels of pulmonary function (3).
r Address reprint requests to the Coordinating Centers for Biometric Research, 2221 University Avenue SE., Suite 200, Minneapolis, MN 55414-3080. * See credit roster for complete listing. 621 0091-7435191 $3.00 Copyright Q 1991 by Academic Press, Inc. All rights of reproduction in any form reserved.
622
TOWNSEND
ET AL.
The Multiple Risk Factor Intervention Trial (MRFIT), a multiple-center randomized study of the effects of intervention on major heart disease risk factors, provides a large cohort in which the effects of smoking cessation can be examined further. More than half of the MRFIT participants smoked at baseline, annual follow-up visits were conducted for 6-7 years, and smoking cessation was strongly emphasized for the half of the cohort that was randomized into the special intervention (SI) group. The present study examines the effects of smoking and smoking cessation on changes in pulmonary function over time among the MRFIT participants. In particular, subjects who stopped smoking in the first 12 months of the study and abstained for the remainder of the trial were compared with those who continued to smoke throughout the trial. METHODS
The Study Population In 1973-1974, 12,866 men, age 35-57 years, were recruited into the MRFIT at 22 clinical centers in the United States. The men were selected because they were free of heart disease but were at high risk of developing it due to the presence of some combination of high blood pressure (BP), high serum cholesterol level (TC), and cigarette smoking. Subjects were randomly assigned into the usual care (UC) or special intervention groups; risk factor modification, including smoking cessation, dietary changes, and use of antihypertensive medications, was strongly emphasized in the SI group. Because pulmonary function may be affected by antihypertensive beta-blocker medications (8), the population of interest in the present study was limited to men who did not use beta-blockers during the trial. In addition, to avoid attributing possible cigar-, cigarillo-, or pipe-related pulmonary function effects to cigarette smoking, the present study population was also limited to men who did not smoke cigars, cigarillos, or pipes during the trial. For a member of this study population to be included in the analysis, both an assessment of cigarette smoking status and measurements of pulmonary function were required. Smoking
Status
Smoking status was determined longitudinally based on self-reporting at all visits attended. In this study, participants were categorized into subjects who: (a) had never smoked cigarettes; (b) stopped smoking cigarettes prior to the trial; (c) smoked cigarettes at either screens 1 or 3 and quit smoking permanently during the first 12 months; (d) continued to smoke cigarettes throughout the trial; or (e) smoked cigarettes intermittently throughout the trial. Subjects who reported not smoking at both screens 1 and 3 but reported smoking at any annual visit were classified as “other.” Participants were also dichotomized into “never” or “ever” smokers of cigars, cigarillos, or pipes during the trial. In an intervention trial such as the MRFIT, biochemical validation of smoking cessation was desirable since smokers might falsely report quitting smoking. The effects of smoking cessation on pulmonary function would be estimated inaccurately if smokers who falsely reported cessation were included in the group of
FORUM:
CIGARETTE
SMOKING
IN
THE
MRFIT
623
quitters. Therefore, for those who reported quitting within the first 12 months, a SCN-adjusted smoking status was defined in addition to the reported cigarette smoking status. In this article, if the serum SCN level exceeded 100 p,mol/liter at any annual visit, the subject’s SCN-adjusted smoking status was “intermittent smoker” rather than “quitter by 12 months.” Thiocyanate adjustment was not applied to other cigarette smoking categories. Results based on SCN-adjusted smoking status are noted only if they differ from results based on reported smoking status. Pulmonary
Function
Testing
Pulmonary function testing was conducted in the MRFIT using IO-liter Stead Wells water-filled spirometers (Warren E. Collins, Inc., Braintree, MA) at baseline and at each of the second through seventh annual visits. However, measurements analyzed in this report were made after trial-wide standardization of spirometry procedures, which occurred in the third year of the trial. At the start of the trial, widespread attention had not yet been drawn to the importance of standardized spirometry testing procedures and calibration checks. Thus, while the trial’s cardiovascular expertise resulted in well-developed quality control procedures for measurements of serum TC and BP and the recording of electrocardiograms, the spirometry testing procedures were inadequately standardized. The need for spirometry quality control was not recognized until spurious changes in pulmonary function were observed between the baseline and second annual examinations. As described elsewhere (9), large deviations from the predicted forced vital capacity (FVC) at many centers prompted an investigation into the quality of the MRFIT pulmonary function testing. Procedural problems, such as nonreproducible and incompletely recorded expirations, and technical problems, such as low water levels and leaks in the spirometers, were discovered. Subsequently, the trial’s pulmonary function testing protocol was rewritten, regional training sessions were conducted for technicians, and extensive quality control procedures were instituted at the coordinating center level (10). However, the initial poor quality control had an adverse effect on many spirograms from the first 3 years of the trial. Therefore, analysis of MRFIT pulmonary function data from all centers combined was limited to the time period following national standardization of the testing technique, from the latter half of the third annual visit cycle to the end of the trial. Following standardization of the national MRFIT spirometry testing, three to five forced expirations were recorded per test (1 l), and reproducibility within 200 ml was sought for the highest recorded forced expiratory volume in 1 set (FEV,) and for the highest FVC (12). All spirometric measurements were corrected to body temperature and pressure saturated (BTPS). For the present study, the highest BTPS-corrected FEV, recorded in a testing session was analyzed if it was reproduced within 200 ml and if the highest and next highest FEV, tracings had abrupt takeoffs, were free of cough/hesitation, and were recorded after the kymograph drum had begun rotating (10).
624
TOWNSEND
Methods for Evaluating
Pulmonary
ET AL.
Function
The initial values of FEV, used in these analyses were the first acceptable FEV, readings after standardization in Year 3. Thus, the initial FEV, was observed near the midpoint of the trial. For this reason, the initial FEV, value used in the analyses is referred to as FEV, at the midpoint of the trial. These midpoint FEV, readings were analyzed cross-sectionally to determine whether the expected association between the level of FEV, and past or current cigarette smoking was observed in the MRFIT. FEV, was also examined from the first acceptable reading after Year 3 to the end of the trial. The longitudinal analysis was performed to determine whether the rate of change in FEV, over the latter half of the trial was modified by permanent smoking cessation in the first year of the trial. The cross-sectional analysis of FEV, was performed by comparing observed FEV,s with predicted values from an internal comparison group. The comparison group was selected from 990 participants who had never used beta-blockers or smoked cigarettes, cigars, cigarillos, or pipes during the trial. Thirty-eight percent of these subjects were excluded from the comparison group based on race or respiratory symptoms: 97 had missing respiratory symptom data, an additional 103 were nonwhite, and an additional 179 had respiratory symptoms. The remaining 611 white participants included in the comparison group were free of respiratory symptoms and had reproducible, good quality FEV, data from measurements taken at the midpoint of the trial. At the time of FEV, measurement, the subjects ranged in age from 38 to 63 years, with a mean of 51.5 years; the subject’s heights ranged from 61 to 79 inches, with a mean of 68.9 inches. The MRFIT prediction equation for FEV, was FEV, (ml) = 94.8 height (in.) - 23.7 age (years) - 1,732. The predicted FEV, was race-adjusted for black participants by multiplying the predicted value by 0.85 (13). The deviation from each subject’s appropriate predicted value yielded an FEV, adjusted for age, height, and, if appropriate, race. This deviation from predicted or “observed minus predicted FEV,” is referred to hereafter as the “adjusted FEV,.” Change in FEV, over time was assessed for each subject by calculating a regression of his FEV,s on calendar time, producing an FEV, “slope” for each participant. The slope was analyzed as a single observation describing the individual’s change in pulmonary function over time. Since slope estimates calculated over only a single year of follow-up are very unstable and exhibit a within-group standard deviation that is at least twice as great as that of slopes measured over longer time periods (14, 15), analyses were restricted to subjects who had 2 or more years of follow-up during the latter half of the trial. Because only two to five FEV,s were used to determine a slope, an FEV, that was far removed from a subject’s remaining FEV,s would exert a marked effect on the calculated slope. Therefore, using the approach of Fletcher and colleagues (3), each subject’s values were checked for outliers before his slope was calculated. Subjects with outliers were excluded from the analyses.
FORUM:
CIGARETTE
SMOKING
IN
THE
MRFIT
625
Mean FEV, slopes were compared, first, for quitters and continuing smokers and, subsequently, for all smoking groups. Slopes were stratified by level of FEV, to examine the effect of this variable on the quitter-continuing smoker difference in FEV, decline. To permit comparison with the Fletcher study (3), the level of FEV, was defined for these analyses by FEV,/height3 and was dichotomized into 100 ~molfliter in the absence of cigarette smoking, the proportion was somewhat higher among self-reported quitters, possibly due to false reports of smoking cessation.
FORUM:
CIGARETTE
SMOKING
627
IN THE MRFIT
TABLE 1 LONGITUDINAL SMOKING STATUS OF MRFIT PARTICIPANTS Participants not using cigars, cigarillos, pipes, or beta-blockers
All participants
Cigarette smoking status
Percentage
Number
Percentage
1,817 2,545
14 20
990 1,434
16 23
1,332 2,156 4,557 226 233
10 17 35 2 2
594 831 2,197 74 165
9 13 35 1 3
Number
Never smoker Former smoker Sustained quitter by first 12 months Intermittent smoker Continuing smoker Othef No follow-up Total
12,866
6,285
a Started to smoke during the trial.
Characteristics
of Excluded
Subjects
The analysis group was compared with each of the excluded groups to determine whether they differed for selected baseline and follow-up characteristics. The groups were compared for age and height at baseline, adjusted FEV, at the midpoint of the trial, and proportion of each group smoking at either screen 1 or 3 and, among the smokers, the number of cigarettes per day, duration, and packyears of smoking prior to the MRFIT (Table 2). The proportion of cigarette smokers was lower in the analysis group than in the groups excluded because of smoking cigars, cigarillos, and pipes or having too few annual visits or too few pulmonary function tests (P < 0.001). However, smokers in the analysis group differed from smokers in the excluded beta-blocker group only in their smoking exposure
(76%)’ ‘“0””
No CCP CCP NWW Smokers
NoCCP CCP Former Smokers (Pre-MRFIT) Longitudinal
NOCCP CCP Quitters by First Annual Smoking
(94%)’
(94%)’
m
El
B
1 NOCCP CCP Intermittent Smokers
I
1 I NOCCP CCP Continuing Smokers
Status
FIG. 2. Serum thiocyanate at 12 months by smoking status (nonusers of beta-blockers only) for MRFIT participants included in the analysis group. *Percentage of group greater than 100 pmol/liter.
4,926 58 46.7 90 -287 27.8 35.8 50.4
Total
12,866
65 46.4 90 -323
28.0 35.3 49.9
28.4* 33.1* 47.6*
60 46.7 91 - 360*
2.737
Beta-blocker users Mb
28.0 35.6 50.3
-341*
76* 45.8s 91
3,844
Cigar, cigarillo, pipe smokers W
28.1 36.6 51.5
- 498d
77* 46.3 89
746
Too few visits (C)
27.7 37.8 52.3
-651d
80* 46.0 90
126
Too few PFTS 03
Excluded participants
TABLE 2 PARTICIPANTS WITH ANALYSISGROUPFORSELECTEDCHARACTERISTKY
0 At baseline, except for FEV,, which was measured at midpoint of trial. b Letters correspond to groups presented in Fig. 1. ’ Mean values. d FEV,s available for ~25% of group. * Different from analysis group, P < 0.05.
n % smoking at screen 1 or 3 Age (years) % white FEV, obs - pred (ml) Screen 1 or screen 3 smokers Years smokedc Cigarettes/day’ Pack-yearsc
Analysis group
COMPARISONSOF EXCLUDED MRFIT
27.6 36.5 51.2
- 298
59 46.7 86
416
Too few acceptable PPTS 03
29.2 30.2 44.0
-415
61 47.9 19
71
FEV, outliers (F)
?
2
s % 3
El
FORUM:
CIGARETTE
SMOKING
IN
THE
MRFIT
629
prior to the MRFIT: analysis group members smoked more cigarettes per day for slightly fewer years, resulting in a larger pack-years of exposure than that for the smokers using beta-blockers (P < 0.01). Analysis group subjects were slightly older than the cigar, cigarillo, and pipe smokers (P < O.Ol), but did not differ significantly from those of any excluded group in terms of height or proportion of white participants. Analysis group subjects had a higher adjusted FEV, than beta-blocker users and cigar, cigarillo, and pipe smokers (P < 0.01). Although the difference in adjusted FEV, between the analysis group and subjects with too few annual visits or too few pulmonary function tests was also significant, too few of the excluded subjects in these groups had FEV,s available for this comparison to be meaningful: only 11% of those with too few visits and 25% of those with too few tests had FEV,s available for analysis. FEV,
by Smoking
Status at the Trial Midpoint
FEV, at the midpoint of the trial showed the expected associations with age (r = -0.293) and height (r = 0.384) for ah smoking groups combined. In Table 3, FEV, is presented by smoking status and age for the 4,412 white and the 349 black participants in the analysis group. Since members of other racial groups were few (n = 165) and heterogeneous, they are not presented here. FEV, decreased with increasing age within each smoking group and with increasing smoking exposure within each age group. These patterns were slightly more erratic among blacks than among whites, probably due to the smaller numbers in the cells. The enhanced age effect seen in smokers (r = - 0.382) compared with never smokers (r = -0.252) is probably due to the increase in duration of smoking that accompanies increasing age among smokers (r = 0.804). As expected, blacks had lower FEV,s than whites in all age-smoking status groups. Among never smokers, the average FEV, among blacks was 88% of that among whites. The relationship between the level of pulmonary function and the smoking exposure variables remained after adjustment for age and height, with the correlations generally remaining the same or increasing slightly (data not shown). However, among smokers the correlation between the adjusted FEV, and years smoked (r = -0.212) was lower than the correlation between the unadjusted FEV, and years smoked (r = -0.364). This decrease in association probably again reflects the collinearity between age and years smoked. Correlations between the adjusted FEV, and the smoking exposure variables were strongest when all smoking statuses were considered together. For example, the correlations between the adjusted FEV, and cigarettes per day, years smoked, and serum SCN were - 0.3 11, - 0.329, and - 0.275, respectively, for all smoking groups combined. For smokers alone, the correlations for the same pairs of variables dropped to -0.166, -0.212, and -0.105, respectively. The correlation between the adjusted FEV, and SCN was also lower among never smokers alone, with r = -0.063. Since the never smokers had high FEV,s and low values for the smoking variables, while the smokers had lower FEV,s and high values for the smoking variables, the increased correlation that was observed for all smoking groups com-
630
TOWNSEND TABLE
MEAN
FEV,(ml)
ET AL. 3
AT MIDPOINTOF TRIAL BY AGE,RACE, AND SMOKING PARTICIPANTS INCLUDED IN THE ANALYSIS
STATUS FORMRFIT
Age (years) at time of FEV, measurement 3w5 Smoking status Never smoker White Black Former smoker White Black Sustained quitter by 12 months White Black Intermittent smoker White Black Continuing smoker White Black Othef White Black Total White Black
46-55
56-63
Total
n
Mean FEV,
n
Mean FEV,
n
Mean FEV,
n
Mean FEV,
128 24
3,892 3,345
386 29
3,563 2,%3
222 8
3,375 3,034
736 61
3,563 3,123
151 10
3,818 3,512
606 33
3,442 3,269
369 14
3,284 2,605
1,126 57
3,441 3,149
109 11
3,735 3,384
235 27
3,279
75 6
WI4
2,904
2,685
419 44
3,330 2,994
191 23
3,607 3,389
303 31
3,182 2,887
109 12
2,740 2,473
603 66
3,237 2,987
438 39
3,498 3,064
815 64
3,136 2,864
219 14
2,848 2,556
1,472 117
3,201 2,894
16
3,700 -
29 3
3,399 2,755
11 1
3,358 2,773
56 4
3,477 2,760
1,033 107
3,642 3,272
2,374 187
3,307 2,959
1,005 55
3,123 2,638
4,412 349
3,343 3,004
-
0 Started to smoke or resumed smoking during the trial.
bined probably reflected the correlation between FEV, and the presence or absence of smoking more than it reflected a correlation with the degree of smoking. Figure 3 summarizes the cross-sectional relationship between the adjusted FEV, and past or present cigarette smoking. Means and 95% confidence intervals are presented for the observed minus the predicted, or adjusted, FEV,. Percentages of subjects with low FEV,, i.e., an FEV, more than 1.645 SD below the predicted value, are also given. The adjusted FEV,s for all smoking groups except the intermittent and the continuing smokers were significantly different from each other based on multiple comparisons tests (P < 0.05). A low FEV, was four times more prevalent among continuing smokers than among never smokers, with 24% vs 6%, respectively. Figure 4 presents adjusted FEV, values for quitters by 12 months (Q), intermittent smokers (I), and continuing smokers (C), who all reported smoking at either screen 1 or 3, by pack-years of smoking prior to the MRFIT. As seen in Fig. 4, pack-years was associated with both FEV, and smoking group. Thirty-two percent of quitters smoked 30 or fewer pack-years prior to the MRFIT compared
FORUM: +lOO = E. 5
0
iif 0 ti e at? ii g
-100
I
CIGARETTE
SMOKING
E 5 %
THE
631
MRFIT
El
(6%)'
El
(9%)’
-200
-300
3 -E z
IN
. II (15%)'
-400 El (22%)'
-500 -600
I
Never Smokers
I
1
Former Smokers
Longltudlnal
I
Ouitlers by First Annual
Smoking
Intermittent Smokers
. 0 (24%)' I
Continuing Smokers
Status
3. Adjusted FEV, by smoking status for MRFIT participants in the analysis group: Mean, 95% confidence interval. FEV, taken at trial midpoint. Adjustment is the observed FEV, minus the FEV, predicted by age and height based on asymptomatic MRFIT participants who did not smoke or use beta-blockers. *Percentage of group more than 1.65 standard deviations (797 ml) below predicted FEV,. FIG.
with 15% of continuing smokers. Conversely, 37% of quitters smoked 50 or more pack-years prior to the MRFIT compared with 51% of continuing smokers. Smoking group retained its association with the adjusted FEV, after adjusting for pack-years of smoking in an analysis of covariance. The pack-years adjusted mean values for the adjusted FEV, were -304.6, -417.3, and -434.9 ml for the quitters, intermittent smokers, and continuing smokers, respectively (P < 0.0001). Pack-years was also significantly associated with deviation from the predicted FEV, after adjusting for smoking group (Z’ < 0.0001). Similar results were obtained when years smoked and the initial number of cigarettes per day were used as covariates in place of pack-years. Rate of FEV, Decline by Smoking
Status
When examined for all smoking groups combined, the FEV, slope showed a very low but significant correlation with age, height, and the smoking exposure indices, with r = -0.02 to - 0.05 (P < 0.05). Correlations of less or similar magnitude were seen when the smoking groups were examined separately, but, in general, these correlations were not statistically significant. To assess the effects of smoking cessation on the rate of FEV, decline over the latter half of the trial, FEV, slopes were compared for quitters and continuing smokers, controlling for the level of FEV,. Table 4 shows that quitting smoking has a modest effect on the FEV, slope. An analysis of covariance confirmed this
632 0 -100 E z
-200
2 P
-300
g 0.z
-400
t k 3
-500
: c 3
-600
P g
-700
p” 0
a .’c n B
(n-154)
(n-194,
F1 (n-246)
TOWNSEND
ET AL.
at c goII. u ’ c a
l-l
(n-222)
(“=+X3,
(“-146,
(“.W,
0 l
l
II
(“-473,
,“.162,
-600 -900
30.49
0.10). Finally, FEV, slopes were examined across all smoking groups (Table 5). The smoking groups had different mean FEV, slopes (P < 0.00001) and continuing smokers differed from former and intermittent smokers (P < 0.05) based on a multiple-comparisons test. However, the other smoking groups were not signiticantly different from each other.
FEV, S~o~~(ml/year)~~
LEVELOF
TABLE 4 FEV, FOR MRFIT QUITTERSAND
CONTINUING
SMOKERS
IN
THEANALYSISGROUP
FEV,/height3” Smoking status Quitters by 12 months Continuing smokers Total
2500 ml/m3
MEDICINE
Pulmonary
20, 621-637 (1991)
Function
in Relation to Cigarette Smoking Smoking Cessation’
and
MARY C. TOWNSEND, DR.PH.,* ALAIN G. DUCHENE,~ JULIET MORGAN, PH.D.,+ AND WARREN S. BROWNER, M.D., M.P.H.§
For the MRFIT
Research Group*
*Department of Epidemiology, Graduate School of Public Health, University of Pittsburgh, Pennsylvania 15260; tDivision of Biostatistics, School of Public Health, Minnesota, Minneapolis, Minnesota 554S5; $Department of Medicine, University Chicago, Illinois 60637; and JGeneral Internal Medicine Section, VA Medical San Francisco, California 94121
Pittsburgh, University of of Chicago, Center,
Background. More than half of the subjects in the MRFIT smoked at baseline and 10% of the subjects stopped smoking permanently during the first year of the trial. In this report, rates of decline in forced expiratory volume in 1 set (FEV,) are compared for early permanent quitters and smokers who continued to smoke throughout the trial. Methods. Since pulmonary function testing was not standardized across all centers until the third annual visit cycle, change in FEV, is examined over the latter half of the trial; the level of FEV, is analyzed cross-sectionally at the midpoint of the trial. Analyses are limited to 4,926 subjects who never used beta-blockers or smoked cigars, cigarillos, or pipes during the trial and who had annual FEV,s measured over 24 years in the latter half of the trial. Results. Quitters during the first 12 months experienced smaller declines in FEV, over the latter half of the trial than continuing smokers, with - 50.7 ml/year versus - 59.0 ml/year, respectively, adjusted for the level of FEV, (P = 0.05). Cross-sectionally, those who had never smoked, former smokers, quitters, and continuing smokers showed a gradient of decreasing FEV,, and all four smoking groups were significantly different from each other (P < 0.05). Conclusions. These data suggest that if a middle-aged, healthy smoker stopped smoking permanently, he could expect his FEV, to deteriorate at a more gradual rate 34 years after stopping smoking than a similar smoker who continued to smoke. No information was available for the complete MRFIT cohort on the pulmonary function effects immediately following smoking cessation. 0 1991 Academic PXSS, 1~.
INTRODUCTION
The effects on pulmonary function of never having smoked, quitting smoking, and continued smoking have been investigated in cross-sectional and longitudinal epidemiological studies (l-7). Cross-sectionally, those who had never smoked and former and continuing smokers exhibit a gradient of decreasing levels of pulmonary function (14). Longitudinally, pulmonary function declines at a more gradual rate in those who had never smoked and former smokers than in continuing smokers (3-7). This difference in rates of decline is more pronounced for those with airway obstruction than for those with normal levels of pulmonary function (3).
r Address reprint requests to the Coordinating Centers for Biometric Research, 2221 University Avenue SE., Suite 200, Minneapolis, MN 55414-3080. * See credit roster for complete listing. 621 0091-7435191 $3.00 Copyright Q 1991 by Academic Press, Inc. All rights of reproduction in any form reserved.
622
TOWNSEND
ET AL.
The Multiple Risk Factor Intervention Trial (MRFIT), a multiple-center randomized study of the effects of intervention on major heart disease risk factors, provides a large cohort in which the effects of smoking cessation can be examined further. More than half of the MRFIT participants smoked at baseline, annual follow-up visits were conducted for 6-7 years, and smoking cessation was strongly emphasized for the half of the cohort that was randomized into the special intervention (SI) group. The present study examines the effects of smoking and smoking cessation on changes in pulmonary function over time among the MRFIT participants. In particular, subjects who stopped smoking in the first 12 months of the study and abstained for the remainder of the trial were compared with those who continued to smoke throughout the trial. METHODS
The Study Population In 1973-1974, 12,866 men, age 35-57 years, were recruited into the MRFIT at 22 clinical centers in the United States. The men were selected because they were free of heart disease but were at high risk of developing it due to the presence of some combination of high blood pressure (BP), high serum cholesterol level (TC), and cigarette smoking. Subjects were randomly assigned into the usual care (UC) or special intervention groups; risk factor modification, including smoking cessation, dietary changes, and use of antihypertensive medications, was strongly emphasized in the SI group. Because pulmonary function may be affected by antihypertensive beta-blocker medications (8), the population of interest in the present study was limited to men who did not use beta-blockers during the trial. In addition, to avoid attributing possible cigar-, cigarillo-, or pipe-related pulmonary function effects to cigarette smoking, the present study population was also limited to men who did not smoke cigars, cigarillos, or pipes during the trial. For a member of this study population to be included in the analysis, both an assessment of cigarette smoking status and measurements of pulmonary function were required. Smoking
Status
Smoking status was determined longitudinally based on self-reporting at all visits attended. In this study, participants were categorized into subjects who: (a) had never smoked cigarettes; (b) stopped smoking cigarettes prior to the trial; (c) smoked cigarettes at either screens 1 or 3 and quit smoking permanently during the first 12 months; (d) continued to smoke cigarettes throughout the trial; or (e) smoked cigarettes intermittently throughout the trial. Subjects who reported not smoking at both screens 1 and 3 but reported smoking at any annual visit were classified as “other.” Participants were also dichotomized into “never” or “ever” smokers of cigars, cigarillos, or pipes during the trial. In an intervention trial such as the MRFIT, biochemical validation of smoking cessation was desirable since smokers might falsely report quitting smoking. The effects of smoking cessation on pulmonary function would be estimated inaccurately if smokers who falsely reported cessation were included in the group of
FORUM:
CIGARETTE
SMOKING
IN
THE
MRFIT
623
quitters. Therefore, for those who reported quitting within the first 12 months, a SCN-adjusted smoking status was defined in addition to the reported cigarette smoking status. In this article, if the serum SCN level exceeded 100 p,mol/liter at any annual visit, the subject’s SCN-adjusted smoking status was “intermittent smoker” rather than “quitter by 12 months.” Thiocyanate adjustment was not applied to other cigarette smoking categories. Results based on SCN-adjusted smoking status are noted only if they differ from results based on reported smoking status. Pulmonary
Function
Testing
Pulmonary function testing was conducted in the MRFIT using IO-liter Stead Wells water-filled spirometers (Warren E. Collins, Inc., Braintree, MA) at baseline and at each of the second through seventh annual visits. However, measurements analyzed in this report were made after trial-wide standardization of spirometry procedures, which occurred in the third year of the trial. At the start of the trial, widespread attention had not yet been drawn to the importance of standardized spirometry testing procedures and calibration checks. Thus, while the trial’s cardiovascular expertise resulted in well-developed quality control procedures for measurements of serum TC and BP and the recording of electrocardiograms, the spirometry testing procedures were inadequately standardized. The need for spirometry quality control was not recognized until spurious changes in pulmonary function were observed between the baseline and second annual examinations. As described elsewhere (9), large deviations from the predicted forced vital capacity (FVC) at many centers prompted an investigation into the quality of the MRFIT pulmonary function testing. Procedural problems, such as nonreproducible and incompletely recorded expirations, and technical problems, such as low water levels and leaks in the spirometers, were discovered. Subsequently, the trial’s pulmonary function testing protocol was rewritten, regional training sessions were conducted for technicians, and extensive quality control procedures were instituted at the coordinating center level (10). However, the initial poor quality control had an adverse effect on many spirograms from the first 3 years of the trial. Therefore, analysis of MRFIT pulmonary function data from all centers combined was limited to the time period following national standardization of the testing technique, from the latter half of the third annual visit cycle to the end of the trial. Following standardization of the national MRFIT spirometry testing, three to five forced expirations were recorded per test (1 l), and reproducibility within 200 ml was sought for the highest recorded forced expiratory volume in 1 set (FEV,) and for the highest FVC (12). All spirometric measurements were corrected to body temperature and pressure saturated (BTPS). For the present study, the highest BTPS-corrected FEV, recorded in a testing session was analyzed if it was reproduced within 200 ml and if the highest and next highest FEV, tracings had abrupt takeoffs, were free of cough/hesitation, and were recorded after the kymograph drum had begun rotating (10).
624
TOWNSEND
Methods for Evaluating
Pulmonary
ET AL.
Function
The initial values of FEV, used in these analyses were the first acceptable FEV, readings after standardization in Year 3. Thus, the initial FEV, was observed near the midpoint of the trial. For this reason, the initial FEV, value used in the analyses is referred to as FEV, at the midpoint of the trial. These midpoint FEV, readings were analyzed cross-sectionally to determine whether the expected association between the level of FEV, and past or current cigarette smoking was observed in the MRFIT. FEV, was also examined from the first acceptable reading after Year 3 to the end of the trial. The longitudinal analysis was performed to determine whether the rate of change in FEV, over the latter half of the trial was modified by permanent smoking cessation in the first year of the trial. The cross-sectional analysis of FEV, was performed by comparing observed FEV,s with predicted values from an internal comparison group. The comparison group was selected from 990 participants who had never used beta-blockers or smoked cigarettes, cigars, cigarillos, or pipes during the trial. Thirty-eight percent of these subjects were excluded from the comparison group based on race or respiratory symptoms: 97 had missing respiratory symptom data, an additional 103 were nonwhite, and an additional 179 had respiratory symptoms. The remaining 611 white participants included in the comparison group were free of respiratory symptoms and had reproducible, good quality FEV, data from measurements taken at the midpoint of the trial. At the time of FEV, measurement, the subjects ranged in age from 38 to 63 years, with a mean of 51.5 years; the subject’s heights ranged from 61 to 79 inches, with a mean of 68.9 inches. The MRFIT prediction equation for FEV, was FEV, (ml) = 94.8 height (in.) - 23.7 age (years) - 1,732. The predicted FEV, was race-adjusted for black participants by multiplying the predicted value by 0.85 (13). The deviation from each subject’s appropriate predicted value yielded an FEV, adjusted for age, height, and, if appropriate, race. This deviation from predicted or “observed minus predicted FEV,” is referred to hereafter as the “adjusted FEV,.” Change in FEV, over time was assessed for each subject by calculating a regression of his FEV,s on calendar time, producing an FEV, “slope” for each participant. The slope was analyzed as a single observation describing the individual’s change in pulmonary function over time. Since slope estimates calculated over only a single year of follow-up are very unstable and exhibit a within-group standard deviation that is at least twice as great as that of slopes measured over longer time periods (14, 15), analyses were restricted to subjects who had 2 or more years of follow-up during the latter half of the trial. Because only two to five FEV,s were used to determine a slope, an FEV, that was far removed from a subject’s remaining FEV,s would exert a marked effect on the calculated slope. Therefore, using the approach of Fletcher and colleagues (3), each subject’s values were checked for outliers before his slope was calculated. Subjects with outliers were excluded from the analyses.
FORUM:
CIGARETTE
SMOKING
IN
THE
MRFIT
625
Mean FEV, slopes were compared, first, for quitters and continuing smokers and, subsequently, for all smoking groups. Slopes were stratified by level of FEV, to examine the effect of this variable on the quitter-continuing smoker difference in FEV, decline. To permit comparison with the Fletcher study (3), the level of FEV, was defined for these analyses by FEV,/height3 and was dichotomized into 100 ~molfliter in the absence of cigarette smoking, the proportion was somewhat higher among self-reported quitters, possibly due to false reports of smoking cessation.
FORUM:
CIGARETTE
SMOKING
627
IN THE MRFIT
TABLE 1 LONGITUDINAL SMOKING STATUS OF MRFIT PARTICIPANTS Participants not using cigars, cigarillos, pipes, or beta-blockers
All participants
Cigarette smoking status
Percentage
Number
Percentage
1,817 2,545
14 20
990 1,434
16 23
1,332 2,156 4,557 226 233
10 17 35 2 2
594 831 2,197 74 165
9 13 35 1 3
Number
Never smoker Former smoker Sustained quitter by first 12 months Intermittent smoker Continuing smoker Othef No follow-up Total
12,866
6,285
a Started to smoke during the trial.
Characteristics
of Excluded
Subjects
The analysis group was compared with each of the excluded groups to determine whether they differed for selected baseline and follow-up characteristics. The groups were compared for age and height at baseline, adjusted FEV, at the midpoint of the trial, and proportion of each group smoking at either screen 1 or 3 and, among the smokers, the number of cigarettes per day, duration, and packyears of smoking prior to the MRFIT (Table 2). The proportion of cigarette smokers was lower in the analysis group than in the groups excluded because of smoking cigars, cigarillos, and pipes or having too few annual visits or too few pulmonary function tests (P < 0.001). However, smokers in the analysis group differed from smokers in the excluded beta-blocker group only in their smoking exposure
(76%)’ ‘“0””
No CCP CCP NWW Smokers
NoCCP CCP Former Smokers (Pre-MRFIT) Longitudinal
NOCCP CCP Quitters by First Annual Smoking
(94%)’
(94%)’
m
El
B
1 NOCCP CCP Intermittent Smokers
I
1 I NOCCP CCP Continuing Smokers
Status
FIG. 2. Serum thiocyanate at 12 months by smoking status (nonusers of beta-blockers only) for MRFIT participants included in the analysis group. *Percentage of group greater than 100 pmol/liter.
4,926 58 46.7 90 -287 27.8 35.8 50.4
Total
12,866
65 46.4 90 -323
28.0 35.3 49.9
28.4* 33.1* 47.6*
60 46.7 91 - 360*
2.737
Beta-blocker users Mb
28.0 35.6 50.3
-341*
76* 45.8s 91
3,844
Cigar, cigarillo, pipe smokers W
28.1 36.6 51.5
- 498d
77* 46.3 89
746
Too few visits (C)
27.7 37.8 52.3
-651d
80* 46.0 90
126
Too few PFTS 03
Excluded participants
TABLE 2 PARTICIPANTS WITH ANALYSISGROUPFORSELECTEDCHARACTERISTKY
0 At baseline, except for FEV,, which was measured at midpoint of trial. b Letters correspond to groups presented in Fig. 1. ’ Mean values. d FEV,s available for ~25% of group. * Different from analysis group, P < 0.05.
n % smoking at screen 1 or 3 Age (years) % white FEV, obs - pred (ml) Screen 1 or screen 3 smokers Years smokedc Cigarettes/day’ Pack-yearsc
Analysis group
COMPARISONSOF EXCLUDED MRFIT
27.6 36.5 51.2
- 298
59 46.7 86
416
Too few acceptable PPTS 03
29.2 30.2 44.0
-415
61 47.9 19
71
FEV, outliers (F)
?
2
s % 3
El
FORUM:
CIGARETTE
SMOKING
IN
THE
MRFIT
629
prior to the MRFIT: analysis group members smoked more cigarettes per day for slightly fewer years, resulting in a larger pack-years of exposure than that for the smokers using beta-blockers (P < 0.01). Analysis group subjects were slightly older than the cigar, cigarillo, and pipe smokers (P < O.Ol), but did not differ significantly from those of any excluded group in terms of height or proportion of white participants. Analysis group subjects had a higher adjusted FEV, than beta-blocker users and cigar, cigarillo, and pipe smokers (P < 0.01). Although the difference in adjusted FEV, between the analysis group and subjects with too few annual visits or too few pulmonary function tests was also significant, too few of the excluded subjects in these groups had FEV,s available for this comparison to be meaningful: only 11% of those with too few visits and 25% of those with too few tests had FEV,s available for analysis. FEV,
by Smoking
Status at the Trial Midpoint
FEV, at the midpoint of the trial showed the expected associations with age (r = -0.293) and height (r = 0.384) for ah smoking groups combined. In Table 3, FEV, is presented by smoking status and age for the 4,412 white and the 349 black participants in the analysis group. Since members of other racial groups were few (n = 165) and heterogeneous, they are not presented here. FEV, decreased with increasing age within each smoking group and with increasing smoking exposure within each age group. These patterns were slightly more erratic among blacks than among whites, probably due to the smaller numbers in the cells. The enhanced age effect seen in smokers (r = - 0.382) compared with never smokers (r = -0.252) is probably due to the increase in duration of smoking that accompanies increasing age among smokers (r = 0.804). As expected, blacks had lower FEV,s than whites in all age-smoking status groups. Among never smokers, the average FEV, among blacks was 88% of that among whites. The relationship between the level of pulmonary function and the smoking exposure variables remained after adjustment for age and height, with the correlations generally remaining the same or increasing slightly (data not shown). However, among smokers the correlation between the adjusted FEV, and years smoked (r = -0.212) was lower than the correlation between the unadjusted FEV, and years smoked (r = -0.364). This decrease in association probably again reflects the collinearity between age and years smoked. Correlations between the adjusted FEV, and the smoking exposure variables were strongest when all smoking statuses were considered together. For example, the correlations between the adjusted FEV, and cigarettes per day, years smoked, and serum SCN were - 0.3 11, - 0.329, and - 0.275, respectively, for all smoking groups combined. For smokers alone, the correlations for the same pairs of variables dropped to -0.166, -0.212, and -0.105, respectively. The correlation between the adjusted FEV, and SCN was also lower among never smokers alone, with r = -0.063. Since the never smokers had high FEV,s and low values for the smoking variables, while the smokers had lower FEV,s and high values for the smoking variables, the increased correlation that was observed for all smoking groups com-
630
TOWNSEND TABLE
MEAN
FEV,(ml)
ET AL. 3
AT MIDPOINTOF TRIAL BY AGE,RACE, AND SMOKING PARTICIPANTS INCLUDED IN THE ANALYSIS
STATUS FORMRFIT
Age (years) at time of FEV, measurement 3w5 Smoking status Never smoker White Black Former smoker White Black Sustained quitter by 12 months White Black Intermittent smoker White Black Continuing smoker White Black Othef White Black Total White Black
46-55
56-63
Total
n
Mean FEV,
n
Mean FEV,
n
Mean FEV,
n
Mean FEV,
128 24
3,892 3,345
386 29
3,563 2,%3
222 8
3,375 3,034
736 61
3,563 3,123
151 10
3,818 3,512
606 33
3,442 3,269
369 14
3,284 2,605
1,126 57
3,441 3,149
109 11
3,735 3,384
235 27
3,279
75 6
WI4
2,904
2,685
419 44
3,330 2,994
191 23
3,607 3,389
303 31
3,182 2,887
109 12
2,740 2,473
603 66
3,237 2,987
438 39
3,498 3,064
815 64
3,136 2,864
219 14
2,848 2,556
1,472 117
3,201 2,894
16
3,700 -
29 3
3,399 2,755
11 1
3,358 2,773
56 4
3,477 2,760
1,033 107
3,642 3,272
2,374 187
3,307 2,959
1,005 55
3,123 2,638
4,412 349
3,343 3,004
-
0 Started to smoke or resumed smoking during the trial.
bined probably reflected the correlation between FEV, and the presence or absence of smoking more than it reflected a correlation with the degree of smoking. Figure 3 summarizes the cross-sectional relationship between the adjusted FEV, and past or present cigarette smoking. Means and 95% confidence intervals are presented for the observed minus the predicted, or adjusted, FEV,. Percentages of subjects with low FEV,, i.e., an FEV, more than 1.645 SD below the predicted value, are also given. The adjusted FEV,s for all smoking groups except the intermittent and the continuing smokers were significantly different from each other based on multiple comparisons tests (P < 0.05). A low FEV, was four times more prevalent among continuing smokers than among never smokers, with 24% vs 6%, respectively. Figure 4 presents adjusted FEV, values for quitters by 12 months (Q), intermittent smokers (I), and continuing smokers (C), who all reported smoking at either screen 1 or 3, by pack-years of smoking prior to the MRFIT. As seen in Fig. 4, pack-years was associated with both FEV, and smoking group. Thirty-two percent of quitters smoked 30 or fewer pack-years prior to the MRFIT compared
FORUM: +lOO = E. 5
0
iif 0 ti e at? ii g
-100
I
CIGARETTE
SMOKING
E 5 %
THE
631
MRFIT
El
(6%)'
El
(9%)’
-200
-300
3 -E z
IN
. II (15%)'
-400 El (22%)'
-500 -600
I
Never Smokers
I
1
Former Smokers
Longltudlnal
I
Ouitlers by First Annual
Smoking
Intermittent Smokers
. 0 (24%)' I
Continuing Smokers
Status
3. Adjusted FEV, by smoking status for MRFIT participants in the analysis group: Mean, 95% confidence interval. FEV, taken at trial midpoint. Adjustment is the observed FEV, minus the FEV, predicted by age and height based on asymptomatic MRFIT participants who did not smoke or use beta-blockers. *Percentage of group more than 1.65 standard deviations (797 ml) below predicted FEV,. FIG.
with 15% of continuing smokers. Conversely, 37% of quitters smoked 50 or more pack-years prior to the MRFIT compared with 51% of continuing smokers. Smoking group retained its association with the adjusted FEV, after adjusting for pack-years of smoking in an analysis of covariance. The pack-years adjusted mean values for the adjusted FEV, were -304.6, -417.3, and -434.9 ml for the quitters, intermittent smokers, and continuing smokers, respectively (P < 0.0001). Pack-years was also significantly associated with deviation from the predicted FEV, after adjusting for smoking group (Z’ < 0.0001). Similar results were obtained when years smoked and the initial number of cigarettes per day were used as covariates in place of pack-years. Rate of FEV, Decline by Smoking
Status
When examined for all smoking groups combined, the FEV, slope showed a very low but significant correlation with age, height, and the smoking exposure indices, with r = -0.02 to - 0.05 (P < 0.05). Correlations of less or similar magnitude were seen when the smoking groups were examined separately, but, in general, these correlations were not statistically significant. To assess the effects of smoking cessation on the rate of FEV, decline over the latter half of the trial, FEV, slopes were compared for quitters and continuing smokers, controlling for the level of FEV,. Table 4 shows that quitting smoking has a modest effect on the FEV, slope. An analysis of covariance confirmed this
632 0 -100 E z
-200
2 P
-300
g 0.z
-400
t k 3
-500
: c 3
-600
P g
-700
p” 0
a .’c n B
(n-154)
(n-194,
F1 (n-246)
TOWNSEND
ET AL.
at c goII. u ’ c a
l-l
(n-222)
(“=+X3,
(“-146,
(“.W,
0 l
l
II
(“-473,
,“.162,
-600 -900
30.49
0.10). Finally, FEV, slopes were examined across all smoking groups (Table 5). The smoking groups had different mean FEV, slopes (P < 0.00001) and continuing smokers differed from former and intermittent smokers (P < 0.05) based on a multiple-comparisons test. However, the other smoking groups were not signiticantly different from each other.
FEV, S~o~~(ml/year)~~
LEVELOF
TABLE 4 FEV, FOR MRFIT QUITTERSAND
CONTINUING
SMOKERS
IN
THEANALYSISGROUP
FEV,/height3” Smoking status Quitters by 12 months Continuing smokers Total
2500 ml/m3
