| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
Guest Access | Sign In via User Name/Password
|
|
|
|||||||
Guest Access | Sign In via User Name/Password |
|||||||||
* From the Center for Health Research (Dr. Vollmer and Ms. Pedula), Portland, OR; Respiratory Sciences Center (Dr. Enright), University of Arizona, Tucson, AZ; Harvard Medical School (Dr. Speizer), Boston, MA; University of Pittsburgh (Dr. Kuller), Pittsburgh, PA; and Division of Lung Diseases (Drs. Kiley and Weinmann), National Heart, Lung, and Blood Institute, National Institutes of Health, Bethesda, MD.
Correspondence to: William M. Vollmer, PhD, Center for Health Research, 3800 N. Kaiser Center Dr, Portland, OR 97227-1098; email: william.vollmer{at}kp.org
 |
Abstract |
|---|
 TOP Abstract IntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
Design: A random-effects metaregression analysis to synthesize results from common cross-sectional regression models fit to participants in each of 10 gender-race strata in each of eight large population-based observational studies or clinical trials.
Setting: Source data collected as part of the most recently completed examination cycle for each of the participating studies.
Participants: Participants ranged in age from 30 to 85 years, although the age, race, gender, and general health characteristics of each of the populations varied greatly.
Interventions: Most of the studies were observational in nature, although some did involve lifestyle interventions. All treatment assignments were ignored in the analysis.
Measurements and results: All studies measured lung function using standardized methods with centrally trained and certified technicians. Study findings confirm statistically significant, dose-related smoking effects in all race-gender groups studied. Significant gender differences in the effects of cigarette smoking were seen only for blacks; black men who smoked had greater smoking-related declines in FEV1 than did black women. This effect was present in only one of two smoking models, however. Significant racial differences in the effects of smoking were seen only for men, with Asian/Pacific Islanders having smaller smoking-related declines than white men in both models.
Conclusions: In summary, this analysis generally failed to support the hypothesis of widespread differences in the effects of cigarette smoking on lung function between gender or racial subgroups.
Key Words: gender • lung function • race • smoking
|
Introduction |
|---|
|
TOPAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
During the last 45 years, the National Heart, Lung, and Blood Institute (NHLBI) has supported the use of spirometry in several large, population-based observational studies and clinical trials. During their most recent examination cycles, the spirometric methods and respiratory questionnaires used by these studies have been standardized, facilitating between-study comparisons of lung function. In 1996, the NHLBI sponsored a workshop designed to encourage the investigators who had collected these data to collaborate, by suggesting hypotheses that could only be answered by pooling and then jointly analyzing the lung function results from these studies.14 This report is based on the analysis used to answer one of the hypotheses examined for the workshop: that the relationship between smoking and lung function differs by gender and race/ethnicity.
|
Materials and Methods |
|---|
|
TOPAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
|
30 observations were dropped from
the analysis and are not shown in Table 1
. The Multiple Risk Factor
Intervention Trial (MRFIT), which was restricted to men at high risk
for coronary heart disease, had two gender-race strata with between 50
to 100 participants; all remaining gender-race strata included in this
analysis contained at least 239 participants. Data on Native Americans
derive solely from the Strong Heart Study (SHS), while data on Asians
and Pacific Islanders derive overwhelmingly from the Honolulu Heart
Program (HHP). As with the MRFIT, the HHP included only men. Technically, Table 1 presents the number of individuals included in model 1 (see below). For the Coronary Artery Risk Development in Young Adults Study (CARDIA), dose data were not available for ex-smokers. The CARDIA coordinating center fit model 1 ([1] excluding ex-smokers, and [2] including them in the low-dose category). Since the results were very similar, we used the results of the latter models in this analysis. We did not use CARDIA data in the pack-years analyses (model 2), however. We also discovered that for MRFIT, the number of individuals included in model 2 was substantially less than the number used in model 1 for some of the racial groups. Still, all of the gender–race-specific models for each study were based on at least 35 individuals, and all but two were based on at least 141 individuals.
Table 2 shows the age ranges and smoking status of study participants with acceptable spirometry results. Participants in the Cardiovascular Health Study (CHS) and the HHP clearly are much older than those in the remaining studies, while the CARDIA reflects a very young study cohort. In general, a greater proportion of men than women were ever smokers. The one exception, perhaps reflecting a cohort effect, was the CARDIA, which showed no difference in ever smoking between men and women. The distributions of age and, among smokers, the number of years smoked were generally very similar for men and women within each study. Only for the SHS, which was limited to Native Americans, was there a notable difference in number of years smoked between men and women. Although not shown in Table 2 , the amount smoked per day was also similar for male and female smokers within each study, although male subjects tended to smoke slightly more, on average, than did female subjects.
|
Age, gender, and race were also self-reported.
Spirometry
Spirometry was performed following American Thoracic Society
recommendations25
by centrally trained and certified
technicians using nearly identical software and manuals of procedures
across the eight studies, which included daily leak and volume
calibration checks. A water-sealed volume spirometer connected to a
personal computer was used by six of the studies, while two other
studies used functionally equivalent dry rolling seal volume
spirometers. Standing height (in meters) and weight (in kilograms) were
measured in stocking feet.
For this analysis, we report only on regression equations for the FEV1.
Statistical Analysis
Each of the participating studies provided us with the
results from a common set of cross-sectional regression models fit to
each race and gender stratum. Although cross-sectional regression
models cannot provide direct (ie, person-specific) estimates
of lung function decline with age, they can provide indirect estimates
of annual lung function decline.26
These estimates may not
agree with direct, longitudinally derived estimates,
however.26
27
28
Two cross-sectional regression models were used to address the relationship of lung function with cigarette smoking while adjusting for age and height. Both models account for the effects of amount and duration of cigarette smoking on lung function, but they do so in different ways. The two models are listed below:
Model 1
 |
ß1 = the estimated annual
change in FEV1 (milliliters per year) in the
absence of current cigarette smoking, (ie, among
nonsmokers); ß4 = the estimated
irreversible excess annual change in FEV1
(milliliters per year) in light smokers ( 10 cigarettes/d) when
compared to nonsmokers; ß5 = the
estimated irreversible excess annual change in
FEV1 (milliliters per year) in heavy smokers
(> 10 cigarettes/d) when compared to nonsmokers;
and ß6 = reversible decrement in
FEV1 among current smokers.
In addition, ß2 and
ß3 represent the linear and quadratic influence
of height on lung function, and 
is a normally distributed error
term having mean zero.
Model 2
 |
ß1 = estimated annual change in FEV1 (milliliters per year) in nonsmokers; ß4 = estimated irreversible excess change in FEV1 associated with smoking (milliliters per pack-year); ß5 = reversible decrement in FEV1 among current smokers.
Once again, ß2 and
ß3 represent the linear and quadratic influence
of height on lung function, and is a normally distributed error
term having mean zero.
The pack-years model (model 2) is potentially more powerful, in that it more fully utilizes the dose information. It does, however, make stronger assumptions about the effects of both dose and duration of smoking on lung function. By contrast, the dose-duration model (model 1) allows for more general relative dose effects, but is limited in that it only uses two dose groups. Consistent with the literature,2 5 both models tacitly assume that rates of decline among ex-smokers revert to those of never smokers on cessation. Both models also allow for the irreversible decline in lung function that is associated with smoking as well as the reversible decline in lung function that may be due in part to a bronchoconstrictive effect of cigarette smoke or to the increased mucous secretion that characterizes current smokers.2 5 6 This study focuses only on the estimated irreversible effects of smoking.
Implicit in both models 1 and 2 is the assumption that lung function decline, both among smokers and nonsmokers, is independent of lung volume. While some investigators assume that decline should be proportional to volume and hence divide their lung function measurements by some power of height (eg, FEV1/height2), this does not necessarily provide any improvement in fit, at least for cross-sectional data, relative to models that assume decline is not proportional to volume.29
Although age tends to be highly correlated with both number of years smoked and pack-years among smokers, the inclusion of never smokers in these models permits valid and reliable estimation of the age and smoking coefficients in these models. Essentially, the analysis uses the never smokers to estimate the normative aging effect (ß1) and then subtracts out this effect from the smokers’ data to estimate the incremental effect of smoking. This can be readily shown by limiting the analysis to never smokers and rerunning the models without the smoking terms.
We used a random effects metaregression model30 to synthesize the results across the studies. Specifically, we used the parameter estimates (the ßs) from the preceding models as dependent variables and regressed these estimates on gender, race, and study. This allowed us to obtain pooled estimates of the influence of gender and race on smoking-related decline in lung function. Conceptually, the model is of the following form:
Model 3
 |
Because standard regression techniques do not account for the fact that
the ßijks are themselves estimated with varying degrees
of precision, we used a two-stage approach to fit this model.
Initially, we used an unweighted (ordinary) least squares analysis of
model 3 to provide estimates of both the variation in estimating the
ßs in the original models (models 1 and 2), as well as the variation
in these estimated ßs from study to study and, within a given study,
from stratum to stratum. We then used these two variance components to
calculate the weights for a weighted least squares regression analysis,
which provided the final (correct) estimates of the regression
coefficients (
s) in model 3.
We actually used three different analytic models for the random-effects analysis. Model 3 above is what we refer to as the overall model. It assumes constant, additive gender and race effects for each study and uses the entire data set to estimate these effects. The results of this model are summarized in Table 3 (for gender) and in Table 4 (for race). We also conducted analyses of gender effects stratified by race (summarized in Table 3 ), and of racial effects stratified by gender (summarized only in the text). Because of the way the estimates are constructed, these latter analyses result in slightly different estimates (and standard deviations) of smoking effects in the gender-race subgroups.
|
|
0.05. |
Results |
|---|
|
TOPAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
|
Gender differences in these smoking effects tended to be small and inconsistent. Overall, the effects of smoking appeared to be slightly greater in male subjects than in female subjects based on the dose-duration model (about 2 mL/yr of smoking). However, these differences were statistically significant only for the low-dose group, and no evidence of gender differences was seen in the pack-years model. To the extent that they exist, gender differences appeared to be limited to blacks. For both the low- and high-dose groups in the dose-duration model, the estimated excess declines in FEV1 among black smokers were 4 to 5 mL/yr greater in male subjects than in female subjects, and these differences were statistically significant in both dose categories. As with the overall analysis, however, we found no evidence of differential smoking effects between black men and women in the pack-years model.
The widths of the confidence intervals for the female-male differences in smoking effects shown in Table 3 reflect both the sample sizes on which the estimates were based and the consistency of these differences from study to study. This latter factor is why, for example, the confidence intervals for blacks are tighter than those for whites, despite the smaller number of black than white participants in these studies. The gender differences for blacks were much more consistent for blacks than for whites (data not shown).
Table 4 presents data on smoking effects separately for each racial group after adjusting for gender; all but one of the estimated excess smoking effects are significantly less than zero. As with Table 3 , the estimated smoking effects in the dose-duration models are greatest among the heavier smokers, and the pack-years coefficients are typically intermediate to those seen for light and heavy smokers. The excess declines associated with smoking were greatest among whites, with the other racial groups being fairly homogeneous. Only in one racial group, Asians/Pacific Islanders, did the estimated smoking effects differ from those for whites, however, and then only among heavier smokers in the dose-duration model. When stratified by gender, these latter differences were significant for heavy smokers in the dose-response model and also based on the pack-years model. No other consistently significant racial differences were seen among men, and no significant differences of any kind were seen among women.
When the estimated declines in Tables 3 , 4 were expressed as a percentage of the nonsmoker declines shown in Table 5 , the same patterns persisted, although we did not test for the statistical significance of these latter differences.
|
Discussion |
|---|
|
TOPAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
The cross-sectional nature of this analysis represents an important potential limitation, since numerous factors, such as cohort and survivor effects, can bias cross-sectionally estimated rates of decline.26 27 28 Less clear is the extent to which subgroup comparisons should differ between cross-sectional and longitudinal analyses. Another limitation of cross-sectional analyses, relative to longitudinal analyses, is the reduced power of cross-sectional analyses due to the reliance on between-subject sources of variability.31 Nonetheless, the studies included in this analysis are all large, population-based studies conducted under rigorous standards of quality control, which both enhance the credibility of these results.
Also, as our own results clearly indicate, the estimation of the effects of smoking on lung function can be very dependent on the nature of the model used to describe these effects, not to mention the quality of smoking history information that is available. Most reasonable analyses should be able to demonstrate adverse main effects for smoking, if they exist. In our case, both of our smoking models clearly showed statistically significant adverse effects of smoking in all race-gender groups, which increased with increasing dose. However, these models often provided very different estimates of subgroup differences.
Gender Differences
The only significant gender difference in estimated smoking
effects occurred among blacks, for whom both light and heavy male
smokers appeared to have excess, smoking-related declines in
FEV1 of 4 to 5 mL/yr of smoking when compared to
black women. However, we observed no gender differences in estimated
smoking effects among blacks when these effects were expressed as
milliliter of lung function per pack-year smoked. This pattern was
observed very consistently across the different studies.
The reason for this lack of consistency is not certain. The dose-duration model is more nonparametric in nature than the pack-years model, but has the disadvantage that it offers only a partial adjustment for dose. If men tended to be heavier smokers than women, it could help to explain the fact that the majority of the dose-duration models resulted in greater estimates of smoking-related decline in men than in women, even though no such pattern was observed for the pack-years models. The available data for these studies, however, suggests that in general the amount smoked was similar for male and female smokers, so that differential smoking does not explain the difference. Black men smoked, on average, zero to three more cigarettes per day than did black women, for instance. By contrast, the pack-years model may give somewhat misleading results due to the nature of its assumptions. For instance, this model assumes that smoking four packs per day for 5 years has the same effect as smoking one pack per day for 20 years. The dose-duration model is much less rigid, and only assumes different rates of decline for heavy and light smokers.
If the differences in the effects of smoking for black men and women are real, they may be explained by several factors, most of which would apply to other racial groups as well. First, black women may have reduced susceptibility to the deleterious effects of smoking. This reduced susceptibility could be due, for instance, to genetic polymorphisms or to other factors (such as differences in dietary intake or the use of vitamin supplements) that we did not ascertain. Second, black women may tend to smoke cigarette brands that produce less airway irritation (eg, due to differences in filters or tar content) than those smoked by black men, or they may tend to smoke cigarettes in a way (eg, depth of inhalation) that reduces airway exposure to the smoke derived from each cigarette.32 Third, black women may have less occupational exposure than do black men to factors that increase the risk of developing COPD or asthma. Finally, estrogen levels in black women may also play a role by influencing airway epithelial cell function33 or airway responsiveness.34
Xu and coworkers10 35 suggest that differences in the never-smoking reference groups used for men and women can bias estimates of the effects of smoking on lung function, and may help explain some of the differences in gender effects seen in various studies. They postulate that the higher the prevalence of smoking, the greater will be the prevalence of cardiovascular and respiratory comorbidities in the remaining never-smoking population. This increased comorbidity, in turn, leads to overestimates of "normal" rates of decline and, hence, to underestimates of the effect of smoking. Following this logic, studies in which the prevalence gap for smoking among male and female subjects is large will tend to underestimate male-female differences in smoking effects relative to studies in which the prevalence of smoking for men and women is more comparable. In these studies, the prevalence of ever smoking was, on average, 19 percentage points higher in male than in female participants.
Racial Differences
Among both men and women, racial differences in smoking-related
declines in lung function tended to occur only among heavier smokers,
with the effects most pronounced among whites. However, the only
significant racial differences that were observed occurred among male
subjects, for whom both heavy smoking Asian/Pacific Islanders and
Hispanics had significantly smaller smoking-related declines than did
heavy smoking whites, and only the Asian/Pacific Islander-white
difference was also significant in the pack-years model.
Those racial differences we did observe, if real, could reflect real genetic differences in susceptibility to cigarette smoke or they could reflect the confounding effects of differences between the various racial groups in environmental exposures (eg, air pollution and passive smoking), occupational exposures, the method of smoking (eg, depth of inhalation), amount smoked, and age of onset of smoking.
Other potential confounders include between-study differences in equipment, testing methods, environment (eg, ambient temperature and humidity), or participant makeup or recruitment, as well as differences in the types of cigarettes smoked and the effect of health on smoking habits and lung function. These issues may be especially relevant, since some of the racial comparisons were necessarily forced to rely exclusively, or almost so, on between-study comparisons. For example, the SHS was the only study to include Native Americans, and it did not include any other racial groups. Similarly, almost all of the Asian/Pacific Islanders included in the study were participants in the HHP, and no female Asian/Pacific Islanders were included in the analysis. The HHP cohort was also considerably older at the time the pulmonary function data were collected than all but the CHS cohort. For the HHP cohort (and hence for comparisons involving Asians/Pacific Islanders), not only do these differences force a (largely untestable) extrapolation of aging effects from the younger to the older cohorts during model fitting, but the smokers in the HHP cohort are by definition survivors and, therefore, are likely biased toward the inclusion of more "nonsusceptible" smokers. Such a bias is less likely to be the case for the younger cohorts. The data on Hispanics are more broadly based than those for Asian/Pacific Islanders and Native Americans, but still derive primarily from just two studies, one of which included only Hispanics.
The two racial groups that are best represented in the data are whites and blacks. Four of the eight studies used in the analysis included substantial number of black and white male and female subjects, and a fifth study, MRFIT, included large numbers of black and white men. Comparisons of smoking effects between blacks and whites are thus much more likely to be reliable and valid than are comparisons between other racial groups.
Finally, it is important to acknowledge that what we have been calling race is really a mixture of race/ethnicity, since "Hispanic" does not actually represent a racial grouping. Rather, it refers to a cultural identity or ethnic background. Further, both Hispanics and many of the true racial classifications encompass broad, heterogeneous populations with widely varying genetic profiles and ethnic heritages. This heterogeneity inevitably blurs what may be true differences in the effects of cigarette smoking, and also makes it difficult to generalize findings even within a single classification. For example, the Hispanic population in New Mexico may not be at all representative of Hispanics from Miami or New York in terms of their response to cigarette smoke.
Comparisons With Previous Studies
Two population-based cross-sectional studies of whites showed that
the prevalence of airways obstruction was higher in female smokers when
compared to male smokers.7
8
Similarly, Chen and
coworkers,9
in their cross-sectional study of 1,149 white
adults living in Humboldt, Saskatchewan, found that women smokers
declined in FEV1 at a rate 4.1 mL/pack-year
faster than did male smokers. In contrast, investigators from the Six
Cities study initially estimated that the loss of height-adjusted
FEV1 was 7.4 mL/pack-year for men vs 4.4
mL/pack-year for women.4
After 6 years of follow-up of
this cohort,5
the longitudinally estimated excess rate of
FEV1 decline associated with smoking was again
greater in men (12.6 mL/pack-year) than in women (7.2 mL/pack-year).
Consistent with our results, the Tucson Airways Study36 reported no significant cross-sectional gender differences in the rate of decline of lung function (FEV1 and forced expiratory flow after 75% of vital capacity has been expelled) with pack-years in white participants after controlling for respiratory trouble before age 16. Subsequent longitudinal analysis of this cohort, however, revealed greater excess declines in male smokers (13 mL/yr) than in female smokers (7 mL/yr). Interestingly, the authors noted that rate of decline was significantly associated with amount smoked but not with pack-years in this latter analysis.2
Longitudinal analysis of spirometry data from two population samples in the Netherlands showed no gender difference in the effects of continuing smoking on FEV1 or FVC, while a cross-sectional analysis showed a smaller effect of current smoking in women when compared to men.37 The authors observed that participants with better-than-average lung function were more likely to remain in a longitudinal study (a healthy survivor effect) and that lung function seemed to improve in successive birth cohorts. These and other factors make it difficult to directly compare longitudinal and cross-sectional estimates of lung function decline.26 27 28 31
In summary, this analysis of large, NHLBI-funded studies generally failed to support the hypothesis of widespread differences in the effects of cigarette smoking on lung function between gender or racial subgroups. This finding is generally consistent with the existing literature on this topic, which show very conflicting results. This analysis also served to highlight the fact that, apart from blacks, the data to address the question of racial differences in the effects of cigarette smoking on lung function are fairly limited. Only one of the eight studies in this analysis included Native Americans, and only one included Asians. Further, both of these studies were limited exclusively to those minority groups, thus precluding the possibility of direct comparisons based on populations selected and tested using the same methodology and personnel. While Hispanics were represented in three studies, > 90% of them came from just two studies, and again only one of these included other racial groups. By contrast, the vast majority of the studies used in this analysis included both men and women from each of the racial groups they sampled, thus permitting very reliable comparisons of gender differences in susceptibility to cigarette smoke.
|
Acknowledgements |
|---|
|
Footnotes |
|---|
Received for publication May 18, 1999. Accepted for publication September 22, 1999.
|
References |
|---|
|
TOPAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
This article has been cited by other articles:
|
|
 |
|
  R. McGhan, T. Radcliff, R. Fish, E. R. Sutherland, C. Welsh, and B. Make Predictors of Rehospitalization and Death After a Severe Exacerbation of COPD Chest, December 1, 2007; 132(6): 1748 - 1755. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
M. T. Dransfield, G. R. Washko, M. G. Foreman, R. S. J. Estepar, J. Reilly, and W. C. Bailey Gender Differences in the Severity of CT Emphysema in COPD Chest, August 1, 2007; 132(2): 464 - 470. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
W. M. Chatila, E. A. Hoffman, J. Gaughan, G. B. Robinswood, G. J. Criner, and for the National Emphysema Treatment Trial Researc Advanced emphysema in african-american and white patients: do differences exist? Chest, July 1, 2006; 130(1): 108 - 118. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 R. G. Barr, C. C. Wentowski, F. Grodstein, S. C. Somers, M. J. Stampfer, J. Schwartz, F. E. Speizer, and C. A. Camargo Jr Prospective Study of Postmenopausal Hormone Use and Newly Diagnosed Asthma and Chronic Obstructive Pulmonary Disease Arch Intern Med, February 23, 2004; 164(4): 379 - 386. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
W. M. Chatila, W. A. Wynkoop, G. Vance, and G. J. Criner Smoking Patterns in African Americans and Whites With Advanced COPD Chest, January 1, 2004; 125(1): 15 - 21. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 A. Langhammer, R. Johnsen, A. Gulsvik, T.L. Holmen, and L. Bjermer Sex differences in lung vulnerability to tobacco smoking Eur. Respir. J., June 1, 2003; 21(6): 1017 - 1023. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
I. Cerveri, S. Accordini, G. Verlato, A. Corsico, M.C. Zoia, L. Casali, P. Burney, and R. de Marco Variations in the prevalence across countries of chronic bronchitis and smoking habits in young adults Eur. Respir. J., July 1, 2001; 18(1): 85 - 92. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
