INTRODUCTION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Keywords: smoking; respiratory function tests; genetics

The rate of decline of forced expiratory volume in one second (FEV₁) is a measure of changes in the peripheral and large airways (1) and is considered a sensitive indicator of chronic obstructive pulmonary disease (COPD) (2, 3). Although COPD is primarily a disease of cigarette smokers, the accelerated rate of decline which characterizes COPD is present in only 15 to 20% of smokers, indicating that other genetic and environmental factors are involved (4, 5).

The rate of decline or slope of FEV₁ is a function of age and sex (6), smoking habit (5), body size (7, 10), a history of respiratory infections, airway hyperresponsiveness, atopy, and environmental and occupational exposures (14). Excessive rates of decline in smokers have been observed in numerous studies (2, 14, 15). Ryan and colleagues reported a mean rate of decline approximately 15 ml/yr greater for moderate-heavy smokers than for never-smokers (14).

Widely divergent rates of decline in FEV₁ have been reported in the literature (7). Cross-sectional analyses have shown greater apparent age effects and earlier age of onset of decline than longitudinal analyses from the same populations. These discrepancies may be due in part to overly simplistic cross-sectional models and to a survivor effect, i.e., earlier death in elderly individuals with lower lung function leading to "under-representation" in longitudinal analyses (7, 16). A number of studies have reported a nonlinear relationship between FEV₁ and age, in which the slope of FEV₁ is less steep in early adulthood than at age 35 and older (7, 17). In general, researchers have used quadratic and cubic terms for age and height in regression analysis to correct for this nonlinearity (7).

Several studies in diverse populations have found significant familial correlation of FEV₁, ranging from 0.11 to 0.66 in nuclear families (4, 18). In 1984, however, Lebowitz and colleagues (24) reported no familial aggregation of initial FEV₁ in nuclear families enrolled in the Tucson Epidemiological Study of Airway Obstructive Diseases (TESAOD) after controlling for body habitus (height/ ³ weight). The results of more recent analysis of familial correlation and inheritance of FEV₁ in the TESAOD population demonstrated significant parent-offspring (r = 0.142; p < 0.001) and sibling (r = 0.228; p < 0.001) correlations in FEV₁, after fitting continuous equations (breakpoint regression model) throughout the observed age span and adjusting for height, height-squared, and smoking variables (25).

To our knowledge, familial correlation in the slope of FEV₁ has not been reported. This report examines correlation in the rate of decline of FEV₁ among family members in the Tucson Epidemiological Study population, and attempts to identify increased susceptibility to an accelerated rate of decline associated with cigarette smoking in subgroups of the population.

	METHODS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Study Population

The original study population was comprised of a random, stratified cluster sample of white, non-Mexican-American households in the Tucson area who were enrolled in the TESAOD beginning in 1972 and followed for up to 25 yr. Those subjects who had family members also enrolled in the study (n = 4,399) comprised the total study population. Every 1.5-2 yr, subjects received pulmonary function tests (PFTs) and completed questionnaires that included respiratory health and smoking histories. PFTs were performed during 11 surveys, using a pneumotachograph according to the criteria of the American Thoracic Society (26). Because the measurement of change in FEV₁ over time is potentially subject to small survey biases (7), the calculation of the slope of FEV₁, for the purposes of our study, was based on a minimum of three FEV₁ assessments over a period of five or more years in subjects over 18 yr of age at their first included PFT. A total of 1,279 individuals in 392 families met these criteria for inclusion.

Smoking status was self-reported in each survey questionnaire. Subjects who reported current or past smoking of an average of five or more cigarettes per day were classified as ever-smokers or smokers. Subjects who consistently reported never having smoked cigarettes were classified as never-smokers or nonsmokers.

Statistical Methods

Differences in mean characteristics and prevalence of smoking and disease between included and excluded subjects and between included parents and offspring were analyzed using analysis of variance (ANOVA) for continuous variables and chi-square distribution (²) for discrete variables.

The slope of FEV₁ was calculated by fitting a simple linear regression line to each subject's FEV₁ data with age. The individual "raw" slope values were then adjusted by sex for "mean" age (calculated as the age midway between the initial and final PFT), age², age³, adult height and height², mean body mass index (BMI), change in BMI, initial FEV₁, mean number of cigarettes smoked per day (over the duration of surveys with PFTs), and number of data points. First-order multiplicative interaction terms were also included, using the forward selection method in multiple linear regression analysis. To correct for the difference in number of data points and the error inherent in the slope estimate, we included a weighting factor in the regression that was equal to the reciprocal of the square root of the standard error of the estimate for each individual's slope value (27).

We used the FCOR program in the Statistical Analysis for Genetic Epidemiology software package (28) to calculate Pearson's correlation coefficients for standardized residual FEV₁ slope values between spousal pairs, all parent-offspring pairs, and sibling pairs using the equal weight to all pairs option. These were calculated for all familial pairs in the total population, regardless of smoking status, and in smoking-concordant and nonsmoking-concordant familial pairs. A one-sample t test was used to determine whether the correlation coefficients were significantly different from zero, and Fisher's z test was used to compare the coefficients to each other (29).

	RESULTS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Table 1 compares smoking, lung function, and respiratory disease characteristics of included and excluded Tucson Epidemiological Study participants who were at least 18 yr old at their initial survey and for whom data on family members were available. Those included in our study had a minimum of three PFTs over a period of five or more years (n = 1,279); those who were excluded (n = 876) did not meet these testing criteria. Burrows and colleagues (7) examined the variability and validity of the TESAOD spirometry data and found a high degree of consistency in the data. In addition, they found that the variability in longitudinal changes in FEV₁ resulting from random variability in FEV₁ decreased as an inverse function of time. There were significant differences in the proportion of males and females, age at initial survey, proportion of ever-smokers and never-smokers, and physician-diagnosed respiratory diseases in those included and excluded from our analyses. The mean age of subjects included in our longitudinal analyses was slightly younger than that of the excluded subjects (31.5 versus 33.4 yr, respectively, p < 0.008), but the mean initial FEV₁ values for included and excluded subjects were not significantly different (p < 0.188). Included individuals were less likely to have ever smoked (56.1 versus 60.3%, p < 0.05), and smoked fewer cigarettes daily (p < 0.001) than those excluded. They were also significantly more likely to have reported ever having been diagnosed by a medical doctor (either in the initial survey or in later surveys) with asthma (18.6 versus 13.3%, p < 0.002), chronic bronchitis (27.0 versus 16.6%, p < 0.0001), and emphysema (4.4 versus 2.5%, p < 0.03).

Because the measurement of change in FEV₁ over time is potentially subject to small survey biases (7), the calculation of the slope of FEV₁, for the purposes of our study, was based on a minimum of three FEV₁ assessments over a period of five or more years. Those who met the criteria for inclusion in the longitudinal analysis participated in an average of seven PFTs and survey questionnaires over almost 14 yr, and thus had many more opportunities to report new diagnoses of conditions such as asthma and chronic bronchitis than those who were excluded from the analysis owing to insufficient data points and duration. It is also possible that those subjects who remained in the study over an extended period of time were motivated to participate because of underlying pulmonary conditions and thus may have been at greater risk of increased decline in lung function over time. The raw FEV₁ slope of asthmatics in the study was not significantly different from that of nonasthmatics (p < 0.605), though the adjusted slope of the asthmatics was significantly lower (p < 0.001). For those who reported ever having been diagnosed with chronic bronchitis, both unadjusted and adjusted slopes were significantly lower than that of subjects with no reported chronic bronchitis (p < 0.016 and 0.001, respectively).

We also compared demographic characteristics, respiratory disease prevalence, smoking, and lung function for parents and offspring who met the criteria for the slope analysis and had genetically related family members who met the criteria for inclusion (Table 2). As expected, the mean (midpoint) age, initial adult FEV₁ measures, raw slope values, number of (pulmonary function) data points, and years in the study were significantly different for parents and offspring (all p < 0.001). Parents were more likely to have ever smoked cigarettes (66.9 versus 54.5%, p < 0.010) and smoked more cigarettes per day on average than their adult offspring (p < 0.001). The prevalence rate of physician-diagnosed asthma was higher in offspring than parents, but not significantly higher (18.3% of offspring and 14% of parents, p < 0.251). The prevalence of chronic bronchitis was similar among parents and offspring (p < 0.687), but the emphysema rate was significantly higher in the parents than in their offspring (p < 0.001)a prevalence of almost 10% in parents, and less than 0.5% in offspring.

Partial regression coefficients, standard errors, and significance values for all of the variables used to compute FEV₁ slope residuals are presented in Table 3. The independent variables were entered one at a time by forward selection, with selection criteria set to maximize inclusion of variables. The same variables were included in the regression equations for both males and females.

More than 20% of the total variance in slope values for males and almost 25% for females could be explained by the regression models. The interaction between initial FEV₁ and mean age accounted for most of the variance in both sexes (14.0% in males and 17.9% in females). Age-squared, change in BMI, and usual number of cigarettes smoked per day each contributed approximately 1% to the total variance in males. In females, the interaction of BMI and age, usual number of cigarettes smoked, initial FEV₁, and mean age each accounted for approximately 1% of the total variance.

Familial correlation coefficients for residual FEV₁ slope values were calculated for all of the families whose members had FEV₁ measures that met the inclusion criteria (Table 4). In these families, standardized residual FEV₁ slope values were significantly correlated only among siblings (r = 0.256, p < 0.001, n = 166) (see also Figure 1A). Although overall parent-offspring correlation was very low (r = 0.041, p < 0.40, n = 522), father-son correlation approached significance (r = 0.161, p < 0.10, n = 111).

View larger version (34K): [in this window] [in a new window]   Figure 1.   Sibling correlation of standardized residual FEV1 slope values.

We also calculated the heritability of the adjusted slope values using the methods described by Falconer and MacKay (30), to assess the genetic versus environmental components of variance in the phenotype defined. Simple linear regression models of parental values (mother, father, and mid-parent) with the residual slope values of their offspring and of siblings with full siblings were run, and the regression coefficients () were used to calculate heritability. In this analysis, as well as in the familial correlation analysis, there was no evidence of mother- offspring inheritance of lung function decline (2 = h² = 0.006). Heritability was estimated at 0.46 for fathers and offspring (2 = h²), and 0.335 for mid-parent and offspring (where  = h²). The heritability estimate of full siblings ranged from 0.406 to 0.516 (sibling 1 against sibling 2 and sibling 1 against sibling 3), which should be considered the upper limit of heritability and the least reliable estimate of heritability (where 2  h²) due to the effects of a common environment.

To separate out the effects of smoking on familial correlation of FEV₁ decline, correlation coefficients were also calculated for pairs of family members who were concordant for smoking status (either smoking-concordant or nonsmoking-concordant). Despite very highly significant correlations in the number of cigarettes smoked daily within families (p < 0.001, for spousal, parent-offspring, and sibling pairs), the standardized residual slope values were not significantly correlated among either smoking-concordant or nonsmoking-concordant spousal or parent-offspring pairs (Table 4). However, siblings, especially smoking-concordant pairs, had standardized residual slopes that were very highly correlated (r = 0.483, p < 0.01, n = 39) (Figure 1B). Residual slope values for nonsmoking sibling pairs were also significantly correlated (r = 0.280, p < 0.05, n = 63) (Figure 1C). However, no correlation was observed among smoking-discordant sibling pairs (r = 0.031, p < 0.77, n = 68) (Figure 1D).

An additional analysis of sibling correlation in residual slope values was conducted only on individuals who had never had a diagnosis of asthma. Sibling correlations were essentially unchanged when asthmatic individuals were removed from the analyses (r = 0.307, p < 0.001, n = 139).

To enable a comparison of the mean FEV₁ slope of smokers to that of nonsmokers, standardized residual slope values were recalculated using a multiple regression model that adjusted for all of the independent variables included in the original regression model, except for number of cigarettes smoked daily. These residuals, unadjusted for smoking, were then expressed as z-scores in ml/yr by standardizing the mean residuals for ever-smokers and never-smokers by the population mean and standard deviation. The mean raw slope of smokers in the total population was significantly lower than that of nonsmokers (20 and 10 ml/yr, respectively, p < 0.001). The difference between z-scored slope values for smokers and nonsmokers approached significance (15 and 11 ml/yr, respectively; p < 0.054). Within the parent population, the mean raw slope values of the smokers and never-smokers were significantly different (p < 0.001), but did not differ from the population means. However, the z-scored means for the parent population were lower than those for the total population (18 ml/yr for ever-smokers and -14 ml/yr for never-smokers, p < 0.055). In the offspring population, there were no significant differences in the mean raw slope or z-scored slope by smoking status. Ever- and never-smokers, respectively, had mean raw slope values of 6 and 2 ml/yr (p < 0.355) and z-scores of 11 and 4 ml/yr (p < 0.140).

When the smokers were dichotomized by persistence of smoking into those who currently smoke at least five cigarettes per day (persistent smokers) versus those who currently smoke less than five cigarettes per day (which includes never-smokers and ex-smokers), the adjusted mean slopes in the parent population were 19 and 14 ml/yr, respectively (p < 0.005), and 13 and 4 ml/yr, respectively, in the offspring population (p < 0.048).

	DISCUSSION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

In multiple analyses, we consistently found significant correlation among adult sibling pairs in the rate of decline of FEV₁. Although there was higher correlation in the standardized residual FEV₁ slopes of smoking-concordant sibling pairs (r = 0.483, p < 0.01, n = 39) than in nonsmoking-concordant sibling pairs (r = 0.280, p < 0.05, n = 63), these correlations were not significantly different (z test, p < 0.258). Smoking-discordant sibling pairs, however, showed no significant correlation in the slope of FEV₁.

Several studies have reported that even short periods of smoking can lead to significant reductions in FEV₁ (1, 2, 5, 31). When the adult offspring population (mean age 27 yr) was dichotomized into current, persistent smokers versus nonpersistent and never-smokers, we found significant differences in the adjusted mean slopes by smoking habit. However, when we classified the offspring population by ever versus never smoking status, there were no significant differences in raw or adjusted mean slopes. Camilli and colleagues (2), in an analysis from the same general TESAOD population, reported no significant effects of light smoking (defined as less than 10 cigarettes per day), and observed the most marked effects of smoking at 50 to 69 yr of age.

It is possible that parental or sibling smoking or some other aspect of a shared past home environment could significantly influence the development of lung function in siblings. However, the siblings in our study were adults at the time of their PFTs, and presumably living independently of their families. If the sibling correlation in lung function decline is a result of a common familial environment, those siblings discordant in smoking habit would presumably also show some correlation.

The high correlation of FEV₁ residual slope values among concordant-nonsmoking siblings suggests that environmental effects of smoking are not the source of the observed familial correlations. Further corroboration comes from the finding of no correlation in lung function decline between smoking-concordant spousal pairs. Genetic susceptibility to smoking may be evidenced in the especially high correlation in slope of FEV₁ among smoking-concordant siblings, and especially in the high proportion of smoking siblings who fall in the lowest tertile of lung function. The lack of correlation among siblings discordant in smoking habit suggests that familial correlation may be attenuated by genetic susceptibility to smoking effects in some or all of the smoking siblings in the discordant pairs.

If FEV₁ decline is genetically determined, it is necessary to reconcile the lack of correlation between parents and offspring. Evidence of high sibling correlation in the absence of parent-offspring correlation is compatible with a recessive model of inheritance. An alternative explanation is that the mechanisms involved in determining the decline in lung function early in adult life are different from those determining loss of lung function in older adults.

Evidence of a plateau phase in growth of pulmonary function in early adulthood has been extensively debated (6, 8, 15, 31), but could be considered a possible explanation for this lack of correlation between parents and offspring, if older adults are declining and their young adult offspring are still increasing or in a plateau state. Tager and colleagues (31) describe a prolonged plateau phase up to the age of 35 yr in male nonsmokers, a less discernible plateau phase in female nonsmokers, and no evidence of a plateau phase in male and female current smokers.

In our population, the small percentage of subjects younger than 45 yr of age who had significant slopes suggests a plateau phase. Based on simple linear regression of each subject's FEV₁ data, only 15.1% had significantly negative slopes and 2.6% had significantly positive slopes. However, because 58.0% of those younger than 45 yr of age with significant negative slopes were current or ex-smokers, and because more than twice as many ever-smokers than never-smokers in our population had raw and residual slope values in the lowest tertile for lung function, our results imply an early decline of FEV₁ in ever-smokers. A number of other researchers have also reported early onset of FEV₁ decline in young adult smokers (2, 15, 31).

The lack of parent-offspring correlation in lung function decline could also be evidence of a "survivor effect" (7, 8, 16). We found that the mean raw slope values for subjects in the population over 70 were less negative than those for subjects in their 60s. Higher mortality in older subjects with higher rates of decline may have resulted in their underrepresentation in the analysis of parent-offspring correlations.

That sibling correlation in FEV₁ decline remained significant regardless of stratification by smoking status suggests that it is not an artifact. Although concordance of smoking habit appears to critically influence the correlation in the slope of FEV₁ among adult siblings, smoking effects alone do not explain the observed correlations. Instead, we may be seeing evidence of genetic susceptibility to an accelerated rate of decline associated with smoking, causing a lack of correlation among adult siblings who are discordant in smoking habit and increased correlation among smoking-concordant sibling pairs.