The Framingham Offspring cohort was started between 1971 and 1975 and includes 5,124 males and females (8). Participants are the adult children, and their spouses, of the participants in the original Framingham cohort, a seminal study that started in 1948 (10).

The Framingham Offspring participants underwent periodic clinical examinations every 4 years up to exam 7, conducted in 1998. In the analysis presented here, we include data obtained in exam 1 (1971–1975), exam 2 (1979–1982), exam 5 (1991–1995), and exam 6 (1996–1997), all of which had reliable spirometric measurements and relevant clinical information according to internal National Heart, Lung, and Blood Institute–National Institutes of Health standards (10). We included in the analysis all participants with two or more valid spirometry measurements during follow-up (n = 4,391; age range at baseline 13–71 yr). Permission to analyze the Framingham database was obtained from the National Heart, Lung, and Blood Institute–National Institutes of Health according to a prespecified research proposal previously approved by our Institutional Review Board.

Spirometry was performed using a 6L Collins water-sealed bell spirometer connected to an Eagle II microprocessor that provided automatic correction for body temperature, pressure saturated with water vapor conditions, based on calibrations performed daily by technicians. Measurements were obtained with the subject standing and wearing nose clips. Maneuvers were repeated (up to a maximum of 8) until at least three acceptable spirograms were obtained. The largest FVC and the largest FEV₁ from all acceptable maneuvers were selected (10), as per later recommendations (11). No bronchodilator was used in any of the Framingham cohort studies, so all values reported are prebronchodilator (pre-BD).

We considered “never-smokers” those individuals who declared that they never smoked before and during the study, “continuous smokers” those who declared active smoking in all available examinations, and “former smokers” those smokers who declared not smoking during the last year in one of the examinations (10). We did not include in the analysis intermittent smokers. Pack-years were calculated as the number of packs of cigarettes smoked per day times how many years the participant smoked.

We considered “healthy” participants those who reported “No” responses to all questions on dyspnea on exertion, increased dyspnea, nocturnal dyspnea, nocturnal cough or wheezing at exam 1, and/or were never reported by an examiner with a clinical diagnostic impression of asthma, chronic obstructive lung disease, chronic bronchitis, or emphysema during follow-up examinations. By contrast, those fulfilling affirmatively any of these criteria were considered “unhealthy,” as per standard published Framingham Original and Offspring cohorts methodology and definitions (10).

We considered airflow obstruction as defined by a pre-BD FEV₁/FVC ratio lower than 0.7 and an FEV₁ value lower than 80% of predicted.

Results are presented as mean ± SD and 95% confidence intervals (95% CI) or percentage, as appropriate. Mean comparisons between groups were analyzed by analysis of variance (followed by post hoc test) or Welch's t test, as required. To best fit the curves describing FEV₁ changes with age to raw data, we used a generalized additive model (12), with smooth terms and incorporating the subject-specific (longitudinal) random effects. Scatter plots smoother of Gaussian-type like splines are also included. Statistical analysis was performed using R statistical software version 2.6.1 (13) and the package mgcv 1.3–27 (14). In this analysis, we expressed FEV₁ in absolute values (liters) or, as Fletcher and Peto did, as percentage of the predicted value at the age of 25 years (3). To calculate the predicted value at the age of 25, we used the Third National Health and Nutrition Examination Survey (NHANES-III) Hankinson's reference equations for whites (15), according to age, sex, and height, because after reviewing the website and related publications it was assumed that all Framingham Offspring participants were white (10). Because the latter was not always measured when the participant was precisely at the age of 25, we used the closest height available after this age, to avoid using a height value of someone who might be still growing. Finally, to calculate FEV₁ decline in the different study groups we first calculated it in each participant by dividing the difference in milliliters between measurements obtained in the first and last available spirometry in each individual, out of a maximum of four (as per Figure 1 and Table E2) by the time (years) between these measurements, and individual values were averaged. The slope from the first adult measurement (after 20 yr) to the last available lung function measurement was adjusted by body mass index, height, and age for comparisons within groups. A P value lower than 0.05 was considered statistically significant.

Figure 1. Flow diagram of participants from the Framingham Offspring cohort by available spirometry and smoking information in exams 1, 2, 5, and 6.

Download Figure | Download Powerpoint

[More] [Minimize]

From the original 5,124 participants at inception of the Framingham Offspring cohort, 4,391 (85.7%) contributed with at least two lung function measurements for this study, and the process of determining smoking status and other related information is presented in Figure 1. A comparison of baseline characteristics with the 699 participants with a single spirometry measurement, not included in later analyses, is presented in Table E3. These participants with only one spirometry measurement had a shorter follow-up, were slightly older (significant only in men), and had a reduced smoking exposure history (all P < 0.05) when compared with participants included in this research.

Table 1 presents the demographic and clinical characteristics of the 2,121 males and 2,270 females included in the study at their first examination evaluation. Of them, 1,017 contributed with two lung function measurements, 1,269 with three, and 2,105 with four. Mean age at the time of the first available spirometry was 36 years (range from 13–71 yr) in both males and females. More males were current or former smokers than females (P < 0.05), and the mean pack-years were 31.2 for males and 19.0 for females. FEV₁ in absolute values (L) was greater in males than in females. Table 1 also presents FEV₁ and FVC values as percent predicted, as well as percent predicted at age 25 for FEV₁.

Figure 2 presents FEV₁ changes with age in healthy never-smoking males and females, expressed in absolute values (Figure 2A) and as percent predicted at the age of 25 (Figure 2B). As expected, males had higher absolute FEV₁ values during their lifetime than females (Figure 2A). Three other observations are of note: First, in males, FEV₁ values increased up to a peak at the age of 23 years but not in females in the age range of this study. Second, FEV₁ in females (but not males) plateaus until about the age of 40 years. Third, the annualized rate of decline for healthy never-smoking males (19.6 ml/year; 95% CI, 17.1–22.1 ml/yr) was slightly but not significantly greater (P = 0.266) than that observed in females (17.6 ml/yr; 95% CI, 13.8–21.4 ml/yr). Similar results were observed when FEV₁ was expressed as percent of the predicted value at the age of 25 (Figure 2B).

Figure 2. Mean FEV₁ values by age (and 95% confidence intervals [CI]) in male (black line) and female (gray line) healthy never-smokers, expressed (A) in absolute values and (B) as percentage of the FEV₁ value at the age of 25. The mean FEV₁ decline value (and 95% CI) for males was 19.6 ml (17.1–22.1) and for females 17.6 ml (13.8–21.4), with a P value of 0.266. For further explanation, see text.

Download Figure (Part 1) | Download Figure (Part 2) | Download Powerpoint

[More] [Minimize]

For completeness, the same figures are presented for FVC (Figure E4) and the ratio FEV₁/FVC (Figure E5).

The mean age at starting smoking in this cohort was 17.5 ± 3.6 years in male and 18.8 ± 4.2 years in female participants. Figure 3 compares FEV₁ decline in healthy never-smokers and continuous smokers both in males (Figure 3A) and females (Figure 3B). We observed, first, that smoking reduced the maximal FEV₁ value achieved in males, but not in females. Second, continuous smoking increased the rate of FEV₁ decline (vs. never-smokers), in males (38.2 ml/yr; 95% CI, 33.9–42.6), and females (23.9 ml/yr; 95% CI, 20.9–27.0). Third, the rate of FEV₁ decline was higher in continuously smoking males than females (P < 0.001). We also confirmed that there is significant variability in the rate of decline of lung function in continuous smokers, both in males (from 8 to 63 ml/yr) and females (from 14 to 49 ml/yr) (first vs. fourth quartiles of lung function decline, data not shown).

Figure 3. Mean FEV₁ values (expressed as percent of its value at the age of 25) by age, for healthy never-smokers (NS), and continuous smokers (CS). (A) Data for males and (B) for females. The mean FEV₁ decline value (and 95% confidence intervals) for males was 38.2 ml (33.9–42.6) and for females 23.9 ml (20.9–27.0), with a P value < 0.001. *P < 0.05 versus healthy never-smokers. For further explanation, see text.

Download Figure (Part 1) | Download Figure (Part 2) | Download Powerpoint

[More] [Minimize]

It is often acknowledged that only a fraction of smokers (the so-called susceptible smokers) will develop airflow obstruction. According to our definition (pre-BD FEV₁/FVC ratio less than 0.7 and an FEV₁ value less than 80% of predicted), we found that 33.0% of continuous-smoker males and 24.2% of continuous-smoker females developed airflow obstruction during follow-up, higher than the proportions observed in our group of never-smokers, 7.4% of men and 5.6% of women. To further explore the issue of susceptibility to smoking in the Framingham Offspring cohort, we investigated if the presence of respiratory symptoms at baseline and/or a respiratory diagnosis in continuous smokers (unhealthy smokers; see operational definitions above) identifies a group of individuals particularly susceptible to smoking. As shown in Figure 4, the rate of FEV₁ decline in unhealthy continuous smokers was increased versus healthy continuous smokers and versus healthy never-smokers, in both sexes (P < 0.05).

Figure 4. Mean FEV₁ values (expressed as percent of its value at the age of 25) by age, for healthy never-smokers (NS), healthy continuous smokers (HCS), and unhealthy continuous smokers (UCS). (A) Data for males and (B) for females. The mean FEV₁ decline value (and 95% confidence intervals) for males in each subgroup were 32.5 ml (24.9–40.2) for healthy continuous smokers and 42.3 ml (37.2–47.3) for unhealthy continuous smokers; for females, 17.1 ml (12.9–21.2) for healthy continuous smokers and 30.7 ml (26.3–35.0) for unhealthy continuous smokers. *P < 0.05 versus healthy never-smokers. ^#P < 0.05 versus healthy continuous smokers. For further explanation, see text.

Download Figure (Part 1) | Download Figure (Part 2) | Download Powerpoint

[More] [Minimize]

Figure 5 presents the effects of quitting smoking at different ages in males (Figure 5A) and females (Figure 5B). Smokers who quit before the age of 30, between 30 and 40 years of age, and after the age of 40 years are compared with healthy never-smokers and continuous smokers. The rate of FEV₁ decline in both male and female smokers who quit before age 30 was indistinguishable from healthy never-smokers. In contrast, smokers who quit after the age of 40 years showed a significantly enhanced rate of decline of FEV₁ versus healthy never-smokers and earlier quitters, but not significantly different from continuous smokers (Figure 5).

Figure 5. Mean FEV₁ values (expressed as percent of its value at the age of 25) by age in smokers who quit smoking before the age of 30 (Q < 30), between 30 and 40 years of age (Q30–40) and after the age of 40 (Q40+). Curves from healthy never smokers (NS) and continuous smokers (CS) are also included for comparison. (A) Results in males. (B) Results in females. The mean FEV₁ decline value (and 95% confidence intervals) for each curve of quitters for males was 15.5 ml (11.3–19.8) in quitters before the age of 30, 24.0 ml (20.0–28.1) in quitters between 30 and 40 years of age, and 28.9 ml (26.4–31.1) in quitters after the age of 40; for females, 10.4 ml (6.3–14.5) in quitters before the age of 30, 16.5 ml (14.0–19.0) in quitters between 30 and 40 years of age, and 21.0 ml (18.8–23.2) in quitters after the age of 40. *P < 0.05 versus healthy never-smokers (NS). ^#P < 0.05 versus CS. For further explanation, see text.

Download Figure (Part 1) | Download Figure (Part 2) | Download Powerpoint

[More] [Minimize]

Several previous population-based studies have analyzed lung function decline over time (16–21). To the best of our knowledge, however, this is the first time that a large cohort of both males and females, with a wide age range, is followed for up to 26 years to describe lung function changes from adolescence to old age in healthy never-smokers and to investigate the effects of smoking and smoking cessation.

Our current understanding of the natural history of COPD is mostly based on the seminal publication of Fletcher and Peto from the MRC cohort (3, 4). Based on a limited number of observations, these authors proposed that FEV₁ declines physiologically with age, that this decline was significantly enhanced in a subgroup of smokers (so-called susceptible smokers) and that the rate of FEV₁ decline would be “normalized” after quitting smoking at any age (3, 4). We believe that our study is the first attempt to recreate that model with data.

Within the age range studied, we could identify a peak of FEV₁ only in males (Figure 2). This is in keeping with other studies (22) and suggests that females achieve lung growth earlier than males. In these Framingham Offspring participants, males started smoking at a mean age of 17.5 years, when the lung is still growing. This was not the case in females, who started smoking at a mean age of 18.8 years, when their lungs are fully developed. Nowadays, teens start smoking much earlier, which could have even more harmful effects on lung growth.

We also observed that females appear to have a longer FEV₁ plateau phase, as other studies have reported in the past (16, 21–23), and that their annual rate of FEV₁ decline with age was slightly but not significantly lower than in healthy never-smoking males (17.6 vs. 19.6 ml/yr, P = 0.266) (Figure 2). The above-mentioned plateau phase refers to pooled group data, but large variability at the individual level and by subgroups of individuals has been reported elsewhere (24). The trend in Figure 2A of a small FEV₁ increase in absolute values in males in the oldest age categories could indicate a healthy survivor bias, or mere variation due to a shrinking sample size, as this trend is not seen in women.

Finally, previous longitudinal studies have published figures of FEV₁ decline in never-smokers that range from 10 to 50 ml/yr (6). Our results are within this range but it is of note that we used a more stringent definition (healthy never-smokers) than most previously published longitudinal studies, which reported results on never-smokers.

Our results confirm that tobacco smoking increases the rate of lung function decline in males and females (16–18, 20, 24–27) (Figure 3). There is also significant variability in the rate of decline of lung function in continuous smokers and, in particular, the rate of FEV₁ decline was slightly but significantly higher in continuous smoking males than females (Figure 3). This is consistent with data from the Tucson respiratory study 20 years ago (28). However, in our study the percentage of continuous smokers developing airflow limitation was similar in males (33%) and females (24%), suggesting that males and females may be equally susceptible to the deleterious effects of tobacco smoking.

Fletcher and Peto reported that phlegm production or infective episodes did not increase lung function decline (3, 4). In contrast we found that the presence of respiratory symptoms at baseline and/or a respiratory diagnosis during follow-up appears to identify a subpopulation of smokers who are particularly susceptible to the deleterious effects in their lungs of smoking, extending the conclusions of a recent large European follow-up study (29). This novel information has important clinical relevance.

Finally, we observed that smoking appeared to reduce the FEV₁ peak in males (Figure 3A). Because we did not identify such a peak in females (Figure 3B) we do not know if this occurred earlier also in this group. In any case, at least in males, our results suggest that smoking during adolescence may reduce lung growth as has been shown by others (27). The clinical importance of this observation is that failure to reach optimal lung growth is a potential risk factor for the subsequent development of airflow obstruction later in life (28, 29).

Many previous studies have investigated the effects of quitting smoking on lung function decline (16, 28). We found that smoking cessation had a beneficial effect on the rate of FEV₁ decline by age (Figure 5). However, previous studies generally investigated the beneficial effects of quitting in middle-aged individuals only (mean age of about 45 yr) (16, 28, 30). In contrast, our study explored the effects of quitting at younger ages. We found that the rate of FEV₁ decline of smokers (both males and females) who quit before the age of 30 was indistinguishable from that of healthy never-smokers, but it was significantly enhanced (and not different from that of continuous smokers) in those who quit after the age of 40 (Figure 5). The observation that the rate of FEV₁ decline between smokers who quit after the age of 40 and continuous smokers is not significantly different could be due to lack of power, because in Figure 5 in males and also in females, the curves appear to have different slopes. It is not our intention to send a misleading and potentially discouraging public health message by implying that late quitting may not have a beneficial impact on the course of subsequent lung function, as this beneficial effect was elegantly reported in the Lung Health Study (31, 32). On the contrary, all these observations highlight the importance of early quitting and do not support the model proposed by Fletcher and Peto, in which it was suggested that the rate of lung function decline returns to normality after quitting independently of the age of the individual (3, 4).

Our study has some strengths and potential limitations that deserve comment. Among the former, the inclusion of female participants, its large sample size, its long follow-up time, and the study of individuals both during adolescence and old age are particularly worth noting. Among the latter, we acknowledge that, first, reliable lung function data were available in only four examinations (Figure 1 and Table E2), and that the accuracy of most estimates would have been larger if more spirometries had been available in the same individual, reducing any potential cohort effect (33). Connected with this, there is potential bias due to applying the Hankinson's NHANES-III prediction equations (15), which are more recent, to the Framingham data, but it appears there is no other satisfactory alternative. It is counterintuitive that Framingham healthy, never-smoker participants have on average lower than the 100% expected mean percent predicted spirometry values according to Hankinson's equations. A likely cohort effect is envisioned, as generations grow healthier and leaner, similar to the one demonstrated in the Dutch population (33). Second, all information on smoking and quitting was self-reported and not objectively (biologically) verified, although it has been extensively used to date (10). Third, several potential confounders, such as passive smoking and environmental, occupational, and allergen exposures, were not available and were therefore not taken into account in the analysis. Fourth, one of many ongoing controversies in COPD research is the use of a fixed ratio or a lower limit of normal to diagnose airflow obstruction (34). As per current Global Initiative for Chronic Obstructive Lung Disease (GOLD) recommendations, we used the former, but it is worth noting that spirometric results correspond to pre-BD values, and because the FEV₁/FVC ratio decreases with increasing age, as in Figure E5, any fixed-ratio definition leads to the potential for overestimation of airflow obstruction in older participants and underestimation of airflow obstruction in young participants. Finally, an inherent limitation of any serial spirometric assessment is that individuals who die, or are very sick, cannot possibly contribute more lung function data. This selection bias is strongly related to smoking status, as it can be seen in Table E6. Death was more frequent in continuous smokers (33% of male and 19% of female) than in the respective comparator groups of never-smokers or intermittent smokers. Accordingly, any results on fast decliners, or other individuals more susceptible to the harmful effects of smoking, are diluted due to this survivor bias. Finally, and not an original objective of our research, we confirm that survival of continuous smokers is much reduced when compared with never-smokers, both in males and in females (Figure E7).

Stanojevic and colleagues (35) have recently proposed a new way of collating and analyzing spirometric data from population surveys, aimed at developing reference ranges that more accurately describe the nonlinear relationship between spirometric lung function and height by age, by means of an extension of the LMS (lambda, mu, sigma) method. We agree that this new modeling technique provides an elegant solution and a statistically robust means of developing continuous reference ranges from early childhood to old age. However, given the limitations stated above, and as we are not pooling different surveys, we decided to sustain the analytical approach as per the original UK MRC study and curves.

There is an urgent need to agree on the interpretative strategies of pulmonary function, at the individual and population levels, and this is a matter of ongoing debate (6). Quanjer and colleagues (36) have elegantly discussed the advantages of using reference equations that use height, age, or other parameters, such as z-scores, in schoolchildren and adolescents in particular. We concur with previous investigators (35, 36) that limitations remain for describing the progression of pulmonary function and the natural course of respiratory disease, as highlighted in our research.

Our results provide longitudinal evidence that recreates the Fletcher and Peto lung function curves in males and, for the first time, extends them to females. This allowed us to identify sex differences in normal lung growth and aging. We confirm that continuous smoking enhances the rate of FEV₁ decline and causes airflow obstruction only in a percentage of smokers. Contrary to Fletcher and Peto (3), the presence of respiratory symptoms at baseline and/or a respiratory diagnosis during follow-up appears to identify a particularly susceptible phenotype, which may be helpful in clinical practice. Our results clearly support the benefits of quitting smoking at early ages.