Decline in FEV₁ in Relation to Incident Chronic Obstructive Pulmonary Disease in a Cohort with Respiratory Symptoms

Abstract

Data on the relationship between decline in lung function and development of COPD are sparse. We assessed the decline in FEV₁ during 10 years among subjects with respiratory symptoms by two different methods and evaluated risk factors for decline and its relation to incident Chronic Obstructive Pulmonary Disease, COPD. A cross-sectional postal questionnaire was in 1986 sent to 6610 subjects of three age strata. All subjects reporting respiratory symptoms were invited to a structured interview and spirometry. A follow-up survey was performed 10 years later, and totally 1109 subjects performed spirometry in both 1986 and 1996. COPD was defined according to the ATS/ERS standards (FEV₁/FVC ≤0.70). The decline in FEV₁ was 39 ml/year in men vs. 28 ml/year in women, p = < 0.001 (−1.53 vs. −0.12 change in percent of predicted normal value over 10 years (pp), p = 0.023), among smokers 39 vs. non-smokers 28 ml/year, p < 0.001 (−3.30 vs. 0.69 pp, p < 0.001), in subjects with chronic productive cough 36 vs. not 32 ml/year, p = 0.044 (−2.00 vs. −0.02 pp, p = 0.002). Incident cases of moderate COPD (n = 83) had a decline of 62 ml/year (−12.6 pp) and 22.9% of them had a decline > 90 ml/year (−27.8 pp over 10 years). Gender-specific analysis revealed that smoking was a stronger risk factor in women than in men, while higher age was a significant risk factor in men only. In conclusion, decline in FEV₁ was associated with age, smoking, and chronic productive cough, but the risk factor pattern was gender-dependent. Among incident cases of COPD the decline was steeper and close to a quarter had a rapid decline.

Key words: :

• Incidence
• COPD
• Decline in FEV₁
• Risk factors
• Respiratory symptoms
• Gender

INTRODUCTION

Chronic Obstructive Pulmonary Disease (COPD) is characterized by obstructive lung function impairment and symptoms such as cough, sputum production and dyspnea. The progressive lung function impairment among affected subjects gradually limits physical activity, decreases quality of life, and may result in chronic respiratory failure ([1]).

There are several reports on the prevalence of COPD, most in the range from 4 to 10% ([2], [3], [4], [5], [6], [7]). Reports on incidence of COPD are few and not consistent with respect to the definition of COPD ([8], [9], [10], [11], [12]). Even though smoking and age are well known risk factors for COPD ([2], [4], [5], [6], [7]), the inter-individual difference in regard to development of COPD among those at risk is not fully understood. Since obstructive lung function impairment is fundamental for the diagnosis of COPD, we presume that the rate of decline in FEV₁ is a marker of risk for developing COPD. The classical report by Fletcher demonstrated an accelerated decline in lung function among smoking men ([13]). Decline in FEV₁ is most often estimated as mean ml/year and sometimes corrected for age, gender and length ([14], [15], [16], [17], [18]). It is common knowledge that smoking deteriorates lung function in both men and women ([14], [15], [16], [17]), but there are also other risk factors such as occupational airborne exposure, air pollution, low socioeconomic status, low educational level, family history of obstructive airway disease, airway hyper-responsiveness, respiratory tract infections, mucus hyper secretion and other respiratory symptoms ([1], [14], [16], [19]).

The epidemiological project Obstructive Lung Disease in Northern Sweden Studies (OLIN) has collected longitudinal data since 1985 with a focus on obstructive airway diseases. The first cohort, an age-stratified general population sample, was recruited in 1985. All subjects reporting any respiratory symptom in a postal questionnaire survey were examined in 1986. The original aim of the first cross-sectional survey was to estimate the prevalence of respiratory symptoms and obstructive airway diseases, thus all subjects reporting respiratory symptoms were invited to structured interview and spirometry ([20], [21]).

The cohort was re-examined in 1996, and based on the longitudinal data the cumulative incidence of COPD among subjects with respiratory symptoms has been estimated ([12]). Though a longitudinal follow up was not one of the original aims, this study population is highly clinically relevant. The examined subjects are considered representative for subjects in the society with respiratory symptoms and thus represent the population from which our patients seeking medical advice due to respiratory complaints are recruited.

The main aim of this study was to estimate decline in FEV₁ and identify risk factors for accelerated decline during a 10-year period in this symptomatic cohort and also in relation to development of COPD according to the spirometric criteria of the American Thoracic Society and the European Respiratory Society, the recently published ATS/ERS standards ([22]). A second aim was to evaluate to what extent two different methods of assessing decline in FEV₁, crude mean ml/year and units change in percent of predicted normal value, affected the results.

MATERIAL AND METHODS

Study population

In 1986, a postal questionnaire regarding respiratory symptoms and diseases was sent to all 6610 subjects born 1919–1920, 1934–1935 and 1949–1950 in eight areas of northern Sweden. The response rate was 86% (n = 5698) ([20]). All subjects reporting any respiratory symptom as cough, sputum, chronic productive cough, wheeze or dyspnea (n = 1655) were identified and invited to structured interviews and spirometry where 1506 subjects (91%) participated ([21]). In 1996, they were invited to a re-examination, and totally 1109 subjects performed spirometry in both 1986 and 1996 ([12]). In 1986, 23.8% of men were non-smokers, 36.0% ex-smokers and 40.2% smokers (in women 47.9%, 20.0% and 32.9%). Baseline data are shown in Table 1. The main reason for not participating at follow-up was death (n = 165).

Table 1 Basic characteristics of study population at entry in 1986 (N = number of subjects, M = men, W = women), including lung function (FEV₁ percent predicted, pp) by gender, age group, smoking categories, and chronic productive cough

Variable	Category	N (M/W)	FEV1 pp (M)	FEV1 pp (W)	p-value	All	p-value
Age group	Born 1919–1920	156/145	90.53 (±18.58)	89.17 (±21.05)	0.875	90.35 (±19.77)
	Born 1934–1935	230/202	92.39 (±16.12)	90.20 (±17.27)	0.174	91.37 (±16.69)	0.466^= < 0.001
	Born 1949–1950	189/187	94.98 (±13.48)	95.32 (±11.11)	0.792	95.15 (±12.34)	< 0.001^=
Smoking	Non-smoker	126/252	95.83 (±15.40)	94.09 (±16.55)	0.326	94.67 (±16.18)
	Persistent ex-smoker	198/44	92.39 (±17.48)	92.65 (±16.28)	0.903	92.47 (±17.07)	0.089^§
	Quitter	80/55	94.19 (±16.16)	88.75 (±18.55)	0.073	91.00 (±15.08)	0.103^§
	Restart smoking	9/13	89.89 (±15.88)	95.85 (±8.04)	0.259	93.41 (±11.92)	0.719^§
	Persistent smoker	151/116	89.81 (±14.43)	88.06 (±16.88)	0.363	89.05 (±15.54)	< 0.001^§
Chron prod cough	No	317/327	94.91 (±14.76)	93.98 (±15.55)	0.442	94.44 (±15.16)
	Yes	258/207	90.07 (±17.20)	88.83 (±18.13)	0.450	89.52 (±17.66)	< 0.001^‖

*p-value comparing gender.

⁼ p-value (t-test) comparing means between age group born 1919–1920 and 1934–1935, 1919–1920 and 1949–1950.

^‡ p-value (t-test) comparing means between age group born 1934–1935 and 1949–1950.

^§ p-value (t-test) comparing means between non-smokers and the other five smoking categories, one at a time.

^‖p-value (t-test) comparing means between chronic productive cough positive or negative answer.

Methods

The postal questionnaire and the interview questionnaire were developed from the British Medical Research Council questionnaire ([23]). The questionnaires have previously been described in detail ([20], [24]) and have been used in studies in several countries ([7], [20], [24], [25]). The spirometry was performed according to the ATS recommendations ([26]). All lung function tests, 1986 and 1996, have been performed on four parallel sets of the dry volume spirometer Minjhard Vicatest 5, the Netherlands. The spirometers have been calibrated in a standardised manner at the start of every working day. A limited number of well-trained nurses and research-assistants have performed all lung function tests. Reversibility tests were performed in 1996 but not in 1986. In order to use comparable spirometric values when estimating decline in lung function from 1986 to 1996 only pre-bronchodilator values have been used. Swedish reference values were used ([27]), which reflect well the adult population of Norrbotten ([21]).

Definitions

The ATS/ERS standards spirometric classification of COPD ([22]), FEV₁/FVC ≤0.70, was used, including disease severity by level of FEV₁ (mild ≥80% predicted, moderate 50–80% predicted, severe 30–50% predicted and very severe < 30% predicted). The ATS/ERS criteria were used in a modified form as pre-bronchodilator values were used. The mean decline (ml/year) was calculated as the crude difference in pre-bronchodilator FEV₁divided by the number of years observation time, and subdivided into the following groups: ≤30 ml/year, > 30– ≤ 60 ml/year, > 60– ≤ 90 ml/year, and > 90 ml/year. Decline in FEV₁ was also estimated as units' change of percent predicted normal value over 10 years (pre-bronchodilator FEV₁% predicted 1996-pre-bronchodilator FEV₁% predicted 1986). The ATS/ERS standards do not exclude asthma as a cause of chronic airway obstruction, thus subjects with self-reported asthma fulfilling the spirometric criteria of airway obstruction have not been excluded.

Smoking habits at the interviews in 1986 and 1996 were divided into the six smoking categories (Figure 1); non-smoker (non-smoker–non-smoker), persistent ex-smoker (ex-smoker–ex-smoker), quitter (smoker–ex-smoker), re-starter (ex-smoker–smoker), persistent smoker (smoker–smoker) and other (difficult to classify).

Figure 1 Smoking categories, by age group (year of birth) and gender.

The Swedish socioeconomic classification system based on occupation, SEI ([27]), was used. The socioeconomic group used in the analyses were divided into professionals and executives, assistant non-manual employees, manual worker in industry, manual worker in service, self-employed non-professionals, housewives, and occupation unknown.

Analysis

Statistical calculations were made using the software Statistical Package for the Social Sciences (SPSS) version 10.0. The chi-squared test was used for bi-variate analysis of proportions and test-for-trend, and the t-test for comparing means. The 95% significance level (p < 0.05) was used. Interview-data from 1986 regarding presence of cough, sputum production, chronic productive cough, dyspnea (corresponding to Medical Research Council dyspnea scale II), wheeze and attacks of shortness of breath were analysed. The change of FEV₁ was estimated separately in men and women by linear regression in a multivariate model adjusting for age, smoking categories, family history of obstructive lung disease and socioeconomic status (appendix).

A multivariate logistic regression model including the above independent variables and gender was also created with decline > 60 ml/year compared to decline ≤ 30 ml/year as dependent variable. The incident cases of COPD were subjects fulfilling the spirometric criteria of COPD in 1996 but not 1986 (thus the prevalent cases of COPD in 1986 were excluded in the population at risk for developing COPD).

RESULTS

Changes in lung function and decline in FEV₁

Lung function deteriorated significantly during the 10 years of observation time (Table 2). The mean decline in FEV₁ in the study population was 33 ml/year (−0.85 change in percent of predicted normal value over 10 years). Non-smokers had a significantly lesser decline in FEV₁ (absolute value and percent of predicted) compared to re-starters and persistent smokers. Re-starters tended to have an even greater decline than smokers (Table 3 and Table 4).

Table 2 Lung function in 1986 and 1996, by gender

Categories	Year	Variable	Men	Women	p-value^=	All
All	1986	FEV1	3.47 (±0.79)	2.47 (±0.60)	< 0.001	2.99 (±0.87)
	1996	FEV1	3.08 (±0.83)	2.20 (±0.59)	< 0.001	2.66 (±0.85)
		p-value^‡	< 0.001	< 0.001		< 0.001
	1986	FEV1 pp	92.74 (±16.11)	91.99 (±16.77)	0.446	92.38 (±16.43)
	1996	FEV1 pp	91.21 (±18.96)	91.87 (±18.11)	0.557	91.53 (±18.55)
		p-value^‡	0.001	0.777		0.006
	1986	FEV1/FVC	0.77 (±0.09)	0.78 (±0.09)	< 0.001	0.77 (±0.08)
	1996	FEV1/FVC	0.74 (±0.10)	0.76 (±0.09)	0.001	0.75 (±0.10)
		p-value^‡	< 0.001	< 0.001		< 0.001

*mean ± standard deviation.

⁼comparing gender.

^‡comparing 1986 and 1996.

Table 3 Gender-specific mean decline in FEV₁ (ml/year ± SD) by smoking categories, age group, and chronic productive cough

Variable	Category	Men	Women	p-value	All	p-value
Gender		39 (±34)	28 (±23)	< 0.001	33 (±30)
Smoking	Non-smoker	36 (±40)	25 (±21)	< 0.001	28 (±29)
	Persistent ex smoker	35 (±32)	27 (±22)	0.018	32 (±30)	0.081^=
	Quitter	43 (±34)	24 (±21)	< 0.001	35 (±31)	0.020^=
	Re-starter	44 (±35)	43 (±38)	0.911	43 (±36)	0.024^=
	Persistent smoker	43 (±32)	35 (±22)	0.034	39 (±28)	< 0.001^=
Age group, born	1919–20	41 (±39)	26 (±22)	< 0.001	34 (± 32)
	1934–35	43 (±35)	29 (±23)	< 0.001	36 (±31)	0.264^‡
	1949–50	33 (±30)	28 (±23)	0.052	30 (±26)	0.136^‡, 0.003^§
Chron productive cough	Yes	40 (±37)	30 (±24)	< 0.001	36 (±32)
	No	38 (±33)	26 (±21)	< 0.001	32 (±28)	0.044^‖

*p-value (t-test) comparing means between men and women.

⁼ p-value (t-test) comparing means between non-smokers and the other five smoking categories, one at a time.

^‡ p-value (t-test) comparing means between age group born 1919–1920 and 1934–1935, 1919–1920 and 1949–1950.

^§ p-value (t-test) comparing means between age group born 1934–1935 and 1949–1950.

^‖ p-value (t-test) comparing means between chronic productive cough positive or negative answer.

Table 4 Gender-specific decline in FEV₁ estimated as changes from 1986 to 1996 in percent of predicted normal value by smoking habits, age group and chronic productive cough (FEV₁ % predicted 1996-FEV₁ % predicted 1986; negative result is consistent with decline)

Variable	Category	Men	Women	p-value	All	p-value
Gender		−1.53 (±10.7)	−0.12 (±9.8)	0.023	−0.85 (±10.3)
Smoking	Non-smoker	−0.70 (±11.62)	1.38 (±9.77)	0.068	0.69 (±10.5)
	Persistent ex-smoker	−0.16 (±10.37)	0.37 (±9.10)	0.672	0.01 (±9.96)	0.398^=
	Quitter	−3.00 (±11.83)	1.25 (±8.24)	0.023	−1.27 (±10.69)	0.064^=
	Re-starter	−3.11 (±10.47)	−6.31 (±13.12)	0.550	−5.00 (±11.95)	0.014^=
	Persistent smoker	−2.95 (±9.63)	−3.75 (±9.58)	0.499	−3.30 (±9.60)	< 0.001^=
Age group, born	1919–1920	−2.16 (±13.59)	0.31 (±11.06)	0.086	−0.97 (±12.47)
	1934–1935	−2.80 (±10.69)	−0.80 (±10.42)	0.050	−1.86 (±10.60)	0.297^‡
	1949–1950	0.55 (±7.37)	0.28 (±7.83)	0.734	0.41 (±7.59)	0.076^‡ 0.001^§
Chron productive cough	Yes	−2.50 (±11.74)	−1.38 (±10.59)	0.287	−2.00 (±11.25)
	No	−0.74 (±9.80)	0.68 (±9.16)	0.059	−0.02 (±9.50)	0.002^‖

*p-value (t-test) comparing means between men and women.

⁼ p-value (t-test) comparing means between non-smokers and the other five smoking categories, one at a time.

^‡ p-value (t-test) comparing means between age group born 1919–1920 and 1934–1935, 1919–1920 and 1949–1950.

^§ p-value (t-test) comparing means between age group born 1934–1935 and 1949–1950.

^‖ p-value (t-test) comparing means between chronic productive cough positive or negative answer.

The decline in FEV₁ was significantly higher in men than women independently of method assessing decline (Table 3 and Table 4). The gender differences for FEV₁ decline in different smoking categories and age groups measured by mean ml/year were reduced when decline in FEV₁ was estimated as changes in percent of predicted normal value over 10 years (Table 3 and Table 4).

The annual decline in FEV₁ was less in the youngest age group compared to the middle-aged group. Chronic productive cough at the start of the observation period was associated with a greater decline (Table 3 and Table 4), while cough, sputum production, wheeze, dyspnea and attacks of shortness of breath were not significantly associated with increased decline in FEV₁. Forty-eight percent of the participants had a mean decline of ≤30 ml/year, 38% > 30–60 ml/year, 10% > 60–90 ml/year, and 4% > 90 ml/year corresponding to a mean change in percent predicted of FEV₁ of 6.4, −4.3, −12.8, and −25.2 percent units over 10 years, respectively. The distribution of severity was similar in all age groups.

Decline in FEV₁ among incident cases of COPD

There were 127 (56 women) incident cases of COPD according to the ATS/ERS standards, corresponding to a 10-year cumulative incidence of 13.5%, of whom 60 (31 women) fulfilled the criteria of moderate COPD, and 6 cases (2 women) of severe COPD. Mean decline in FEV₁ among the incident cases was 51 ml/year (mild COPD 43 ml/year, moderate COPD 54 ml/y and severe COPD 102 ml/y) corresponding to −7.4 percent units (−2.8, −10.2, and −26.3) change of mean percent predicted value over 10 years. Among all incident cases 27.6% had a mean decline of ≤ 30 ml/year, 41.7% > 30–60 ml/year, 18.1% > 60–90 ml/year, and 12.6% > 90 ml/year.

When moderate COPD (FEV₁/FVC ≤0.70 and FEV₁ < 80% predicted) was used to define disease, which means that mild COPD at the beginning of the observation period was included in the population at risk, we found 83 incident cases (37 women) of COPD. Among them the mean decline in FEV₁ was 62 ml/year corresponding to −12.6 percent units change of mean percent predicted value over 10 years. Subdivision of the incident cases by decline in FEV₁ revealed: 14.5% ≤30 ml/year, 47.0% > 30–60 ml/year, 15.7% > 60–90 ml/year, and 22.9% > 90 ml/year. More than a half (53.5%) of the incident cases of COPD according to the ATS/ERS standards reported chronic productive cough in 1986 (incident moderate COPD 65.1%), but the decline in lung function did not differ significantly between incident cases with and without chronic productive cough. For comparison, 37.6% of all participants reported chronic productive cough at start of the observation period.

Among the incident cases of COPD the mean decline in FEV₁ was significantly greater in men than women, 57 ml/year vs. 42 ml/year (p = 0.010). When moderate COPD was used to define disease the corresponding figures were 74 ml/year and 47 ml/year (p = 0.001). However, decline in FEV₁ expressed as change of percent predicted normal value over 10 years demonstrated no gender differences by either definition of COPD.

Multivariate analysis

According to the linear regression model (appendix) the change in FEV₁ was significantly influenced by smoking and increasing age (Figure 2). Family history of OLD, socioeconomic status, and chronic productive cough did not significantly affect the decline in FEV₁. In the gender-specific analyses we found that the effect of smoking was significant in women and close to in men (p = 0.055), while the effect of increasing age reached statistical significance in men only (Table 5). The greatest decrement tended to occur among re-starters of smoking. This was true particularly among women, p = 0.004 (Table 5). The annual FEV₁ decline in non-smokers and persistent ex-smokers was similar.

Figure 2 Decline in FEV₁ during 10 years in men and women by smoking habits (never = never smoker, ex = ex-smoker, smoker = current smoker) and age at start of the observation period adjusted for family history of OLD and socio-economic class in a multiple linear regression model.

Table 5 The influence of age and smoking categories on the decline in FEV₁ during the observation period by logistic linear regression. Other independent variables in the model were family history of OLD, socioeconomic class and chronic bronchitis at start of the observation period

Variable	Men			Women
	Constant	SE	p-value	Constant	SE	p-value
Constant	0.467	0.041	< 0.001	0.267	0.027	< 0.001
Age^=	0.049	0.019	0.013	0.009	0.013	0.390
Re-starters^‡	0.080	0.116	0.492	0.182	0.063	0.004
Persistent smokers^‡	0.063	0.033	0.055	0.106	0.024	< 0.001

*express the 10 year decline of FEV₁ in liters.

⁼effect of 10 years ageing.

^‡effect compared with effect of non-and ex-smoker.

In the multivariate logistic regression model, smoking, male gender and higher age were significant risk factors for decrement ≥ 60 ml/year compared to < 30 ml/year. Gender-specific analyses were also performed in an analogous model, in which higher age was found to be a significant risk factor in men (born 1934–1935 and 1919–1920 adjusted ORs 2.38, CI 1.32–4.73 and 2.50, CI 1.32–4.73), while persistent smoking and restart of smoking were significant risk factors in women (adjusted ORs 3.74, CI 1.54–9.09 and 10.12, CI 1.99–51.43). Thus a similar pattern was found as in the linear regression model.

DISCUSSION

There are only a few reports on longitudinal data allowing estimates on incidence of COPD ([8], [9], [10], [11], [12]), and there are hardly any population-based data on decline in lung function in relation to development of COPD. The bases of the longitudinal data in this study were subjects reporting respiratory symptoms in an age stratified general population sample ([21]). The study population is thus considered to be representative of subjects with respiratory symptoms in the general population in current age groups. From a clinical point of view these subjects represent the population from which subjects seeking medical advice due to respiratory complaints are recruited.

Even though the results cannot be regarded as representative for the general population we do consider the results highly relevant in a clinical setting. The diagnosis of COPD was based on established spirometric criteria and the quality of spirometry was essential when evaluating both the incidence of COPD and the rate of decline in FEV₁. The rate of decline in FEV₁ was based on two lung function tests performed 10 years apart. More frequent performed lung function tests would have strengthened the results. However, the high participation rate, the use of the same equipment, and the standardised performance of the lung function tests supports the validity of the results.

Already at study start in 1986 we found an impaired lung function among smokers and subjects with chronic productive cough (Table 1), a symptom consistent with the clinical diagnosis of chronic bronchitis ([20]). The overall decrease in lung function was greater than the normal decline (Table 3 and Table 4), and the decline was most prominent in men, smokers and subjects reporting chronic productive cough at entry. However, the greater decline in FEV₁ expressed as mean ml/year among men was almost eliminated when the method of measurement included correction for age, gender and height as change in percent of predicted did. Weight changes, above all weight gain, could also be expected to affect the lung function. However, data on weight was not consistently collected in the examination in 1986, and therefore we cannot evaluate this possible confounder.

The benefit of smoking cessation was obvious in this symptomatic cohort. In accordance with other studies, the decline in FEV₁ was similar in ex-smokers and non-smokers ([3], [15], [17], [29]) and, as expected, smokers had a significantly greater decline in FEV₁ compared to non-smokers (Table 3 and Table 4). The decline was found to be even steeper in re-starters compared to persistent smokers as indicated in a previous report from the Tucson, AZ studies ([30]). According to an 11-year follow-up of the Lung Health Study intermittent quitters had a decline between that of quitters and persistent smokers ([18]). However, the study population of the Lung Health Study were middle-aged smokers with obstructive lung function impairment at entry thus not quite comparable to our study population, and their intermittent smokers were not equivalent to our re-starters, which make the results difficult to compare. Further, our results must be interpreted with caution due to the limited number of subjects classified as re-starters.

Subjects in the youngest age group (35–36 years at study start and 45–46 years at time for follow-up) had a lower decline than subjects aged 45–46 years at study start (55–56 years at follow-up). This conforms well to the known plateau phase with stable lung function among adults until the normal decline due to ageing starts after 40 years of age ([19]). The decline did not differ significantly between the oldest age group and the two younger, possibly due to a “healthy survivor” effect, and also to the gender-dependent different smoking habits analysed by age group. Our finding of an association between chronic productive cough and increased decline corroborates results from other studies ([11], [14], [16]). However, this association did not reach statistical significance in the multiple logistic regression model.

Findings by others have to some extent been contradictory with respect to the relationship between respiratory symptoms and increased decline in lung function, which is necessary for the development of COPD. Chronic productive cough was found to be associated with increased lung function decline in COPD ([16]) in one study, but the same group, the Copenhagen City Heart Study, has also reported respiratory symptoms in subjects with normal lung function (GOLD stage 0) not to predict the development of COPD ([11]). Further longitudinal studies of general populations samples are needed to increase the knowledge of the relationship between respiratory symptoms, decline in lung function and the development of COPD.

Current smoking and higher age were risk factors for increased decline in FEV₁ in the multivariate analyses. However, the gender-specific analyses revealed smoking to be a stronger risk factor in women compared to in men, while increasing age was a significant risk factor in men only. Women re-starting smoking had the steepest decline in FEV₁ of all sub-groups. There was a trend of a more pronounced benefit of total smoking abstinence or cessation among women, increasing their FEV₁ percent predicted over 10 years, while men continued to decline (Table 4). This may be related to airborne occupational exposure in men. The 11-year follow-up of the U.S. Lung Health Study ([18]), originally a cohort of smokers with airway obstruction, showed as expected a greater decline in FEV₁ compared to our results. Their findings of a greater benefit of smoking cessation in women compared to in men are in accordance with the results in the present study. Our study population confirmed further that women are catching up on male smoking behavior.

Differences in smoking habits probably cause most of the reported gender related differences in prevalence of COPD (men > women) found in the literature ([1], [30]), and may also explain why COPD is becoming more common in women ([32]). There are, however, studies suggesting that female smokers are more susceptible to tobacco smoke with a more rapid decline in lung function compared to male smokers ([15], [16], [33], [34]). The results of our gender-specific multivariate analyses support this hypothesis. We can thus expect an increase in COPD among women of greater magnitude than the changes in smoking habits may explain.

The use of different spirometric criteria for COPD may change the prevalence four-fold in the same population ([30], [34]). We consequently expected that different definitions of COPD would influence the incidence of COPD, first when defining the population at risk and secondly when identifying the incident cases. For comparison we used two different levels of disease severity in this study; COPD according to the ATS/ERS standards (FEV₁/FVC ≤0.70) and moderate COPD according to the ATS/ERS standards (FEV₁/FVC ≤0.70 and FEV₁ < 80% of predicted) which conforms with the former British Thoracic Society (BTS) criteria of COPD (35) and COPD stage II according to the Global Initiative for Chronic Obstructive Lung Disease ([36]).

Of all subjects developing COPD according to ATS/ERS nearly 50% (60 out of 127) developed moderate COPD and approximately 5% (6 out of 127) developed severe COPD. As expected, the decline in FEV₁ was related to incident COPD, and increased decline was associated with development of more severe COPD. For comparison, when moderate COPD was used to define the disease, we found consequently a higher number of incident cases (n = 83) with a more rapid decline in lung function compared to the 66 incident cases of moderate COPD originating from the 127 incident cases of COPD according to ATS/ERS. This difference reflects the effect of using different populations at risk, which is a very important factor to consider when comparing incidence and incident cases of COPD.

Another important issue when using the fixed ratio 0.70 for definition of airway obstruction is that COPD may be overestimated in elderly and underestimated among younger subjects. There is, among others, a report that the GOLD criteria will identify non-symptomatic elderly without a smoking history ([37]). However, in our study the study sample consisted of symptomatic subjects. Thus we can expect that the risk of over diagnosis is less prominent than in a general population sample.

The spirometric criteria of GOLD stage ≥II ([36]) (similar to the BTS criteria ([35]) and also to moderate COPD according to ATS/ERS ([22])) has been suggested to be a practical threshold to identify symptomatic subjects with COPD ([38]). Consequently, close to a half of the incident cases of COPD in this symptomatic cohort (60/127) developed COPD of possible clinical relevance during the 10-year observation period (or 83 cases if moderate COPD is used for definition of COPD).

In prevalence studies of COPD it is impossible to distinguish between the presences of former lung function impairment due to reduced childhood growth, premature decline and accelerated decline in lung function. Analyses of incident COPD provide the opportunity to study populations with normal or close to normal lung function that during the observation period have a sufficient decline in lung function to develop COPD. Almost every fourth incident case had a decline > 90 ml/year corresponding to a change in percent predicted FEV₁ of −27.8 over 10 years when moderate COPD according to ATS/ERS standards was used to define COPD. No similar results have been published to date, as there are very few reports on incidence of COPD in the literature over the last 25 years ([8], [9], [10], [11], [12]).

In the analyses asthma has not been excluded as a cause of obstructive lung function impairment. Most guidelines agree on that asthma developing chronic obstructive lung function impairment is almost impossible to distinguish from COPD. Some of the guidelines include chronic asthma as a cause of COPD. Further, asthma and COPD may co-exist in the same subject. According to earlier OLIN-studies the prevalence of co-existent asthma and chronic bronchitis/chronic obstructive lung function impairment was < 1%, and approximately 1–2% in higher ages ([39]). Approximately 30% of the study population reported that they had or had had asthma. We did not analyze this subpopulation separately. However, symptoms commonly associated with asthma, as wheeze and attacks of shortness of breath were not associated with a higher decline in lung function.

We conclude that increased decline in FEV₁ is associated with smoking, higher age and chronic productive cough among subjects with respiratory symptoms but the risk factor pattern is gender-dependent, and women seems to be more vulnerable to smoking than men. A considerable proportion of the incident cases of COPD have a rapid decline in lung function. Further, an adjusted method of measuring decline in FEV₁ is recommended to avoid influence by mainly gender but also age and height. Units change in percent of predicted normal value of FEV₁ is an easy method, well suited for analysis in daily clinical practice. However, the predictive value for development of COPD and disease progress needs to be further evaluated.

APPENDIX

Decline in FEV₁ during the 10 years can be calculated from the linear logistic model in men and women using age, smoking categories, socio-economic status and family history of OLD as discrete variables. Three age groups were included in the study (coded 1, 2, and 3). By defining the variable age as (36.5-age)/15+3 in the model calculations for all ages from age 36–37 (36.5) to age 66–67 years can be calculated. Thus the outcome by age in the different smoking categories can be estimated. The smoking categories re-starters and smokers were compared to a reference group consisting of non-smokers, ex-smokers and intermittent smokers.

Table A1 Decline in FEV₁ (litre) during the 10-year period, by age at start of follow-up and smoking category

	Men			Women
Age at start	Re-starter*	P smoker^=	Reference^‡	Re-starter	P smoker^=	Reference^‡
35	0.402	0.386	0.323	0.421	0.346	0.240
40	0.418	0.401	0.338	0.424	0.348	0.243
45	0.433	0.417	0.354	0.427	0.351	0.246
50	0.449	0.433	0.369	0.430	0.354	0.248
55	0.465	0.448	0.385	0.433	0.357	0.251
60	0.480	0.464	0.400	0.436	0.360	0.254
65	0.496	0.479	0.416	0.439	0.363	0.257

*Re-start of smoking.

⁼Persistent smoking.

^‡Reference group including non-smoking, ex-smoking and intermittent smoking.

The difference between re-starters and persistent smokers did not reach statistical significance. The difference was small among men, while it was more obvious among women but the study was probably under-powered in this aspect. The difference between all smoking categories was not significant in men, though it was close to between persistent smokers and the reference group. In women both the difference between persistent smokers and the reference group and between re-starters and the reference group were significant (p < 0.001 and p = 0.040, respectively). The outcomes are illustrated by the following table.

The authors thank Anders Oden for statistical advice and support and the OLIN study staff for help with collecting the data. The study was approved by the Regional Committee for Ethics at Umeå University.

Funding from: Swedish Heart-Lung Foundation and Local Health Care Authority, Norrbotten