Prediction of COPD and Related Events Improves by Combining Spirometry and the Single Breath Nitrogen Test

Abstract

Chronic obstructive pulmonary disease (COPD) develops in small airways. Severity of small airway pathology relates to progression and mortality. The present study evaluated the prediction of COPD of a validated test for small airway disease, i.e. a slope of the alveolar plateau of the single breath nitrogen test (N₂-slope). The N₂-slope, spirometry, age, smoking habits, and anthropometric variables at baseline were obtained in a population-based sample (n = 592). The cohort was followed for first COPD events (first hospital admission of COPD or related conditions or death from COPD) during 38 years. During follow-up, 52 subjects (8.8%) had a first COPD event, of which 18 (3.0%) died with a first COPD diagnosis. In the proportional hazard regression analysis adjusted for age and smoking habits, the cumulative COPD event incidence increased from 5% among those with high forced expired volume in one second (FEV₁) to 25% among those with low FEV₁, while increasing from 4% among those with the lowest N₂-slope to 26% among those with the highest. However, combining the N₂-slope and FEV₁ resulted in considerable synergy in the prediction of first COPD event and even more so when taking account of smoking habits. The cumulative COPD event incidence rate was 75% among heavy smokers with the highest N₂-slope and lowest FEV₁, and less than 1% among never smokers with the lowest N₂-slope and highest FEV₁. Thus, combining the results of the single breath N₂-slope and FEV₁ considerably improved the prediction of COPD events as compared to either test alone.

Keywords:

• Epidemiology
• longitudinal study
• COPD event
• respiratory functions tests

Introduction

Careful morphological studies by Hogg et al. (1) have shown that chronic obstructive pulmonary disease (COPD) develops in small airways, notably those with an internal diameter of 2 mm or less. Furthermore, COPD progression and mortality have been shown to be related to pathology of the small airways (2, 3).

According to the Global Initiative for Chronic Obstructive Lung Disease (GOLD), spirometry is required to make the diagnosis of COPD (4). However, the forced expired volume in one second (FEV₁) is rather insensitive to small airway dysfunction (1), indicating that spirometry may not diagnose COPD at an early stage. Nevertheless, low FEV₁ within the normal range, may represent a pathologic decline and therefore predicts development of COPD, as exemplified by GOLD stage 1 (4, 5).

The slope of phase three of the single breath nitrogen test (the N₂-slope) relates to small airway abnormalities/COPD as shown by Cosio et al. (6), who compared structural changes in small airways with the N₂-slope as obtained prior to open lung biopsy or lobectomy. The N₂-slope was more reliable than spirometry in detecting early lesions in small airways. In smokers, the N₂-slope is abnormal more frequently than FEV₁ (7) and predicts an increased decline of FEV₁ (8), indicating that the N₂-slope may predict the development of COPD. Furthermore, there is an association between an abnormal N₂-slope and an increased risk of hospital admission owing to COPD according to a 9-year follow up by Vestbo et al. (9).

Thus, both low FEV₁ within the normal range and a steep N₂-slope predict the development of COPD, but there is no knowledge of the predictive ability of combining FEV₁ and a steep N₂-slope. In the present investigation, we analysed the prediction of first COPD events by the N₂-slope and FEV₁ in a systematic population sample of 50- and 60-years old men followed up for 38 years. The aim was to test the hypothesis that the combination of FEV₁ and the N₂-slope has complimentary predictive power for future COPD.

Methods

Study design

A systematic sample of middle-aged men with a follow-up period of 38 years formed the basis of this study. Oral informed consent, standard procedure at the time, was obtained from all study subjects. The study was approved on several occasions, first by the Research Ethics Committees in Gothenburg and Uppsala, Sweden, and later by the National Research Ethics Board.

Study population

In 1973, two systematic samples were drawn from the national population register. The first sample consisted of all men born in 1913 on a date divisible by 3 (i.e. day 3, 6, 9, and so on of each month) and living in the city of Gothenburg. Of the 945 men fulfilling the criteria, 787 (83%) agreed to participate in the study, which was part of the Study of Men Born in 1913 and 1923. Those of the men born on a date of a month divisible by 6 (n = 387) were included in the lung function part of the study. The second sample consisted of all men born in 1923 on day 3, 15 and 27 of each month, and living in the city of Gothenburg. A total of 292 men met these criteria and 220 men (75%) participated in the study (7).

The lung function part of the database from 1973 (7) was not fully accessible due to outdated storage of data. Preserved original hard copy data from 1973 were, however, available, and a new database was established, and the quality of data was carefully examined. This new database contains data from 595 men, of whom 212 were 50 years old and 383 were 60 years old in 1973 (10). Data from 12 men could not be retrieved. The sampling process of the subjects included in the present study considering the inclusion criteria is presented in Figure 1. The present study population contains data from 592 men, of whom 210 were 50 years old, and 382 were 60 years old in 1973.

Figure 1. Flow-chart of the study population.

Smoking habits data obtained in 1973 were classified in three different ways. First, individuals were categorised as never-smokers, ex-smokers, or smokers. Second, a smoking habit score ranging from 1 to 5 was assigned: (1) never smoked, (2) ex-smoker since more than 6 months, (3) currently smoking 1–14 g/day, (4) smoking 15–24 g/day, or (5) smoking more than 24 g/day, one cigarette equalling 1 g, one cheroot 2 g, and one cigar 5 g. For pipe-smokers, the amount smoked was obtained by dividing the traditional 50 g tobacco pack by the number of days it lasted. Finally, pack years were calculated as the product of smoking intensity (i.e. cigarettes or grams of tobacco per day) and duration (years).

Lung function tests

All subjects were examined with spirometry and a single breath nitrogen (N₂) test, lung function measurements being performed by qualified technicians. Details of spirometry measurements, the single breath N₂ test, equipment, and procedures have been published previously (7). Briefly, each subject performed two satisfactory slow vital capacity (VC) manoeuvers and three satisfactory forced vital capacity (FVC) manoeuvers. The largest VC of the slow and forced manoeuvers and the largest FEV₁ were used in the analyses. VC, FEV₁, and FEV₁/VC were expressed as a percent of predicted normal according to Hankinson et al. (11), thus disregarding any differences between FVC and VC.

The N₂-slope was calculated as the increase in the nitrogen concentration from the point where 825 ml (body temperature and pressure, saturated) had been expired from total lung capacity (TLC) until the beginning of phase IV (closing point), divided by the corresponding expired volume. The N₂-slope was also expressed as a percent of predicted normal (% predicted) according to Sixt et al. (12). In each subject, two satisfactory tracings were attempted, and the mean of these tracings was used. All tracings were coded and examined by an experienced investigator who was unaware of the characteristics of the subjects. Closing volume (%VC) and closing capacity (%TLC) were also calculated.

Outcome measure

The outcome measure of this study was a first COPD event, measured either as the first hospital admission with a diagnosis of COPD or related diagnoses, or death from COPD for subjects with no previous hospital admission for the disease until 31 December 2011. Data on all hospital admissions in the study population from baseline and onwards were obtained from the National Hospital Discharge Register, and data on all deaths were obtained from the National Cause of Death Register. These registers contain all hospital admissions and all mortality events for Swedish residents, whether citizens or not, and are maintained by the National Board of Health and Welfare.

Discharge diagnoses and causes of death were coded according to the International Classification of Diseases (ICD), version 8 from 1968 through 1986, version 9 from 1987 through 1996, and version 10 from 1997 and onwards. First COPD and related event diagnoses were defined and identified by ICD-8 codes 490 (bronchitis, unspecified as to acute or chronic), 491 (chronic bronchitis), and 492 (emphysema); ICD-9 codes 490, 491, 492, and 496 (chronic obstructive lung disease); and ICD-10 codes J40 (bronchitis, unspecified as to acute or chronic), J41 (simple and mucopurulent chronic bronchitis), J42 (unspecified chronic bronchitis), J43 (emphysema), and J44 (COPD).

In the hospital discharge data, the main diagnosis and eight possible additional diagnoses were scrutinised. Respiratory diagnoses, such as asthma, pneumonia, or lung cancer, were not considered as related diagnoses to COPD or causes of death and therefore not included in the outcome criteria. In the mortality data, the underlying cause of death and 20 possible additional diagnoses were analysed. Three subjects had a hospital discharge involving a COPD related diagnosis before baseline and were therefore excluded (Figure 1).

Statistical considerations

Data were analysed using the SAS Software version 9.3 (SAS Institute, Inc., Cary, NC, USA). Simple differences between groups were tested with the Mann–Whitney U test for continuous data and with the chi-square test for categorical data.

During the planning of the study, no power analysis was performed. However, a post hoc analysis showed a need of 80 subjects for a power of 80% to identify factors associated with the outcome variables. The power with the actual sample size was 99.99% using all diagnoses in the discharge diagnoses and cause of death diagnoses. Using only the first discharge diagnosis and the underlying cause of death gave a need of 100 subjects for 80% power and 99.99% with the actual sample.

Smoking was measured in three ways. In multivariate proportional hazard regression analyses of the three measures of smoking on the first COPD event, the smoking habit score had the best impact on outcome: Wald’s chi-square 23.8 as compared with 14.0 for never-, ex-, or current smokers, and 21.4 for pack years. Similarly, there were three relevant spirometry measures including FEV₁ of which FEV₁ (% predicted) had the best impact on outcome: Wald’s chi-square 38.0, as compared to 29.2 for FEV₁/VC, and 32.7 for FEV₁/VC (% predicted). Finally, two measures of the N₂-slope were available, of which the N₂-slope (% predicted) had the best impact on outcome: Wald’s chi-square 38.6 as compared to 37.5 for the N₂-slope (%N₂/L). Based on these circumstances, smoking habit score, FEV₁ (% predicted), and N₂-slope (% predicted) were used in the subsequent analyses. To facilitate the interpretation of the effects of FEV₁ (% predicted) and N₂-slope (% predicted), data were subdivided into quintiles.

In the analyses of the effects of exposure on outcome, follow-up time was measured as a number of days from the baseline to outcome or end of follow-up. Hazard rates (HR) for FEV₁ (% predicted) and N₂-slope (% predicted) on a first COPD event were computed with the SAS “Lifetest” procedure. The hazards of the groups with FEV₁ (% predicted) and the N₂-slope (% predicted) above or below mean value, respectively, were approximately proportional across time, allowing proportional hazard regression analysis.

The FEV₁ (% predicted) and the N₂-slope (% predicted) measures were closely correlated (r=0.66), indicating that they might be the measures of the same underlying process and thereby cause collinearity. Possible collinearity was measured with the SAS “Reg” procedure, and none was found.

The influences of age, smoking habit score, FEV₁ (% predicted) quintiles, and N₂-slope (% predicted) quintiles on a first COPD event were tested with proportional hazard regression analysis. Censoring events were suffering a first COPD event, death from COPD with no previous event, death from other causes, or no event at end of follow up, whichever came first, thereby adjusting for competing risk. Additive synergy was measured by an additive term of FEV₁ (% predicted) quintiles and N₂-slope (% predicted) quintiles, and multiplicative synergy by a multiplicative term of the quintiles. The risk diagrams and the nomograms were compiled based on data from the proportional hazard regression model. All analyses were two-tailed, and p < 0.05 was set as the significance level.

Results

Characteristics at baseline

Mean age was 56.7 years, mean height was 175 cm, and mean weight was 78 kg. These anthropometric data did not differ between subjects with and with no COPD event. During the 38-year follow-up, 52 subjects (8.8%) had a first COPD event, of which 18 (3.0%) died from a first COPD event. Tobacco smoking was more common among subjects with a COPD event than among those without a COPD event. FEV₁ (% predicted) was significantly lower and N₂-slope (% predicted) significantly higher in COPD event subjects than in subjects without a COPD event (Table 1).

Table 1. Study population characteristics at baseline classified according to the presence or absence of first COPD event during follow-up.

	No COPD event (n = 540)		COPD event (n = 52)
	n	Mean (SD)	n	Mean (SD)	p
Age at baseline, years		56.8 (4.8)		56.3 (4.9)	n.s.
FEV1 (% of predicted normal)^a		94.5 (15.7)		76.8 (21.0)	<0.0001
Quintile 1 < 79.8	89	69.3	27	60.9
Quintile 2 ≥ 79.8 to <90.6	109	85.4	10	84.9
Quintile 3 ≥ 90.6 to <98.9	111	94.8	8	94.3
Quintile 4 ≥ 98.9 to <106.7	115	102.2	4	102.2
Quintile 5 ≥ 106.7	116	114.3	3	112.6
N2-slope (% of predicted normal)^a		152.8 (108.7)		254.3 (154.6)	<0.0001
Quintile 1 < 80.5	118	59.4	1	7.6
Quintile 2 ≥ 80.5 to <114.0	115	99.0	4	101.5
Quintile 3 ≥ 114.0 to <149.0	112	130.7	7	129.4
Quintile 4 ≥ 149.0 to <220.2	102	178.5	16	177.0
Quintile 5 ≥ 220.2	93	336.3	24	378.1
Smoking habit score (1–5)^b		2.9		3.6	<0.001
Never smoked (%)	115	21.3	4	7.7
Ex-smoker (%)	178	33.0	12	23.1
Smoking 1–14 g/day (%)	155	28.7	19	36.5
Smokers 15–24 g/day (%)	70	13	12	23.1
Smokers ≥25 g/day (%)	22	4.0	5	9.6

^a Predicted normal values according to Hankinson (spirometry) (11) and Sixt (N₂ test) (12).

^b Smoking amount classified as grams (g) of tobacco consumed per day from cigarettes (1 cigarette = 1 g of tobacco), cheroots (1 cheroot = 2 g of tobacco), cigars (1 cigar = 5 g of tobacco), and pipe tobacco (50 g package divided by number of days it lasted).

The total number of observation years was 13,058. The hazards rate, measuring the percentage of subjects with a COPD event per time unit, increased for both FEV₁ (% predicted) and N₂-slope (% predicted) across the total follow-up time. This indicates that the two risk factors were active throughout life in the study population.

Table 2 shows the result of a proportional hazard regression analysis of age, smoking habit score, FEV₁ (% predicted) quintiles, and N₂-slope (% predicted) quintiles on outcome. After accounting for the effect of age and smoking habit score, N₂-slope (% predicted) quintiles and FEV₁ (% predicted) quintiles were significantly and independently related to the incidence of COPD events.

Table 2. Results of the proportional hazard regression analysis of exposures on the first COPD event during follow-up.

	Parameter estimate (SD)	Wald’s chi-square	HR	95%CI	p
N2-slope (% pred.) quintiles	0.48682 (0.15)	10.1	1.63	1.21–2.20	<0.002
FEV1 (% pred.) quintiles	−0.44011 (0.14)	9.40	0.64	0.49–0.85	<0.005
Smoking habit score	0.25598 (0.14)	3.53	1.29	0.99–1.69	0.06
Age	0.02057 (0.03)	0.47	1.02	0.96–1.08	0.49

Abbreviations: HR, hazards ratio; p, probability; SD, standard deviation; 95%CI, 95% confidence interval.

Closing volume (%VC) quintiles and closing capacity (%TLC) quintiles were in proportional hazard regression analysis not significantly related to the incidence of COPD events after accounting for age, smoking habit score, and FEV₁ (% predicted) quintiles.

For FEV₁, the cumulative incidence of COPD events increased from 5% in quintile 5 to 25% in quintile 1, and for the N₂-slope, the cumulative incidence increased from 4% in quintile 1 to 26% in quintile 5, after adjusting for age and smoking habit score. The two measures appear to have a similar effect on outcome.

In Figure 2, the cumulative COPD event incidence after 38 years of follow-up is shown in the relationship to FEV₁ (% predicted) and N₂-slope (% predicted) quintiles. The term measuring additive synergy and the term measuring multiplicative synergy were significantly related to outcome (p < 0.0001). However, the term measuring multiplicative synergy had a higher Wald’s chi-square (impact on outcome). After 38 years of follow-up, the cumulative COPD event incidence increased from 1.8% with the combination of FEV₁ (% predicted) quintile 5 and N₂-slope (% predicted) quintile 1, to 52.3% with the combination of FEV₁ (% predicted) quintile 1 and N₂-slope (% predicted) quintile 5, thus far higher incidence than each measure per se. The progressive curvilinear increase of the cumulative incidence is illustrated in Figure 2, reflecting the synergy.

Figure 2. COPD event incidence rate during 38 years of follow-up in men classified according to FEV₁ (% predicted) quintiles and N₂-slope (% predicted) quintiles and adjusted for the influence of age and smoking habit score. The progressive curvilinear increase of the incidence rate seen in the model is a reflexion of the synergy between FEV₁ and the N₂-slope.

Taking account also of smoking habits further increased the predictive power. In Figure 3, three nomograms show the probability of suffering a first COPD event during 10, 20, and 38 years of follow-up based on FEV₁ (% predicted) quintiles, N₂-slope (% predicted) quintiles, and smoking habits score. For those who never smoked and had the lowest N₂-slope and the highest FEV₁ values, the probability of suffering a COPD event within 10 years was <1%. For heavy smokers with the highest N₂-slope and lowest FEV₁ values, the corresponding probability was 18%. After 38 years of follow-up, the corresponding probabilities were 1% and 75%.

Figure 3. Nomograms showing the probability of suffering a first COPD event during 10 years (a), 20 years (b) and 38 years (c) of follow-up in the relationship to FEV₁ quintiles, N₂-slope quintiles, and smoking habits for middle-aged men. To obtain the probability of suffering a COPD event, a straight line should connect the patient’s N₂-slope (% predicted) level and FEV₁ (% predicted) level. At the crossing between this line and the bold line among the smoking habit groups in the nomogram, go horizontally to the smoking habit relevant for the patient. The probability of suffering a COPD event is given in the relevant circle.

Discussion

The present study demonstrates that in combination, the single breath N₂-slope, FEV₁ and smoking history considerably improve the prediction of future COPD events compared to any of the measures separately. The N₂-slope improves the prediction when FEV₁ is within normal limits and when FEV₁ is reduced (Figure 2).

The N₂-slope is increased when spirometry remains normal in early COPD and among smokers (6, 7). Thus, the N₂-slope is more sensitive to the early pathology of COPD when FEV₁ remains relatively unaffected. In addition, the N₂-slope predicts an increased decline of FEV₁ in a 7 years follow-up of a population-based sample of 460 men (8). However, no such relationship was found by Buist et al. in a 9–11 years follow-up of 734 subjects in two cohorts of which one was non-population based (13). Closing capacity was associated with the FEV₁ decline in that study. Stanescu et al. studied 105 smokers during a 6 years follow-up (14) and 56 of them during a 13 years follow-up (15) and found that subjects with a low initial FEV₁/VC ratio and a steep N₂-slope developed a lung function profile consistent with COPD. Different selections of study populations and differences in methodology, including different algorithms for determination of the N₂-slope, presumably explain the lack of consistency between the above-cited studies.

To the best of our knowledge, only Vestbo et al. (9) have previously analysed the relationship between the N₂-slope and occurrence of COPD events. These authors studied a random population sample of 876 men, aged 46–69 years followed up for 9 years. The N₂-slope, but not closing volume or closing capacity, was associated with hospital admission due to COPD after accounting for age, smoking, and FEV₁. The authors considered, however, the predictive effect of the N₂-slope to be clinically insignificant in addition to FEV₁ because of a considerable over-lap between the results of the N₂ test for men admitted to hospital and men who were not (9). Unfortunately, there was no corresponding analysis of the overlapping regarding FEV₁. Furthermore, the multiple logistic regression analysis used by Vestbo et al. may not have been as powerful as the Cox regression analysis used in the present investigation. The effect of combining the results of the N₂-slope and FEV₁ was not analysed. Notwithstanding the differences between the Vestbo study and the present, they agree that the N₂-slope is an independent predictor of COPD events, whereas closing volume and closing capacity are not, which is in agreement with the results of the present study.

Low spirometry results are associated with an increased risk of COPD morbidity (16, 17) and exacerbations (5, 18–21). An important reason why low spirometry values, notably low FEV₁, have a predictive value is that a low value may represent a pathologic decline from a higher normal value and therefore indicates COPD, provided other diseases are excluded.

The importance of obstruction of the small airways in COPD was shown by Hogg et al., already in 1968 (1). A more detailed description of the morphologic changes in COPD has emerged in recent years. The airways obstruction in COPD is primarily due to obliteration and disappearance of small airways that potentially precede the development of emphysema, according to a recent study by McDonough et al. (22). The reduction in the total cross-sectional area of terminal bronchioles was 81–99.7% in severe COPD! (22).

Although FEV₁ is insensitive to obstruction of small airways, a pronounced reduction in the total cross-sectional area of terminal bronchioles necessarily reduces FEV₁. The predictive effect of reduced FEV₁ or increased decline of FEV₁ presumably relates mainly to the reduction in a number of small airways, rather than to effects of emphysema. On the other hand, Boeck et al. measured, among other things, the N₂-slope and a double tracer gas single breath washout test among 65 patients with moderate to severe COPD (23). In 35 of these patients, a computed tomography scan was obtained at TLC, and the area of low density (<950 Hounsfield units) was calculated. The N₂-slope correlated significantly with the area of low density but became insignificant in a multilinear regression, whereas FEV₁ remained significant, possibly indicating that FEV₁ is more sensitive to emphysema than the N₂-slope. However, the material was rather small, the statistical power was limited, and the low-density area may not accurately have assessed the degree and inhomogeneity of emphysema (24).

The N₂-slope is a measure of the degree of ventilation inhomogeneity (25). The inhomogeneity has at least two different components: (1) uneven distribution of airway resistance of large and/or small airways and (2) uneven lung elastic properties (26). Uneven elastic properties almost certainly cause heterogeneous distribution of regional and intraregional RV/TLC ratios, obtained after the VC inspiration of oxygen preceding the expired N₂-slope. The more heterogeneous the residual volume/total lung capacity (RV/TLC) ratios are, the steeper the slope, other factors equal. Furthermore, unequal lung elastic properties may increase the asynchronous expiration from TLC, required to disclose the effect of heterogeneous RV/TLC ratios on the N₂-slope. The improved prediction of COPD events by the N₂-slope in addition to FEV₁ may be explained by the generally higher sensitivity of the N₂-slope compared to FEV₁ to disclose COPD related processes, causing uneven distribution of airway resistance of small and larger airways and uneven lung elastic properties, as in emphysema.

Cough symptoms are associated to a steep N₂-slope also after adjusting for spirometry and smoking history (27). Presumably, alterations in small airways trigger some cough symptoms. Thus, the link between a cardinal symptom of bronchitis and a steep N₂-slope adds strength to the importance of the single breath N₂ test.

The multiple breath nitrogen washout test offers the possibility of separating acinar (Sacin) and conductive airways (Scond) inhomogeneity (28) and would probably have contributed to further understanding of the mechanisms involved. The high sensitivity of these indices to smoking induced airway abnormalities as shown by Jetmalani et al. (29) is similar to the high sensitivity of the N₂-slope as shown by Oxhöj et al (7). At the time of the present study, the sophisticated analysis of the multiple breath nitrogen washout was not developed. Interestingly, Boeck et al. (23) found that the N₂-slope and the double tracer gas single breath washout test were uncorrelated, indicating that the double tracer gas test also adds information on the physiological aspects of COPD.

Data on COPD events in the present study derive from a national registration database of diagnosis codes. As the diagnosis code of COPD was not in practice before ICD10, the definition of COPD events included related diagnoses, such as chronic bronchitis and emphysema, to cover COPD outcomes before this time. This may have caused so called potential independent misclassification of outcome, e.g. bronchitis or chronic bronchitis, which may have had various degrees of airways obstruction. Nevertheless, the 52 subjects with a COPD event had a lower FEV₁ and a steeper N₂-slope compared to the no COPD event group (Table 1), indicating that COPD and related diagnoses were generally accurate. The subjects with COPD event in the present study had either COPD or related diagnoses as one of the discharge diagnoses or COPD as one of the diagnoses of the death certificate. The present results were, however, confirmed when the subset of subjects with the principal diagnosis of COPD or related diagnoses, or COPD as the main cause of death was analysed (data not shown). The related diagnoses did not include diagnoses such as asthma, pneumonia, or lung cancer. Notably, cardiac disease could be a confounding factor, e.g. heart failure may mimic COPD exacerbations in some cases (30). A relationship between impaired FEV₁ and heart failure has also been shown (31, 32). In some of these studies, a low FEV₁ was due to COPD (33). In chronic heart failure, increased ventilation heterogeneity has furthermore been shown (34). Conceivably, a steep N₂-slope may be a predictor of cardiovascular disease. Since COPD or related diagnoses were not always the first diagnosis, the importance of COPD and related diagnoses for hospital admission may have varied.

Furthermore, information on COPD and related diagnoses in primary health care was not available for the analysis. Neither did we collect data on the development of airflow limitation over time. Another potential limitation is the relatively small study sample and the consequently limited number of events. However, the post hoc power analysis justifies the sample size and study design. One obvious limitation of this study is that it included men only. Conclusions are therefore restricted to men. Strengths of this study include the systematic population sample, that the investigators measuring the N₂-slope were blinded to subject information and that the subjects were followed for almost 40-years.

In conclusion, the N₂-slope is an independent predictor of COPD events over time, considerably improving the prediction based on FEV₁ and smoking history. The N₂ test should be taken into account when prediction of COPD is relevant.

Conflict of interest statement

No actual or potential conflicts of interest exist between any of the authors (Jan Olofson, Björn Bake, Bengt Bergman, Anders Ullman, or Kurt Svärdsudd) and organisations with financial interest in the subject matter.

Author contributions

Björn Bake and Kurt Svärdsudd are the guarantors for the content of the manuscript, including data and analysis. Jan Olofson, Björn Bake, Bengt Bergman, Anders Ullman, and Kurt Svärdsudd designed and performed study, analysed data, and wrote paper.

Funding sources

The study was financed by grants from several sources. During a long period, the Swedish Medical Research Foundation funded the study. Research grants were also received from the Swedish Heart and Lung Foundation. In recent years, funding has been received from the Swedish state under the agreement between the Swedish government and the county councils relating to the economic support of research and education under the ALF agreement (ALFGBG-721351).