Background Many studies show a link between forced expiratory volume in 1 s (FEV1) and survival in the general population and this has been interpreted as a link between airway obstruction and survival. However, the observation that vital capacity is also associated with survival weakens this interpretation.
Methods Data on spirometry and survival were taken from the Atherosclerosis Risk in Communities (ARIC) limited access dataset. Survival among 7489 participants with usable spirometry and complete data was regressed against measures of ventilatory function after controlling for many other factors likely to be associated with survival.
Results Survival was strongly associated with forced vital capacity (FVC) after adjustment for FEV1, but not the other way round. The fully adjusted hazard ratio (HR) associated with high FVC was 0.90 in men (95% CI 0.80 to 1.00; p=0.049) and 0.82 in women (95% CI 0.70 to 0.95; p=0.01). This compares with 0.98 for FEV1 in men (95% CI 0.90 to 1.07; p.0.72) and 1.01 in women (95% CI 0.89 to 1.15; p=0.84). There was no association between survival and airway obstruction as measured by the FEV1/FVC ratio.
Conclusions FVC but not airway obstruction predicts survival in asymptomatic adults without chronic respiratory diagnoses or persistent respiratory symptoms. The association is not explained by age, anthropometry, smoking, income occupation or blood pressure. As FVC later in life, cardiovascular risk, type II diabetes mellitus and low-grade systemic inflammation are all associated with poor fetal growth, these other conditions may be partly responsible for the poor survival in those with low FVC.
- pulmonary disease
- chronic obstructive
- COPD epidemiology
- respiratory measurement
- systemic disease and lungs
Statistics from Altmetric.com
- pulmonary disease
- chronic obstructive
- COPD epidemiology
- respiratory measurement
- systemic disease and lungs
Over the last 40 years many studies have demonstrated a relation between survival and forced expiratory volume in 1 s (FEV1) in general population samples,1 2 and have shown that low FEV1 is associated with high mortality from cardiovascular causes.3 4 This has been interpreted as evidence for an association between chronic obstructive lung disease (COPD) and atherosclerosis.5 FEV1, however, does not define COPD and an association has also been identified between mortality and vital capacity.3 Very few studies have specifically investigated which component of ventilatory function best predicts survival, and the few studies that have looked at the prognosis associated with a lower ratio of FEV1 to forced vital capacity (FVC) have given inconclusive answers. Anderson showed an association between the FEV1/FVC ratio and survival in New Guinea,6 but neither Friedman3 in his original study nor Lange in the Copenhagen Heart Study4 showed a convincing association with cardiovascular end points.
In this analysis we have used data from the Atherosclerosis Risk in Communities (ARIC) limited access dataset to separate out the effects of different spirometric indices on survival.
Participants in the ARIC study were drawn from the general population of four American communities: Forsyth County, North Carolina; Jackson, Mississippi; the north-western suburbs of Minneapolis, Minnesota; and Washington County, Maryland. In Jackson, only black subjects were included in the sample. Four thousand participants aged 45–64 years were to be selected from each centre.7
The protocol was ambitious and mainly aimed at identifying potential risk factors for cardiovascular disease. Importantly for this analysis, FEV1 and FVC were also measured using a volume displacement water-sealed spirometer.7 From a minimum of five forced expiratory manoeuvres, at least three acceptable blows were required and the best of these selected according to the contemporary ATS guidance.8 9 Quality control was managed from a central monitoring service at Johns Hopkins University.7 Subjects were followed up over the following years and, in particular, any deaths were noted.
The limited access dataset includes those participants in the ARIC cohort who gave informed consent for their data to be used in this way, and in some instances has small subgroups recoded to preserve anonymity. We analysed only individuals who were asymptomatic at baseline—that is, excluding those reporting persistent cough or phlegm, wheeze, breathlessness or doctor-diagnosed asthma, chronic bronchitis or emphysema. Except where otherwise indicated, we analysed only those with usable spirometry (at least two reproducible manoeuvres out of at least three judged to be acceptable).
In those with complete data, we modelled the effects on overall mortality of baseline characteristics observed at entry to the study. These included age, height, body mass index (BMI), waist-hip ratio, sitting height, income category, current working status, most recent occupation, ever smoked, current smoking, pack-years of smoking, education level and systolic blood pressure. Participants for whom the most recent occupation was missing were not distinguished from those with ‘other’ occupations in the limited access dataset, and the relatively high proportion of people with unreported income were analysed as a separate category. We analysed those participants with complete data on the remaining predictors and outcome. We compared characteristics of asymptomatic individuals with and without complete data using a Mann–Whitney test (age) and χ2 tests (sex and smoking status).
Mortality was analysed using Cox regression, looking at men and women separately. We looked at main effects of age, height, BMI, waist-hip ratio, sitting height, income category, current working status, most recent occupation, ever smoked, current smoking, pack-years of smoking, education category, systolic blood pressure and ethnic group (divided into black and non-black subjects in the limited access dataset). We also looked at what happened if we included quadratic effects of the continuous variables age, height, BMI, waist-hip ratio, sitting height, pack-years and systolic blood pressure. We concluded that linear effects were reasonable for all the continuously-scaled variables except for systolic blood pressure, which we divided into quintiles (calculated separately for men and women).
Having obtained in this way a model for the effects of confounding variables on mortality, we considered the additional effects of log(FEV1), log(FVC) and log(FEV1/FVC ratio). Taking logarithms of lung function measurements made conceptual sense because the three log values are linearly related, and also produced better-fitting models.
We also conducted a sensitivity analysis using all asymptomatic subjects with a recorded assessment of lung function regardless of quality.
All analyses were done with Stata 11.
The study began in 1986 and recruited 15 792 participants, 15 732 of whom gave permission for their data to be included in the ARIC limited access dataset;11 106 were asymptomatic, 7631 had usable spirometry and 7489 had complete data. There were 792 deaths recorded over 102 576 person-years of follow-up (figure 1).
There were small but statistically significant differences in the characteristics of asymptomatic individuals with and without complete data. Of those with complete data, 3982 (53%) were women compared with 2045 (57%) of those without complete data (p=0.001). The median interquartile range (IQR) age of those with complete data was 53 (49–58) compared with 54 (49–59) for those without complete data (p<0.001). Of those with complete data, 1478 (20%) were current smokers and 2635 (35%) were ex-smokers, compared with 826 (23%) and 1080 (30%) of those without complete data (p<0.001).
The characteristics of asymptomatic individuals with usable spirometry and complete data are shown in table 1. The association of these factors (excluding lung function) with mortality is given in table 2. Table 3 shows that both FEV1 and FVC were strongly associated with mortality (expressed as a hazard ratio (HR)) even after adjusting for the identified confounders, although there was no association between the FEV1/FVC ratio and survival for either sex. When further mutual adjustment was made for FEV1 and FVC, only FVC remained as an important predictor of mortality in the model, a higher FVC being associated with a lower HR: 0.90 for men (95% CI 0.81 to 1.00; p=0.049) and 0.82 for women (95% CI 0.70 to 0.95; p=0.01). FEV1 had no association with survival independent of its association with FVC.
Among 2988 asymptomatic black people, 1418 (47%) had usable spirometry while, among 8118 asymptomatic non-black people, 6213 (77%) had usable spirometry (p<0.001). However, when we repeated the analyses in table 3 using all individuals with spirometry data regardless of quality, the results were qualitatively the same.
In this general population cohort from the USA, excluding participants with chronic respiratory diagnoses or persistent respiratory symptoms, survival was associated with FVC in both men and women. Neither the FEV1 nor the FEV1/FVC ratio had any independent effect on survival. This result is at variance with the general assumption that the association of FEV1 with mortality—and more specifically with cardiovascular mortality—represents an association with airways obstruction.
Supporting our findings, there have been several other studies showing an association between FVC and both survival6 10 11 and more specific cardiovascular outcomes,3 4 12 and this alone makes interpretation of FEV1 as a marker of obstruction less convincing. Although it is true that some studies have failed to find an association with FEV1,13 this has been uncommon. In addition, where it has been studied, there is very little evidence of the FEV1/FVC ratio being of prognostic significance.3 4
It is extremely unlikely that the restriction observed in this analysis is due to the restrictive lung diseases as they are usually defined. These are rare conditions. Some restriction may be seen where there is a physical constriction of the lungs due either to abdominal obesity or where rib or diaphragm movement is limited.14 We were unable to specifically exclude the latter, but the findings here are adjusted for BMI and waist-hip ratio. Some restriction may also be seen in early obstructive airway disease due to gas trapping.15 16 We have no measurement of residual volume in this study, but it seems extremely unlikely that a subtle measure of early obstruction would have a strong association with mortality but that the FEV1/FVC ratio would not.
The association between mortality and low ventilatory function has been ascribed to comorbidities—specifically cardiovascular disease—and the link with cardiovascular disease is often ascribed to common risk factors such as smoking. The assumption behind many of these speculations, however, has been that the relevant pathology was airway obstruction.5 The same factors may, however, affect FVC. In the ARIC study, FVC was more closely associated with traffic exposure than was the FEV1/FVC ratio,17 and incident coronary heart disease was linked to both FVC and FEV1.18 Interestingly, the link with FVC was particularly noted in non-smokers and after adjusting for many other cardiovascular risk factors.
A link between diabetes (another comorbidity that could explain an excess mortality) and low FVC has been reported from many studies including ARIC,19 and has more often been ascribed to an adverse effect of diabetic pathology on ventilatory function,20 a view that is supported by the more rapid decline of lung function in those with diabetes.21
A further possibility is that the low FVC associated with increased mortality is associated with poor intrauterine lung growth. Barker's early study of adult ventilatory function and birth weight reported an association between lower birth weights and lower FEV1 in adulthood but did not report an association with a reduction in FVC.22 Since then, however, in spite of a few studies reporting no association between birth weight and subsequent ventilatory function,23–25 most studies have shown that the deficit in adult lung function associated with low birth weight is a deficit in FVC as well as FEV126 27 and not in the FEV1/FVC ratio.28–33
As cardiovascular disease is also closely associated with birth weight,34 this may also explain the association between lung function and cardiovascular mortality. The metabolic syndrome has similarly been associated with low birth weight,35 and this could also explain the particularly strong association of low lung function with metabolic syndrome36 and the increased mortality from COPD in those with metabolic syndrome.37
Another common feature of both cardiovascular disease and poor ventilatory function is low-grade systemic inflammation,38 and this has been used to explain at least part of the association between the two conditions.39 40 Low-grade inflammation is, however, also associated with poor intrauterine growth, as shown in both the MIDSPAN study41 and the North Finnish birth cohort,42–44 and this association may also reflect a common origin.
The observation that low FVC, type II diabetes, cardiovascular disease and low-grade systemic inflammation are all associated with each other and with low birth weight does not demonstrate that these conditions have a common cause, nor that low birth weight itself is the cause of the different conditions. Low birth weight is itself a combination of premature birth and poor fetal growth, and its causes are poorly understood. There is no guarantee that increasing birth weights within the normal range would reduce any of the associated conditions. Nor should it be assumed that, because the origins of the conditions are in early life, later efforts to mitigate the consequences would be ineffective. Nevertheless, the observation provides a potentially fruitful hypothesis to explain a very common condition.
The ARIC study is a large prospective study of good quality. Although the conclusions of a multiple regression analysis may be influenced by the relative quality of the information on different variables so that the better measured variables will tend to dominate the analyses, this does not seem to be a likely problem in this instance. If the results in table 3 were simply an artefact related to data quality, it would be expected that the more easily measured variable (in this case FEV1) would dominate the analysis. In the end it is the FVC that dominates the regression. There is no reason to suppose that the results from this large American study would be different in another population.
The results reported here apply to asymptomatic people. We looked at this group because we were interested in the association between mortality and spirometric values in the general population independent of overt obstructive disease. This group nevertheless has a substantial number of people with airway obstruction. Our results should not, however, be extrapolated to symptomatic populations. When we re-ran the analyses on this group, the results were different (see table E1 in the online supplement). Here all three spirometric measures (FEV1, FVC and the FEV1/FVC ratio) were associated with mortality and, when FEV1 and FVC were mutually adjusted for, each provided an estimated HR of 0.92 for both men and women. Only the HR for FEV1 in men was significantly different from 1, which may partly reflect the relatively small sample size of this group.
These findings may provide a partial explanation for why investigators have found that FEV1 rather than a measure of air flow obstruction predicts the prognosis of airway obstruction.45–47 The more obvious staging by the extent of the reduction in FEV1/FVC ratio would clearly not be as effective at estimating prognosis in this group at least. Although FVC might theoretically be more appropriate, the strong correlation with FEV1 makes the latter an adequate marker.
Survival in asymptomatic adults without chronic respiratory diagnoses or persistent respiratory symptoms is associated with FVC and not with airway obstruction as measured by the FEV1/FVC ratio. The relation of FEV1 to survival is indeed strong, but secondary to the association of FEV1 to FVC. The association with survival is not explained by age, anthropometry, smoking, income, occupation or blood pressure. As FVC later in life, cardiovascular risk, type II diabetes mellitus and low-grade systemic inflammation are all associated with poor fetal growth, these other conditions may be partly responsible for the poor survival in those with low FVC.
We would like to express our gratitude to the investigators and staff of the ARIC study and to Professor Neil Pride and Dr Sean Coady for helpful comments on an earlier draft.
See Editorial, p 1
Linked articles 151274.
Funding The Atherosclerosis Risk in Communities Study (ARIC) is conducted and supported by the NHLBI in collaboration with the ARIC Study Investigators. This manuscript was prepared using a limited access dataset obtained from the NHLBI and does not necessarily reflect the opinions or views of the ARIC investigators or the NHLBI. RH was supported by the Department of Health Policy Research Programme.
Competing interests None.
Ethics approval The ARIC study was approved by local institutional review boards in the USA and the analysis was approved by the Imperial College research ethics committee.
Provenance and peer review Not commissioned; externally peer reviewed.
If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.