According to the Global Initiative for Chronic Obstructive Lung Disease (GOLD), COPD is a preventable and treatable disease characterized by airflow limitation that is not fully reversible [1]. Spirometry measurements can be used to define COPD, specifically the forced expiratory volume in the first second (FEV1), and the FEV1 to forced vital capacity (FVC) ratio [1-3]. Accordingly, the American Thoracic Society / European Respiratory Society (ATS/ERS) standards for the diagnosis and management of patients with COPD state that spirometry should be performed in all patients suspected of COPD and is necessary for diagnosis, assessment of disease severity, and monitoring disease progression [3].

Recent estimates based on self-reported data show that 15 million Americans aged 18 years and older have been diagnosed with COPD [4]. Estimating the prevalence of COPD can, however, be challenging and may vary based on the diagnostic method (e.g. self-report or spirometry), the criteria used for defining COPD (e.g. GOLD, ATS/ERS), and the age group analyzed (e.g., 18 years and older or 40 years and older). Estimates will also vary if pre-bronchodilator or post-bronchodilator spirometry results are used with the GOLD guidelines advocating the use of post-bronchodilator spirometry [1] while the ATS/ERS guidelines support the use of pre-bronchodilator spirometry [5]. The GOLD criteria with the fixed cutoff may be simpler to use in daily clinical practice [6,7], however, studies suggest that it overestimates disease burden in the elderly [8,9]. While the ATS/ERS guidelines may be better suited to capture age-related decline in pulmonary function, their use is more computationally intensive and requires appropriate, population-based, reference equations for the interpretation of pulmonary function tests [5].

During 2007–2010, the U.S. National Health and Nutrition Examination Survey (NHANES) conducted a spirometry component which obtained pre-bronchodilator and post-bronchodilator pulmonary lung function data on survey participants 6–79 years of age. Data on respiratory symptoms and diseases were also collected. These data offer a unique opportunity to examine multiple approaches to estimating the prevalence of COPD in a large nationally representative sample.

The main objectives of this study were 1) to estimate the prevalence of COPD in the United States based on pre-bronchodilator and post-bronchodilator spirometry measurements and according to the operational definitions in the two major published guidelines: ATS/ERS and GOLD [1,5,10] and 2) to examine the potential impact of nonresponse on the estimated prevalence.

NHANES is a cross-sectional survey of the civilian, non-institutionalized U.S. population conducted by the National Center for Health Statistics of the Centers for Disease Control and Prevention [11]. Data were collected via household interviews and standardized physical examinations in specially equipped mobile examination centers (MEC). The NHANES survey samples are selected through a complex, multistage, probability design. Each annual sample is nationally representative, however NHANES data are publicly released for 2-year survey periods to protect confidentiality and increase statistical reliability. The 2007–10 NHANES survey cycles oversampled major U.S. demographic subgroups including Hispanic and non-Hispanic black persons, low income white persons, and persons aged 80 years and older. The procedures to select the sample and conduct the interview and examination have been specified elsewhere [12]. Informed consent was obtained from all participants and the National Center for Health Statistics Research Ethics Review Board approved the protocol.

This study was based on analysis of NHANES 2007–2008 and 2009–2010 data for participants 40–79 years. During 2007–2010, 9,985 persons aged 40–79 years were eligible for the survey, 7296 (73%) were interviewed and 7104 (71%) attended the NHANES MEC exam and were invited to participate in the spirometry component, which included pre-bronchodilator and post-bronchodilator lung function assessment. Of the 7104 eligible, 649 participants were excluded from spirometry for safety reasons: current chest pain or pain with forceful expiration, currently taking daytime supplemental oxygen, had recent surgery of the eye, chest or the abdomen; had a recent heart attack, stroke, tuberculosis exposure, hemoptysis, a history of detached retina or pneumothorax. An additional 632 participants did not receive spirometry due to limited time available in the MEC (n=337), subject refusals (n=113), or some other reason (n=182). In total, 5,823 adults aged 40–79, or 82% of those eligible, received spirometry.

Of these 5,823 examinees, 1110 were eligible for the bronchodilator study based on pre-bronchodilator spirometry values that indicated possible airflow obstruction per ATS/ERS or GOLD criteria. Possible airflow obstruction, per the ATS/ERS criterion, was defined as a pre-bronchodilator FEV1/FVC ratio less than the lower limit of normal (LLN) representing the lower 5^th percentile based on person’s age, sex, height, and race/ethnicity [5]. LLN values were determined using normative reference equations developed from NHANES III data by Hankinson et al. [13]. Possible airflow obstruction, per the GOLD criterion, was a pre-bronchodilator FEV1/FVC ratio less than 70%. Of the 1110 who qualified for post-bronchodilator testing, 238 did not perform the post-bronchodilator test due to subject refusals or limited time available in the MEC. An additional 305 were excluded due to safety reasons which included active cardiovascular disease (uncontrolled blood pressure, irregular pulse on examination, taking medication for major arrhythmia, having an implanted defibrillator, or history of congenital heart disease) or taking certain prescription medications (a monoamine oxidase inhibitor, an anticonvulsant, a tricyclic antidepressant plus current treatment for cardiac disease, or potassium lowering drugs). Examinees were also excluded from bronchodilator administration if they had already recently taken a β2-adrenergic bronchodilator to avoid exceeding FDA recommended doses, or if they had had a previous adverse reaction to albuterol. Also, women who were pregnant or breastfeeding were excluded. In total, 543 adults aged 40–79 did not receive the bronchodilator test while 567, or 51% of those eligible, received the test (Figure  1).

Among those who received the pre-bronchodilator spirometry test (n=5823) and the post-bronchodilator spirometry test (n=567), 94% and 100%, respectively, had acceptable quality data for analysis. Acceptable quality was defined as either meeting or exceeding the ATS data quality standard [14] of 3 acceptable and 2 reproducible spirometric curves with 2 observed values for both FEV1 and FVC within 150 ml (87% of all tests), or a lesser standard where there were 2 curves where FEV1 values were within 200 ml, and FVC values were within 200 ml (7% of all tests). The final pre-bronchodilator and post-bronchodilator analytic samples were 5477 and 564, respectively.

Characteristics of persons eligible for pre- and post-bronchodilator spirometry are summarized in Table  1. The overall exam completion rate for pre-bronchodilator spirometry was 77.1 percent. 1,627 eligible individuals were excluded from the analysis due to safety reasons, not having quality spirometry results or other non-response. Spirometry completion rates were higher among 40–59 year olds (81.4%) than 60–79 year olds (72.0%); p<0.001, progressively higher with more education (Less than 12^th grade, 69.2%; HS grad/equivalent, 78.1%; Some college or AA degree, 80.9%; College graduate or above, 84.1%; p<0.001), and higher for those with no self-reported COPD diagnosis (yes, 55.2%; No, 78.5%; p<0.001). There were also significant differences in spirometry completion rates with respect to race/ethnicity (Non-Hispanic white, 80.4%; Non-Hispanic black, 72.8%; Mexican-American, 75.9%; p<0.001).

Of the 1,110 persons eligible for the post-bronchodilator spirometry, 564 persons had complete, acceptable quality bronchodilator results (50.8% response rate). Mirroring the baseline spirometry completion rates, post-bronchodilator completion rates were higher among 40–59 year olds (58.2%) than 60–79 years olds (45.6%); p<0.001, progressively higher with more education (Less than 12^th grade, 44.4%; HS grad/equivalent, 46.8%; Some college or AA degree, 56.6%; College graduate or above, 59.8%; p<0.001), and higher for those with no self-reported COPD diagnosis (yes, 30.2%; No, 53.2.5%; p<0.001). In contrast to the baseline spirometry, more men completed the post-bronchodilator exam than women (54.7% vs. 44.8%; p<0.001) and there were also significant differences in post-bronchodilator completion rates by smoking status (Current smoker, 45.2%; Past smoker, 50.7%; Never smoked, 57.7%; p<.05).

Prevalence estimates of COPD based on the fixed ratio criterion are presented in Table  2. When applying the fixed ratio criterion to pre-bronchodilator spirometry values, the overall prevalence of COPD among adults aged 40–79 years was 20.3% (SE=1.1) when using the original examination weights and 20.9% (SE=1.1) with non-response adjusted weights. The prevalence estimates for demographic subgroups were similar when using the original weights and the nonresponse adjusted weights (0.1-0.6% differences). Regardless of which survey weights were used, about 11% of adults aged 40–79 years had mild disease, 8% had moderate, and 1% had severe/very severe COPD.

When applying the fixed ratio criterion to post-bronchodilator spirometry values, the overall prevalence of COPD was 8.3% (SE=0.7) using the original examination weights. After adjusting for nonresponse by imputation or reweighting, the post-bronchodilator estimate using the fixed ratio was 13.7% (SE=0.8) and 14.0% (SE=1.0), respectively. Estimates were similar for demographic subgroups and disease severity regardless of which non-response adjustment approach was used, that is imputation or reweighting. Specifically, about 7.5% were mild cases, 5% were moderate, and <1% was severe/very severe COPD.

Prevalence estimates of COPD based on the LLN criterion are presented in Table  3. Based on the LLN criterion and pre-bronchodilator spirometry values, the overall prevalence of COPD was 15.0% (SE=0.8) using the original examination weights and 15.4% (SE=0.8) using the adjusted weights. Estimates were similar for demographic subgroups and disease severity when using the original or adjusted weights. About 10% had mild, a little over 3% had moderate or moderately severe, and about 1% had severe/very severe disease. Prevalence was about 14-17% among all gender, age, and race/ethnic groups except among Mexican-American adults who had a prevalence of 8.7%.

Applying the LLN criterion to post-bronchodilator spirometry, the overall prevalence estimates for COPD using the original examination weights was 6.1% (SE=0.5). By comparison, after multiple imputation or after reweighting, the post-bronchodilator estimates using the LLN ratio were 10.2% (SE=0.7) and 10.2% (SE=0.8), respectively. Estimates were similar for demographic subgroups and by disease severity regardless of which non-response approach was used, that is either multiple imputation or reweighting adjustment. Specifically, about 6.5 to 7% had mild disease, about 2% had moderate or moderately severe, and about 0.5% had severe/very severe disease.

Spirometry can be used to define COPD yet currently established guidelines (e.g., ATS/ERS and GOLD) differ in their recommendation for using spirometric measurements to do so. Specifically, the GOLD guidelines recommend using post-bronchodilator FEV1/FVC <0.70 and ATS/ERS recommends pre-bronchodilator FEV1/FVC < LLN. Applying both of these criteria to the 2007–2010 NHANES spirometry data, we found the prevalence of COPD in the U.S. for individuals aged 40 to 79 years to be 14.0% using the fixed ratio criterion with post-bronchodilator spirometry (GOLD) and 15.4% using the LLN criterion with pre-bronchodilator spirometry (ATS/ERS). Currently, no other published post-bronchodilator COPD prevalence estimates exist for the overall U.S. population. A similar study, conducted outside the U.S., which utilized fixed ratio, post-bronchodilator spirometry data in an aged 40+ population was the PLATINO study which found a 14.3% pooled prevalence across five Latin American cities [22]. This is comparable to the estimated U.S. prevalence of 14.0% we found using fixed ratio, post-bronchodilator spirometry data in a population ages 40 to 79 years.

Examination of the differences between using pre-bronchodilator and post-bronchodilator data revealed a reduction in COPD prevalence after bronchodilation, relative to before bronchodilation, of 33% using the fixed ratio criterion (pre-bronchodilator 21% vs. post-bronchodilator 14%) and 34% for the LLN criterion (pre-bronchodilator 15% vs. post-bronchodilator 10%). This puts our pre- to post-bronchodilator reduction estimates at the high end, but within, the 25-35% range of reductions in prevalence found in a number of other studies [23-26]. This finding further strengthens the growing body of evidence that COPD prevalence rates computed using post-bronchodilator spirometry are likely to be 25-35% lower than those computed using pre-bronchodilator spirometry.

Comparing results obtained based on currently established guidelines, the overall COPD prevalence rates using the GOLD post-bronchodilator and ATS/ERS pre-bronchodilator criteria were somewhat similar at 14.0% (SE=1.0) and 15.4% (SE=0.8), respectively. COPD prevalence according to disease severity stage was somewhat similar, as well (overlapping confidence intervals), with some of the residual point estimate differences explained, in part, by the different cut-points used by the GOLD and ATS/ERS systems to define COPD severity stages. For example, the GOLD criteria define mild disease as greater than or equal to a percent of predicted FEV1 (PPFEV1) of ≥0.8, whereas ATS/ERS defines mild COPD with a slightly larger range of PPFEV1 ≥0.7. As a result, the calculated ATS/ERS prevalence estimate for mild COPD (pre-bronchodilator) was 10.2% (SE=0.5), compared to the 7.9% (SE=0.8) rate for the GOLD criteria (post-bronchodilator). Conversely, for moderate disease, the GOLD moderate stage has a slightly wider range of values (.5 ≤ PPFEV1 < .8) than the ATS/ERS moderate and moderately severe stage specifications (.5 ≤ PPFEV1 ≤ .7), hence the prevalence of moderate COPD severity was expectedly higher for the (post-bronchodilator) GOLD criteria (4.7%, SE=0.5) as compared to (pre-bronchodilator) ATS/ERS criteria (3.4%, SE=0.3). When the severity is equivalent for both diagnostic criteria, as is the case with severe and very severe disease (PPFEV1 < .5), the prevalence estimates were again similar with GOLD producing an estimate of 0.7% (SE=0.2) and ATS/ERS producing 1.2% (SE=0.2).

Using the fixed ratio criterion with post-bronchodilator spirometry led to greater variability than when using the LLN criterion with pre-bronchodilator spirometry when estimating the COPD prevalence by demographic subgroups. Most notably, COPD prevalence was 14.2% among those aged 40–59 years and 17.8% among those aged 60–79 years using LLN and pre-bronchodilator data compared to the 9.4% prevalence for 40–59 year olds and 23% prevalence for 60–79 year olds using the fixed ratio and post-bronchodilator data. This appears to be a common finding, as other studies have shown that using the fixed ratio to define airflow limitation will result in more frequent diagnosis of COPD in the elderly [8,9], and less frequent diagnosis among those younger than 45 years [27].

On the other hand, some criticisms of the LLN include the lack of longitudinal studies validating the use of the LLN and the argument that its use may underestimate the true burden of disease.

In this study, disease classification and subsequent prevalence estimates are based solely on spirometric assessments with the exception of self-reported supplemental oxygen use. Using spirometry alone to classify disease is consistent with the methods used in other population based studies, however, this represents a simplified and measurable case definition in a research study setting, as opposed to an actual disease diagnosis in a clinical setting [28] and could potentially lead to disease misdiagnosis. To reduce potential misdiagnosis, the most recent GOLD committee recommendations have included evaluating symptoms and risk factors, in addition to evaluating spirometric results, when clinically diagnosing COPD [1]. The fielding of our study preceded these recommendations and, as a result, it was not possible to evaluate all the symptoms and risk factors from our survey content. As daytime supplemental oxygen use together with self-reported, health care provider-diagnosed emphysema or current chronic bronchitis is thought to be indicative of the disease with a high degree of certainty, even in the absence of a spirometric assessment, persons with these characteristics were also classified as having COPD. Since they were excluded from the exam due to supplemental oxygen use, and subsequently had no spirometric data, the disease stage for these persons could not be determined. However, given their supplemental oxygen use, it is likely they are in one of the more severe stages of disease. These supplemental classification criteria, not based on spirometry measurements, applied to a small number of cases (55) and increased the overall, strictly spirometry-based prevalence rates only by a small amount: about one half of one percentage point.

A limitation of our study was that primarily due to safety exclusions and time available for exams, we had nearly 50% nonresponse for post-bronchodilator testing. This resulted in a significant portion of potentially positive disease cases being missing due to nonresponse. To account for this missing data, estimates were produced, and compared, using two different approaches: multiple imputation of the missing data and re-weighting of those cases with valid spirometry data. Although we applied different strategies to account for non-response, the presented estimates may still be an underestimation as some of the eligible participants were excluded from the bronchodilator portion of the exam due to the recent use of a prescribed bronchodilator. Recent bronchodilator use was not accounted for in the nonresponse adjustment.

Beyond missing data, there were also other limitations in the study. Persons with no race designation on the public use file (other) were assigned predicted pulmonary values which were 88% of the corresponding predicted pulmonary values for whites. This adjustment factor was originally derived for Asians and, since some participants in this “other” group may be other than Asian, this correction factor may cause some misclassification of COPD severity stage. Similarly, application of predicted values derived from Hispanics of Mexican origin to Hispanics of non-Mexican origin could possibly cause some misclassification of COPD severity stage due to differences in smoking habits, culture, and ancestry. The more severe stages of COPD may have been further underreported by limiting our analyses to the use of acceptable quality curves (grades of A, B, or C). Curve quality was determined using both within-maneuver evaluations (no cough during first second, no extra breaths, etc.) and between-maneuver evaluations (number of acceptable spirograms, ATS-defined reproducibility criteria) [14]. As the likelihood of producing reproducible spirometric measurements may decrease with increasing severity of lung disease, the exclusion of nonreproducible tests could potentially and selectively exclude a higher proportion of persons with obstructive airways disease [29,30]. We expected that this exclusion due to lack of test reproducibility would have a negligible impact on our results. The number of persons in the poor quality spirometry curve group was relatively small (349 persons out of 7104 eligible, or 5%) and the 6.3% rate of self-reported, health care provider-diagnosed COPD (affirmative response to either having emphysema or current bronchitis) for these 349 persons was only marginally higher than the 5.2% rate found for the entire group eligible for spirometry (n=7104).

Finally, the most recent GOLD guidelines recommend the use of either the Modified British Medical Research Council (mMRC) questionnaire or the COPD Assessment Test (CAT) for assessing symptoms in patients with COPD [1]. Although NHANES does collect some respiratory symptom data, it was not a close match to the GOLD specifications. ATS guidance regarding COPD symptom criteria is even less specific than the GOLD criteria. We therefore limited the scope of our study primarily to spirometric definitions of COPD to make valid comparisons between ATS and GOLD criteria. It should be noted, however, that the multiple imputation models to predict FEV1 and FVC values included available symptom information to generate maximally informed prediction equations.

In conclusion, the overall prevalence of COPD among U.S. adults aged 40–79 years ranged from 10.2% to 20.9% based on whether pre- or post-bronchodilator values were used and which diagnostic criterion (fixed ratio or LLN) was applied. These estimates provide the first U.S. national estimates of impaired lung function evidence of COPD using both pre- and post-bronchodilator spirometry measurements. Using the GOLD recommended combination of a fixed ratio criterion and post-bronchodilator spirometry resulted in an overall COPD prevalence estimate of approximately 14% while using the ATS/ERS recommended combination of LLN criterion and pre-bronchodilator spirometry resulted in an overall prevalence of about 15%. The overall prevalence decreased by approximately 33% when airflow limitation was based on post-bronchodilator testing as compared to pre-bronchodilator spirometry, regardless of which diagnostic criterion was used. Due to the large percentage of missing post-bronchodilator data, the NHANES post-bronchodilator data required adjustment for nonresponse but either method of adjustment, multiple imputation or reweighting, appeared to provide similar results. Nonresponse adjustment of the pre-bronchodilator data did not lead to significant differences in computed prevalence estimates.

The overall prevalence of COPD among US adults aged 40–79 years varied from 10.2% to 20.9% based on whether pre- or post-bronchodilator values were used and which diagnostic criterion (fixed ratio or LLN) was applied. The overall prevalence decreased by approximately 33% when airflow limitation was based on post-bronchodilator as compared to pre-bronchodilator spirometry, regardless of which diagnostic criterion was used.

The authors declare that they have no competing interests.

TT conceived of the analyses, analyzed and interpreted the data, and drafted the manuscript. CD participated in the design of the study, performed quality control on both exams and data, and helped to draft the manuscript. RP provided significant conceptual input, contributed intellectual content, and helped to draft the manuscript. EH provided clinical diagnostic expertise and provided critical review of the manuscript. BD contributed significant intellectual content and provided critical review of the manuscript. All authors read and approved the final manuscript.