Discriminative Accuracy of FEV1:FVC Thresholds for COPD-Related Hospitalization and Mortality | Chronic Obstructive Pulmonary Disease | JAMA | JAMA Network
[Skip to Navigation]
Sign In
Figure 1.  Incidence Density Rates for COPD-Related Hospitalization and Mortality According to Initial FEV1:FVC
Incidence Density Rates for COPD-Related Hospitalization and Mortality According to Initial FEV1:FVC

The incidence density rate (IDR) for chronic obstructive pulmonary disease (COPD)-related hospitalization and mortality was inversely related to the ratio of the forced expiratory volume in the first second to the forced vital capacity (FEV1:FVC) without clear evidence of an inflection point. IDRs and 95% CIs were computed per 1000 person-years via Poisson regression in mutually exclusive, incremental categories of initial FEV1:FVC. The predicted IDR (curved black line) is generated from the best function (quadratic 0.40 to 0.77; piecewise linear 0.77 to 0.80) describing the FEV1:FVC and the corresponding IDR relationship. The blue line indicates IDR per 1000 person-years (95% CI indicated by blue shading), computed via Poisson regression, for all patients with an FEV1:FVC less than the lower limit of normal per Global Lung Function Initiative reference equations.

Figure 2.  Discriminative Accuracy of Various Fixed FEV1:FVC Thresholds for Airflow Obstruction With Respect to COPD-Related Hospitalization and Mortality
Discriminative Accuracy of Various Fixed FEV1:FVC Thresholds for Airflow Obstruction With Respect to COPD-Related Hospitalization and Mortality

The C statistics (95% CI) for fixed-ratio thresholds (dots) were estimated separately in unadjusted Cox proportional hazards models that included only 1 dichotomous predictor (ie, whether a participant had a baseline ratio of forced expiratory volume in the first second to forced vital capacity [FEV1:FVC] that was above or below a given ratio threshold). The C statistic for the lower-limit-of-normal (LLN) threshold is indicated by the solid blue horizontal line (95% CI indicated by blue shading). The optimal threshold based on highest C statistic was 0.71, but C statistics were not significantly different for 0.70 to 0.72. Fixed ratio thresholds 0.66 to 0.74 yielded C statistics that were significantly higher than the LLN threshold (P values <.05).

Figure 3.  Weighted Youden Index for Various FEV1:FVC Thresholds Across a Range of Relative Weights for Sensitivity and Specificity
Weighted Youden Index for Various FEV1:FVC Thresholds Across a Range of Relative Weights for Sensitivity and Specificity

The weighted Youden index for the 0.70 threshold was higher than the lower limit of normal (LLN) threshold for weights of 0.35 or greater. Where weight equals 0.5, sensitivity and specificity are weighted equally. For each ratio threshold, the sensitivity and specificity were estimated from unadjusted Cox proportional hazards models including ratio thresholds only. Weight (x-axis) indicates relative weight assigned to sensitivity vs specificity. FEV1:FVC indicates the ratio of the forced expiratory volume in the first second to the forced vital capacity.

Table 1.  Baseline Characteristics
Baseline Characteristics
Table 2.  Sensitivity Analyses for the Discriminative Accuracy of Various Fixed FEV1:FVC Thresholds for COPD Event Riska
Sensitivity Analyses for the Discriminative Accuracy of Various Fixed FEV1:FVC Thresholds for COPD Event Riska
Supplement.

eTable 1. Design Features of Cohorts Included

eTable 2. Baseline Characteristics, Stratified by FEV1:FVC <0.70

eTable 3. Reclassification of Airflow Obstruction According to Fixed FEV1:FVC Threshold

eTable 4. Brier Scores (95% Confidence Intervals) for the Prediction of COPD Events by Various FEV1:FVC Thresholds in Unadjusted Models

eTable 5. Sensitivity Analyses for the Discriminative Accuracy of Various Fixed FEV1:FVC Thresholds for COPD Event Risk in Adjusted Models

eTable 6. Brier Scores (95% Confidence Intervals) for the Prediction of COPD Events by Various FEV1:FVC Thresholds in Adjusted Models

eFigure 1. CONSORT

eFigure 2. Venn Diagram Showing the Extent of Overlap for Several Different FEV1:FVC Threshold Groups and, for Each Group, the Incidence Density Rates (IDRs) per 1000 Person-Years for COPD-Related Hospitalization and Mortality

eFigure 3. Calibration Plot Comparing Predicted and Observed 10-Year Risk for COPD-Related Events, by Various FEV1:FVC Stratum for Airflow Limitation, in Covariate-Adjusted Models

eFigure 4. Discriminative Accuracy of Various Fixed FEV1:FVC Thresholds for Airflow Obstruction With Respect to COPD-Related Hospitalization and Mortality, Stratified by Smoking Status

eFigure 5. Discriminative Accuracy of Various Fixed FEV1:FVC Thresholds for Airflow Obstruction With Respect to COPD-Related Hospitalization Versus COPD-Related Mortality

eFigure 6. Discriminative Accuracy of Various fixed FEV1:FVC Thresholds for Airflow Obstruction With Respect to COPD-Related Hospitalization and Mortality, Stratified by Sex

eFigure 7. Discriminative Accuracy of Various Fixed FEV1:FVC Thresholds for Airflow Obstruction With Respect to Events Classified as Primarily Caused by COPD Versus the Primary End Point

eFigure 8. Incremental Discriminative Accuracy of Various Fixed FEV1:FVC Thresholds for Airflow Obstruction With Respect to COPD-Related Hospitalization and Mortality, When Added to a Model Adjusted for Sociodemographic Factors and Smoking Status

eFigure 9. Incremental Discrimination With Sequential Covariate Adjustment to the Fixed FEV1:FVC Thresholds for Airflow Obstruction With Respect to COPD-Related Hospitalization and Mortality

1.
Collaborators GBDCRD; GBD 2015 Chronic Respiratory Disease Collaborators.  Global, regional, and national deaths, prevalence, disability-adjusted life years, and years lived with disability for chronic obstructive pulmonary disease and asthma, 1990-2015: a systematic analysis for the Global Burden of Disease Study 2015.  Lancet Respir Med. 2017;5(9):691-706. doi:10.1016/S2213-2600(17)30293-XPubMedGoogle ScholarCrossref
2.
Mannino  DM, Homa  DM, Akinbami  LJ, Ford  ES, Redd  SC.  Chronic obstructive pulmonary disease surveillance—United States, 1971-2000.  MMWR Surveill Summ. 2002;51(6):1-16.PubMedGoogle Scholar
3.
Celli  BR, Halbert  RJ.  Point: should we abandon FEV1/FVC <0.70 to detect airway obstruction? no.  Chest. 2010;138(5):1037-1040. doi:10.1378/chest.10-2049PubMedGoogle ScholarCrossref
4.
Enright  P, Brusasco  V.  Counterpoint: should we abandon FEV1/FVC <0.70 to detect airway obstruction? yes.  Chest. 2010;138(5):1040-1042. doi:10.1378/chest.10-2052PubMedGoogle ScholarCrossref
5.
Vogelmeier  CF, Criner  GJ, Martinez  FJ,  et al.  Global Strategy for the Diagnosis, Management, and Prevention of Chronic Obstructive Lung Disease 2017 Report: GOLD executive summary.  Am J Respir Crit Care Med. 2017;195(5):557-582. doi:10.1164/rccm.201701-0218PPPubMedGoogle ScholarCrossref
6.
Qaseem  A, Wilt  TJ, Weinberger  SE,  et al; American College of Physicians; American College of Chest Physicians; American Thoracic Society; European Respiratory Society.  Diagnosis and management of stable chronic obstructive pulmonary disease: a clinical practice guideline update from the American College of Physicians, American College of Chest Physicians, American Thoracic Society, and European Respiratory Society.  Ann Intern Med. 2011;155(3):179-191. doi:10.7326/0003-4819-155-3-201108020-00008PubMedGoogle ScholarCrossref
7.
National Clinical Guideline Centre.  Chronic Obstructive Pulmonary Disease: Management of Chronic Obstructive Pulmonary Disease in Adults in Primary and Secondary Care. NICE Clinical Guidelines, No. 101. London, UK: Royal College of Physicians; 2010.
8.
Whelton  PK, Carey  RM, Aronow  WS,  et al.  2017 ACC/AHA/AAPA/ABC/ACPM/AGS/APhA/ASH/ASPC/NMA/PCNA guideline for the prevention, detection, evaluation, and management of high blood pressure in adults: a report of the American College of Cardiology/American Heart Association Task Force on Clinical Practice Guidelines.  J Am Coll Cardiol. 2018;71(19):e127-e248. doi:10.1016/j.jacc.2017.11.006PubMedGoogle ScholarCrossref
9.
American Diabetes Association.  2. Classification and diagnosis of diabetes: standards of medical care in diabetes-2018.  Diabetes Care. 2018;41(suppl 1):S13-S27. doi:10.2337/dc18-S002PubMedGoogle ScholarCrossref
10.
Tilert  T, Dillon  C, Paulose-Ram  R, Hnizdo  E, Doney  B.  Estimating the US prevalence of chronic obstructive pulmonary disease using pre- and post-bronchodilator spirometry: the National Health and Nutrition Examination Survey (NHANES) 2007-2010.  Respir Res. 2013;14:103. doi:10.1186/1465-9921-14-103PubMedGoogle ScholarCrossref
11.
Hankinson  JL, Odencrantz  JR, Fedan  KB.  Spirometric reference values from a sample of the general US population.  Am J Respir Crit Care Med. 1999;159(1):179-187. doi:10.1164/ajrccm.159.1.9712108PubMedGoogle ScholarCrossref
12.
Quanjer  PH, Stanojevic  S, Cole  TJ,  et al; ERS Global Lung Function Initiative.  Multi-ethnic reference values for spirometry for the 3-95-yr age range: the Global Lung Function 2012 equations.  Eur Respir J. 2012;40(6):1324-1343. doi:10.1183/09031936.00080312PubMedGoogle ScholarCrossref
13.
Oelsner  EC, Balte  PP, Cassano  PA,  et al.  Harmonization of respiratory data from 9 US population-based cohorts: the NHLBI Pooled Cohorts Study.  Am J Epidemiol. 2018;187(11):2265-2278. doi:10.1093/aje/kwy139PubMedGoogle ScholarCrossref
14.
ARIC Investigators.  The Atherosclerosis Risk in Communities (ARIC) Study: design and objectives.  Am J Epidemiol. 1989;129(4):687-702. doi:10.1093/oxfordjournals.aje.a115184PubMedGoogle ScholarCrossref
15.
Fried  LP, Borhani  NO, Enright  P,  et al.  The Cardiovascular Health Study: design and rationale.  Ann Epidemiol. 1991;1(3):263-276. doi:10.1016/1047-2797(91)90005-WPubMedGoogle ScholarCrossref
16.
Goodpaster  BH, Carlson  CL, Visser  M,  et al.  Attenuation of skeletal muscle and strength in the elderly: the Health ABC Study.  J Appl Physiol (1985). 2001;90(6):2157-2165. doi:10.1152/jappl.2001.90.6.2157PubMedGoogle ScholarCrossref
17.
Bild  DE, Bluemke  DA, Burke  GL,  et al.  Multi-Ethnic Study of Atherosclerosis: objectives and design.  Am J Epidemiol. 2002;156(9):871-881. doi:10.1093/aje/kwf113PubMedGoogle ScholarCrossref
18.
Miller  MR, Hankinson  J, Brusasco  V,  et al; ATS/ERS Task Force.  Standardisation of spirometry.  Eur Respir J. 2005;26(2):319-338. doi:10.1183/09031936.05.00034805PubMedGoogle ScholarCrossref
19.
Oelsner  EC, Loehr  LR, Henderson  AG,  et al.  Classifying chronic lower respiratory disease events in epidemiologic cohort studies.  Ann Am Thorac Soc. 2016;13(7):1057-1066. doi:10.1513/AnnalsATS.201601-063OCPubMedGoogle ScholarCrossref
20.
Kumar  R, Seibold  MA, Aldrich  MC,  et al.  Genetic ancestry in lung-function predictions.  N Engl J Med. 2010;363(4):321-330. doi:10.1056/NEJMoa0907897PubMedGoogle ScholarCrossref
21.
Rodriguez  J, Jiang  R, Johnson  WC, MacKenzie  BA, Smith  LJ, Barr  RG.  The association of pipe and cigar use with cotinine levels, lung function, and airflow obstruction: a cross-sectional study.  Ann Intern Med. 2010;152(4):201-210. doi:10.7326/0003-4819-152-4-201002160-00004PubMedGoogle ScholarCrossref
22.
Harrell  FE  Jr, Califf  RM, Pryor  DB, Lee  KL, Rosati  RA.  Evaluating the yield of medical tests.  JAMA. 1982;247(18):2543-2546. doi:10.1001/jama.1982.03320430047030PubMedGoogle ScholarCrossref
23.
Harrell  FE  Jr, Lee  KL, Mark  DB.  Multivariable prognostic models: issues in developing models, evaluating assumptions and adequacy, and measuring and reducing errors.  Stat Med. 1996;15(4):361-387. doi:10.1002/(SICI)1097-0258(19960229)15:4<361::AID-SIM168>3.0.CO;2-4PubMedGoogle ScholarCrossref
24.
Kang  L, Chen  W, Petrick  NA, Gallas  BD.  Comparing two correlated C indices with right-censored survival outcome: a one-shot nonparametric approach.  Stat Med. 2015;34(4):685-703. doi:10.1002/sim.6370PubMedGoogle ScholarCrossref
25.
Steyerberg  EW, Vickers  AJ, Cook  NR,  et al.  Assessing the performance of prediction models: a framework for traditional and novel measures.  Epidemiology. 2010;21(1):128-138. doi:10.1097/EDE.0b013e3181c30fb2PubMedGoogle ScholarCrossref
26.
Youden  WJ.  Index for rating diagnostic tests.  Cancer. 1950;3(1):32-35. doi:10.1002/1097-0142(1950)3:1<32::AID-CNCR2820030106>3.0.CO;2-3PubMedGoogle ScholarCrossref
27.
Li  DL, Shen  F, Yin  Y, Peng  JX, Chen  PY.  Weighted Youden index and its two-independent-sample comparison based on weighted sensitivity and specificity.  Chin Med J (Engl). 2013;126(6):1150-1154.PubMedGoogle Scholar
28.
Genant  HK, Cooper  C, Poor  G,  et al.  Interim report and recommendations of the World Health Organization Task-Force for Osteoporosis.  Osteoporos Int. 1999;10(4):259-264. doi:10.1007/s001980050224PubMedGoogle ScholarCrossref
29.
Bhatt  SP.  Diagnosis of chronic obstructive pulmonary disease: breathing new life into an old debate.  Ann Am Thorac Soc. 2018;15(2):163-165. doi:10.1513/AnnalsATS.201709-733EDPubMedGoogle ScholarCrossref
30.
Bhatt  SP, Sieren  JC, Dransfield  MT,  et al; COPDGene Investigators.  Comparison of spirometric thresholds in diagnosing smoking-related airflow obstruction.  Thorax. 2014;69(5):409-414. doi:10.1136/thoraxjnl-2012-202810PubMedGoogle ScholarCrossref
31.
García-Rio  F, Soriano  JB, Miravitlles  M,  et al.  Overdiagnosing subjects with COPD using the 0.7 fixed ratio: correlation with a poor health-related quality of life.  Chest. 2011;139(5):1072-1080. doi:10.1378/chest.10-1721PubMedGoogle ScholarCrossref
32.
Izquierdo Alonso  JL, De Lucas Ramos  P, Rodríguez Glez-Moro  JM; grupo de estudio CONSISTE.  The use of the lower limit of normal as a criterion for COPD excludes patients with increased morbidity and high consumption of health-care resources.  Arch Bronconeumol. 2012;48(7):223-228. doi:10.1016/j.arbres.2012.02.007PubMedGoogle ScholarCrossref
33.
Mannino  DM, Sonia Buist  A, Vollmer  WM.  Chronic obstructive pulmonary disease in the older adult: what defines abnormal lung function?  Thorax. 2007;62(3):237-241. doi:10.1136/thx.2006.068379PubMedGoogle ScholarCrossref
34.
Mannino  DM, Diaz-Guzman  E.  Interpreting lung function data using 80% predicted and fixed thresholds identifies patients at increased risk of mortality.  Chest. 2012;141(1):73-80. doi:10.1378/chest.11-0797PubMedGoogle ScholarCrossref
35.
Çolak  Y, Afzal  S, Nordestgaard  BG, Vestbo  J, Lange  P.  Young and middle-aged adults with airflow limitation according to lower limit of normal but not fixed ratio have high morbidity and poor survival: a population-based prospective cohort study.  Eur Respir J. 2018;51(3):1702681. doi:10.1183/13993003.02681-2017PubMedGoogle ScholarCrossref
36.
Woodruff  PG, Barr  RG, Bleecker  E,  et al; SPIROMICS Research Group.  Clinical significance of symptoms in smokers with preserved pulmonary function.  N Engl J Med. 2016;374(19):1811-1821. doi:10.1056/NEJMoa1505971PubMedGoogle ScholarCrossref
37.
Sood  A, Petersen  H, Qualls  C,  et al.  Spirometric variability in smokers: transitions in COPD diagnosis in a five-year longitudinal study.  Respir Res. 2016;17(1):147. doi:10.1186/s12931-016-0468-7PubMedGoogle ScholarCrossref
38.
Mannino  DM, Diaz-Guzman  E, Buist  S.  Pre- and post-bronchodilator lung function as predictors of mortality in the Lung Health Study.  Respir Res. 2011;12:136. doi:10.1186/1465-9921-12-136PubMedGoogle ScholarCrossref
39.
Fortis  S, Eberlein  M, Georgopoulos  D, Comellas  AP.  Predictive value of prebronchodilator and postbronchodilator spirometry for COPD features and outcomes.  BMJ Open Respir Res. 2017;4(1):e000213. doi:10.1136/bmjresp-2017-000213PubMedGoogle ScholarCrossref
40.
Huiart  L, Ernst  P, Suissa  S.  Cardiovascular morbidity and mortality in COPD.  Chest. 2005;128(4):2640-2646. doi:10.1378/chest.128.4.2640PubMedGoogle ScholarCrossref
Original Investigation
June 25, 2019

Discriminative Accuracy of FEV1:FVC Thresholds for COPD-Related Hospitalization and Mortality

Author Affiliations
  • 1Division of Pulmonary, Allergy, and Critical Care Medicine and the UAB Lung Health Center, University of Alabama at Birmingham
  • 2Division of General Medicine, Columbia University Medical Center, New York, New York
  • 3Division of Nutritional Sciences, Weill Cornell Medical College, Ithaca, New York
  • 4Gillings School of Global Public Health, University of North Carolina, Chapel Hill
  • 5Division of Epidemiology and Community Health, School of Public Health, University of Minnesota, Minneapolis
  • 6Division of Pulmonary and Critical Care Medicine, Northwestern University, Chicago, Illinois
  • 7Division of Pulmonary, Allergy, Sleep, and Critical Care, Boston University, Boston, Massachusetts
  • 8Department of Critical Care Medicine, University of Pittsburgh and Veterans Affairs Pittsburgh Healthcare System, Pittsburgh, Pennsylvania
  • 9Division of Pulmonary and Critical Care Medicine, Brigham and Women’s Hospital, Boston, Massachusetts
  • 10MedStar Health Research Institute, Hyattsville, Maryland
  • 11Benjamin Leon Center for Geriatric Research and Education, Florida International University, Miami
  • 12Undergraduate Training and Education Center, Tougaloo College, Tougaloo, Mississippi
  • 13Department of Epidemiology, Mailman School of Public Health, Columbia University Medical Center, New York, New York.
JAMA. 2019;321(24):2438-2447. doi:10.1001/jama.2019.7233
Key Points

Question  What is the discriminative accuracy of various thresholds for the ratio of the forced expiratory volume in the first second to the forced vital capacity (FEV1:FVC) for predicting chronic obstructive pulmonary disease (COPD)-related hospitalization and mortality?

Findings  Among 24 207 participants from 4 US general population–based cohorts, the optimal fixed threshold for discriminating COPD-related events was 0.71 (C statistic for the optimal fixed threshold, 0.696). The discriminative accuracy of the 0.71 threshold was not significantly different than that of the 0.70 threshold (difference, 0.001) but it was more accurate than a lower-limit-of-normal threshold derived from population-based reference equations (difference between the optimal ratio threshold vs the model using the LLN threshold, 0.034). The 0.70 threshold provided optimal discrimination in a subgroup analysis of ever smokers and in adjusted models.

Meaning  These results support the use of FEV1:FVC less than 0.70 to identify individuals at risk of clinically significant COPD.

Abstract

Importance  According to numerous current guidelines, the diagnosis of chronic obstructive pulmonary disease (COPD) requires a ratio of the forced expiratory volume in the first second to the forced vital capacity (FEV1:FVC) of less than 0.70, yet this fixed threshold is based on expert opinion and remains controversial.

Objective  To determine the discriminative accuracy of various FEV1:FVC fixed thresholds for predicting COPD-related hospitalization and mortality.

Design, Setting, and Participants  The National Heart, Lung, and Blood Institute (NHLBI) Pooled Cohorts Study harmonized and pooled data from 4 US general population–based cohorts (Atherosclerosis Risk in Communities Study; Cardiovascular Health Study; Health, Aging, and Body Composition Study; and Multi-Ethnic Study of Atherosclerosis). Participants aged 45 to 102 years were enrolled from 1987 to 2000 and received follow-up longitudinally through 2016.

Exposures  Presence of airflow obstruction, which was defined by a baseline FEV1:FVC less than a range of fixed thresholds (0.75 to 0.65) or less than the lower limit of normal as defined by Global Lung Initiative reference equations (LLN).

Main Outcomes and Measures  The primary outcome was a composite of COPD hospitalization and COPD-related mortality, defined by adjudication or administrative criteria. The optimal fixed FEV1:FVC threshold was defined by the best discrimination for these COPD-related events as indexed using the Harrell C statistic from unadjusted Cox proportional hazards models. Differences in C statistics were compared with respect to less than 0.70 and less than LLN thresholds using a nonparametric approach.

Results  Among 24 207 adults in the pooled cohort (mean [SD] age at enrollment, 63 [10.5] years; 12 990 [54%] women; 16 794 [69%] non-Hispanic white; 15 181 [63%] ever smokers), complete follow-up was available for 11 077 (77%) at 15 years. During a median follow-up of 15 years, 3925 participants experienced COPD-related events over 340 757 person-years of follow-up (incidence density rate, 11.5 per 1000 person-years), including 3563 COPD-related hospitalizations and 447 COPD-related deaths. With respect to discrimination of COPD-related events, the optimal fixed threshold (0.71; C statistic for optimal fixed threshold, 0.696) was not significantly different from the 0.70 threshold (difference, 0.001 [95% CI, −0.002 to 0.004]) but was more accurate than the LLN threshold (difference, 0.034 [95% CI, 0.028 to 0.041]). The 0.70 threshold provided optimal discrimination in the subgroup analysis of ever smokers and in adjusted models.

Conclusions and Relevance  Defining airflow obstruction as FEV1:FVC less than 0.70 provided discrimination of COPD-related hospitalization and mortality that was not significantly different or was more accurate than other fixed thresholds and the LLN. These results support the use of FEV1:FVC less than 0.70 to identify individuals at risk of clinically significant COPD.

Introduction

Chronic obstructive pulmonary disease (COPD) is the third leading cause of death worldwide.1 In the United States, the prevalence of COPD is estimated to be 24 million, of which half remains undiagnosed.2 Confusion regarding how to diagnose airflow obstruction, the major physiological feature of COPD, remains a major hurdle to improving care for patients with COPD.3,4

Quiz Ref IDMajor respiratory society guidelines recommend diagnosing airflow obstruction when the ratio of the forced expiratory volume in the first second to the forced vital capacity (FEV1:FVC) is less than a fixed threshold of 0.70.5-7 This approach is analogous to current clinical approaches to hypertension and diabetes, for which the identification of fixed disease thresholds has resulted in significant improvements in early detection and treatment.8,9 However, there remains no rigorous, population-based evidence to support the 0.70 threshold, which was set by expert opinion as the FEV1:FVC threshold for defining clinically significant airflow obstruction.

The selection of a threshold for defining airflow obstruction has major implications for patient care and public health as the prevalence of airflow obstruction can vary by as much as 33% depending on which threshold is selected.10 To account for differences in FEV1:FVC according to demographic factors, airflow obstruction can be defined by an FEV1:FVC less than the lower limit of normal (LLN), which can be predicted from population-based normative data adjusted for age, sex, race, and height.11,12 However, in addition to pragmatic issues, concerns regarding the LLN approach include the premise that low absolute levels of lung function could be interpreted as normal in women, individuals who are not white, or elderly individuals.

The aim of this study was to determine the discriminative accuracy of various FEV1:FVC fixed thresholds for predicting COPD-related hospitalization and mortality in a large, multiethnic, US general population–based sample of adults.

Methods
Study Population

Quiz Ref IDThe National Heart, Lung, and Blood Institute (NHLBI) Pooled Cohorts Study harmonized data from 9 US general population–based studies that collected spirometry data.13 The current report is limited to 4 cohorts that completed follow-up for COPD-related clinical events: the Atherosclerosis Risk in Communities Study (ARIC)14; Cardiovascular Health Study (CHS)15; Health, Aging and Body Composition Study (Health ABC)16; and the Multi-Ethnic Study of Atherosclerosis (MESA).17 CHS and Health ABC were designed to study older adults, whereas ARIC and MESA included both middle-aged and older adults (eTable 1 in the Supplement).17 All studies were approved by institutional review boards at participating institutions, and all participants provided written informed consent. Secondary data analysis for this work was approved by the Columbia University institutional review board.

Spirometry

Spirometry was performed using water-seal, dry-rolling seal or flow-sensing spirometers in accordance with the American Thoracic Society criteria and quality controlled using 2005 criteria.18 To minimize measurement error, only participants with valid spirometry measurements were retained for analyses. Using the Global Lung Function Initiative approach, predicted values were calculated based on age, sex, race, and height, and the LLN for the FEV1:FVC was defined as the 5th percentile of the distribution of the standard deviation (Ζ score).12 National Health and Nutrition Examination Survey (NHANES) III spirometric reference equations were used in secondary analyses.11

Outcomes

The primary outcome was a composite of COPD-related mortality and COPD-related hospitalization. A clinical events committee adjudicated COPD-related clinical events in Health ABC (hospitalizations and deaths) and CHS (deaths only). For hospitalizations and deaths in ARIC and MESA and for nonfatal hospitalizations in CHS, International Classification of Diseases, Ninth Revision and International Classification of Diseases, Tenth Revision (ICD-9; ICD-10) codes were used to classify COPD-related events (COPD: ICD-9, 496 and ICD-10, J44; chronic bronchitis: ICD-9, 490-491 and ICD-10 J40-J42; and emphysema: ICD-9, 492 and ICD-10, J43), following a previously validated protocol.13 The primary outcome, a COPD-related event, was defined as first hospitalization or death adjudicated as primarily or secondarily attributable to COPD or, if adjudication was lacking, events with COPD listed in any diagnosis field. In prior work in MESA, 82% of such administratively defined events were confirmed by a physician as evidence of clinical COPD.17,19

In sensitivity analyses, COPD-related events were decomposed into COPD-related hospitalizations and COPD-related mortality. Also, only those events adjudicated or ICD coded as primarily caused by COPD were separately assessed. This end point was previously found to have a positive predictive value of 97% for physician-adjudicated exacerbations.19

Covariates

Age and sex were self-reported at enrollment. Race was self-reported according to fixed categories. The cohorts did not include a separate question regarding ethnicity, although MESA and Health ABC participants were asked to self-report as white, black, Asian/Pacific Islander, or Hispanic/Latino race/ethnicity (eTable 1 in the Supplement). Race/ethnicity was included as a covariate since this study aimed to evaluate discriminative accuracy in a multiethnic US general population setting, and race/ethnicity has been associated with lung function and COPD risk.20 Current smoking status and pack-years were assessed at baseline by self-report, with biochemical verification in a subset.21 Lifetime smoking status was classified as never or ever by comparison of self-reported smoking status over all available examinations.13 Height was measured using standard methods. Due to the extensive quality control and harmonization efforts performed in the NHLBI Pooled Cohorts Study,13 missing covariate data at enrollment were rare (<1% [eFigure 1 in the Supplement]).

Statistical Analyses

The incidence density rate (IDR) of COPD-related events per 1000 person-years of follow-up was plotted by initial FEV1:FVC, which was stratified by 0.01 increments over the range of 0.40 to 0.80, as was the IDR for all participants with FEV1:FVC less than the LLN. The functional form of the relationship between the FEV1:FVC and the IDR was explored by use of deviance statistics.

To evaluate the discriminative accuracy of different threshold-based definitions for airflow obstruction, airflow obstruction was dichotomized according to 11 fixed-ratio definitions (0.01 decrements over the interval of 0.75 to 0.65) and the LLN definition. Each fixed-threshold definition was modeled separately using Cox proportional hazards models. Time-to-event data were defined as time since measurement of FEV1:FVC for each individual. Non-COPD mortality and loss to follow-up were treated as censoring events. The proportional hazards assumption was confirmed by Kaplan-Meier curves and residual plots.

The criterion for identifying the optimal fixed threshold was defined a priori as the threshold that generated the highest Harrell C statistic, which is a rank-correlation measure of the concordance between observed and predicted outcomes in the setting of censored survival data.22,23 The Harrell C statistic is an estimate of the area under the curve for a receiver operating characteristic curve that adjusts for censoring. C statistics were compared between the optimal ratio threshold model, the 0.70 threshold model, and the LLN threshold model. Formal statistical comparisons of C statistics (difference in C statistics from the model using the optimal ratio threshold vs the model using the 0.70 threshold and the difference in C statistics from the model using the optimal ratio threshold vs the model using the LLN threshold) were performed using a nonparametric approach to compare 2 correlated C statistics with right-censored survival outcomes.24 Model fit was assessed by the Brier score.25

The same unadjusted Cox proportional hazards models were used to calculate classification rates for each fixed threshold and the LLN. Based on the sensitivity (true positive rate) and specificity (true negative rate) of each threshold, the Youden index (sensitivity + specificity −1) was calculated.26 In the primary analyses, which included a single binary predictor, maximizing the Harrell C statistic was equivalent to maximizing the Youden index. Because the Youden index assigns equal utility to sensitivity and specificity, which may not be consistent with clinical priorities, public health priorities, or both, a weighted Youden index was plotted across a range of potential relative weights for sensitivity and specificity.27 Positive predictive value and negative predictive value were also calculated based on the observed event rates.

As sensitivity analyses, stratified models were performed according to smoking status and sex. Analyses were repeated for alternative outcome definitions: COPD-related hospitalization, COPD-related mortality, and events with adjudicated or ICD coded as primarily due to COPD. For comparison with the primary unadjusted approach, the incremental improvement in discrimination was evaluated when adding each ratio threshold to a Cox proportional hazards model adjusted for age at enrollment, sex, race/ethnicity, height, birth year, site, and cohort. Discrimination by the LLN-Global Lung Function Initiative was compared with that of the LLN-NHANES.

A 2-tailed alpha of .05 was considered significant for all analyses. Because of the potential for type 1 error due to multiple comparisons, findings for secondary analyses should be interpreted as exploratory. Beyond the exclusion of participants with missing or invalid spirometry and the censoring associated with loss to follow-up and non-COPD mortality, there were no missing data in the primary analyses. Secondary analyses that were stratified, adjusted for covariates, or both were restricted to complete case analyses. Analyses were completed using SAS, version 9.4 (Cary, North Carolina).

Results
Baseline Characteristics

After exclusions (eFigure 1 in the Supplement), there were 24 207 participants (Table 1). Mean age at spirometry was 63 years. Women constituted 54% of the cohort; 69% were non-Hispanic white, and 24% were black. Sixty-three percent were ever smokers and 37% were never smokers. Complete follow-up for COPD-related events was available for 97% of participants at 5 years, 85% at 10 years, and 77% at 15 years.

Prevalence of Airflow Obstruction

According to the LLN threshold, 3646 (15%) of participants had airflow obstruction (eFigure 2 in the Supplement). Compared with the LLN threshold, a fixed threshold of less than 0.66 yielded the most similar prevalence (3576 participants [15%]), but 540 (15%) of those with FEV1:FVC less than 0.66 did not meet the LLN classification, and 610 (17%) of those meeting the LLN criterion were excluded.

There were 6261 (26%) participants with FEV1:FVC less than 0.70, including all but 19 (0.5%) of participants meeting the LLN criterion (eFigure 2 in the Supplement). The IDR was 19.0 for participants with FEV1:FVC less than 0.70 but greater than or equal to the LLN threshold. By comparison, the IDR was 17.2 for participants with FEV1:FVC between 0.66 and 0.70 and less than the LLN threshold. A fixed ratio of less than 0.73 was required to capture 100% of participants below the LLN threshold (eTable 3 in the Supplement).

Incidence of COPD-Related Events

During a median follow-up of 15 years (interquartile range [IQR], 9 to 22), 3925 participants experienced COPD-related events over 340 757 person-years of follow-up (IDR, 11.5), including 3563 COPD-related hospitalizations (IDR, 10.5) and 447 COPD-related deaths (IDR, 1.3).

The IDR for COPD-related events was inversely related to the FEV1:FVC (Figure 1). Initially, a cubic spline–smoothed curve, with smoothing parameter selected by the generalized cross-validation score, was used to describe the FEV1:FVC and corresponding IDR relationship. The optimal functional form for the FEV1:FVC and corresponding IDR relationship was a quadratic model over the interval (<0.40, 0.77) with a piecewise linear component over the interval (0.77, >0.80); this parametric model outperformed the cubic spline based on the deviance statistics for each. This suggested that participants with FEV1:FVC of at least 0.77 had minimal COPD-related event risk; however, it did not indicate an inflection point over the remainder of the FEV1:FVC range.

Discrimination of COPD-Related Events

In the primary analysis, the 0.71 threshold demonstrated the highest C statistic (0.696 [95% CI, 0.688 to 0.703]) (Figure 2). Discrimination by the 0.71 threshold was not significantly different than by the 0.70 threshold (difference, 0.001 [95% CI, −0.002 to 0.004]; P = .57), but was significantly more accurate than that of the LLN threshold (difference, 0.034 [95% CI, 0.028 to 0.041]; P < .001; Table 2). Taking all pairwise comparisons into account, C statistics were not significantly different vs the 0.70 threshold over the fixed threshold interval (0.70, 0.72) and were significantly more accurate than the LLN threshold over the fixed threshold interval (0.66, 0.74) (Figure 2). Brier scores were nominally lower for the LLN threshold (eTable 4 in the Supplement).

Sensitivity and Specificity for COPD-Related Events

The sensitivity for the LLN was 52%, and the specificity for the LLN was 89%, which approximated the results for a fixed 0.66 threshold (Figure 3). By comparison, for the 0.70 threshold, the sensitivity was 66%, and the specificity was 79%. Compared with the LLN, the weighted sum of sensitivity and specificity were greater for the 0.70 threshold under all conditions in which sensitivity was given equal or greater weight than specificity. The negative predictive value was at least 0.90 for fixed thresholds of 0.66 to 0.71, as well as for the LLN. The positive predictive value was 0.44 for the LLN threshold and 0.37 for the 0.70 threshold.

Sensitivity Analyses

In the majority of sensitivity analyses, as in the primary analysis, the C statistic for the optimal fixed threshold was not significantly different from the C statistic for the model using a fixed-ratio threshold of 0.70 and was significantly better than the C statistic for the model using the LLN (Table 2). The 2 exceptions were in groups with relatively low event rates. First, a significant difference between using the C statistic for the optimal fixed threshold and the C statistic for the model using a fixed-ratio threshold of 0.70 was observed in never smokers (IDR [4.04]; optimal fixed threshold [0.74]; eFigure 4 in the Supplement). Second, for COPD-related mortality (IDR [1.31]; optimal fixed threshold [0.69]; eFigure 5 in the Supplement), differences between using the C statistic for the optimal fixed thresholdvs the C statistic for the model using a fixed-ratio threshold of 0.70 were statistically significant, and differences between using the C statistic for the optimal fixed threshold vs the C statistic for the model using the LLN were not significant. The optimal fixed threshold was 0.70 in ever smokers (eFigure 4 in the Supplement) and in men (eFigure 6 in the Supplement), 0.71 for COPD-related hospitalizations (eFigure 5 in the Supplement), and 0.69 for the events for which COPD was the primary or underlying cause (eFigure 7 in the Supplement).

The C statistic for the covariates-only base model was 0.680 (95% CI, 0.671 to 0.689; eFigure 8 in the Supplement). Addition of any ratio threshold to the covariates-only model significantly improved discrimination (P < .001 for all). The optimal fixed threshold, when added to the adjusted model, was 0.70 (C statistic for the optimal fixed threshold, 0.760 [95% CI, 0.752 to 0.768]), although discrimination was not significantly different over the fixed-threshold interval (0.66, 0.71). Whereas the LLN threshold yielded significantly less accurate discrimination than the optimal fixed threshold in the primary analysis, incremental discrimination by the LLN and 0.70 thresholds converged once models were adjusted for age (eFigure 9 in the Supplement), and discrimination by the LLN threshold was not significantly different from the optimal fixed threshold in the fully adjusted analysis (difference in the C statistic for the optimal fixed threshold vs LLN, −0.002 [95% CI, −0.006 to 0.002]; Table 2). Consistent with agreement between predicted and observed outcomes, calibration plots of predicted to observed 10-year event risk showed overlapping predictive performance for LLN and 0.70 thresholds, and Brier scores were similar (eFigure 3 and eTable 6 in the Supplement). Results were similar after adjustment in stratified analyses and in analyses using alternative end points (eTable 5 in the Supplement).

Compared with the Global Lung Function Initiative definition for LLN, the NHANES definition for LLN attained a greater C statistic in unadjusted analyses (0.667 [95% CI, 0.658 to 0.675]), but the difference was small (0.005 [95% CI, 0.002 to 0.008]). Both the 0.70 and 0.71 thresholds were associated with significantly greater discriminative accuracy vs the NHANES LLN (difference in C statistic 0.70 vs NHANES, 0.028 [95% CI 0.022 to 0.034], and difference in C statistic 0.71 vs NHANES, 0.029 [95% CI, 0.023 to 0.035]).

Discussion

Quiz Ref IDIn this study based on pooled data from 4 US general population–based cohorts, a fixed threshold of 0.70 for the FEV1:FVC provided discrimination of COPD-related hospitalization and mortality that was not significantly different or was more accurate than other fixed thresholds and population-based reference equations. Hence, the present work provides population-based evidence to support 0.70 as the optimal FEV1:FVC threshold for defining clinically significant airflow obstruction.

Until the findings of this research, the 0.70 fixed threshold to diagnose airflow obstruction was based on expert opinion. However, expert opinions have historically diverged. Over time, 2 distinct perspectives emerged. The first considers the age, sex, and race/ethnicity dependence of lung function as part of the normal variance and contends that reference equations drawn from the normal population should inform deviation from normality.11,12 This approach is, to some extent, similar to current definitions of osteoporosis,28 yet even these are based on normative values for maximum bone density, not age-specific predictions. An alternative perspective is that the manner by which a certain lung size or degree of airflow limitation is achieved is immaterial, but that beyond this threshold of normal lung function, respiratory reserve is overcome and there are clinical consequences.3 This latter view, positing a fixed threshold for harm, is more consistent with current guidelines for high blood pressure and diabetes.8,9

Establishing a diagnostic threshold that is easy to use is critical to improve adaptation of spirometry in primary care and to facilitate epidemiologic follow-up and multiregional clinical trials. Identifying individuals below the 5th percentile of normal using population-based reference values may be statistically sound, but this approach assumes that the prevalence of airflow obstruction has to be at least 5%. It is also sensitive to population differences: this study found significantly better discrimination by the NHANES III LLN, which was developed from a US population, vs the Global Lung Function Initiative LLN. Furthermore, as populations demonstrate changing demographics such as obesity, which are not accounted for in the reference equations, different reference equations drawn from the same population over time can result in differing definitions of normal and abnormal.29 There are also important differences in reference equations. For instance, the Global Lung Function Initiative equation for FEV1:FVC adjusts for height, whereas the NHANES III LLN does not.

Although the aim of this study was to identify the optimal fixed FEV1:FVC threshold to discriminate risk of COPD-related events in a general population-based context, a number of sensitivity analyses were performed. Of particular clinical interest was a subgroup analysis in ever smokers, who constitute the majority but far from all of COPD cases. Among ever smokers, the optimal ratio threshold was 0.70. In never smokers, the optimal ratio threshold was 0.74, but event rates were low and 0.70 still offered more accurate prediction compared with the LLN. No thresholds were significantly more accurate than 0.70 across strata of sex or in analyses adjusted for sociodemographic and anthropometric characteristics, which suggests that 0.70 may be applicable to all adults.

Quiz Ref IDThe selection of a diagnostic threshold requires trade-offs between sensitivity and specificity with important ramifications for underdiagnosis and overdiagnosis, and no ratio threshold is unassailable. Most reference equations assume uniformity of variance across patient ages, which means they are more likely to yield lower values for LLN and hence, underdiagnose airflow obstruction at older ages. Conversely, due to the age-related decline in lung function, the possible consequences of using a fixed threshold are overdiagnosis in older individuals (which could result in unnecessary medication) and underdiagnosis in younger individuals (which could lead to missed opportunities for recommending smoking cessation and early initiation of therapy).4 With respect to potential overdiagnosis by fixed thresholds, the prior literature has established that 7% to 23% of older adults meet the 0.70 threshold but not LLN criteria, yet longitudinal studies have mostly shown that these discordant participants have a greater degree of structural lung disease on computed tomography, worse quality of life, and greater health care utilization and mortality when compared with individuals without airflow obstruction by either criteria.2,30-34 With regards to potential underdiagnosis of younger individuals by a fixed threshold, evidence suggests this is of minimal importance: only 1% of young adults meeting the LLN criterion was missed by the fixed threshold of 0.70 in the Copenhagen General Population cohort; moreover, compared with those without airflow obstruction by any criteria, these individuals were no different in terms of COPD- or asthma-related exacerbations on follow-up.35

Quiz Ref IDRegardless of the threshold selected, the specificity and particularly sensitivity of airflow obstruction were modest, confirming recent observations that spirometry alone does not detect all individuals at risk of COPD-related events36 and also confirming that some patients with airflow obstruction on spirometry may not report clinical symptoms. For cases in which the FEV1:FVC value is borderline, especially in the absence of symptoms, it may be prudent to recommend close monitoring as recent data suggest the diagnosis of airflow obstruction in these individuals may not be stable.37 Nonetheless, weighted analyses suggested that the 0.70 threshold would be preferred to lower fixed ratio thresholds and the LLN as long as sensitivity carries equal or greater weight than specificity.

Strengths of the current work include the use of a large, US general population–based sample, supporting the generalizability of our results. The sample also included a large number of never smokers and less than 10 pack-year smokers who are commonly excluded from major studies of COPD. Lung function was systematically harmonized, and outcomes were defined by adjudication or a validated protocol using administrative data.13,17

Limitations

This study has several limitations. First, although the GOLD (Global Initiative for Chronic Obstructive Lung Disease) guidelines recommend using postbronchodilator values for confirming airflow obstruction,5 this study used prebronchodilator spirometry. Nonetheless, multiple studies have shown strong correlations between prebronchodilator and postbronchodilator spirometry measures, and both perform similarly in predicting respiratory outcomes.38,39 Furthermore, the LLN, the major comparator for this work, is calculated on prebronchodilator values only.

Second, no adjustments were made for medication use. Third, as participants were selected across cohorts, there are baseline differences in demographics and historical differences in disease management over time. Covariate-adjusted models included birth year, site, and cohort to alleviate potential biases due to this heterogeneity.

Fourth, although outcomes were longitudinal, the FEV1:FVC threshold was determined at baseline. In cases of borderline lung function, it is possible that some participants do not consistently meet criteria for airflow limitation.37

Fifth, there was loss to follow-up among participants, yet potential attrition biases were mitigated by use of survival models designed to account for censored data.

Sixth, the composite of COPD-related hospitalization and mortality was selected as the primary outcome; restricting the outcome to COPD mortality or to clinical events deemed to be primarily caused by COPD would be expected to exclude a large number of participants with mild-to-moderate disease who suffer exacerbations and in whom hospitalizations and mortality are more often due to cardiovascular causes than to respiratory.40 Regardless, sensitivity analyses for events primarily caused by COPD yielded an optimal threshold (0.69) that was not significantly different with respect to discrimination compared with 0.70. COPD hospitalization and mortality can be biased by preexisting knowledge of lung function, but physicians involved in classifying events were blinded to study spirometry,19 and clinical spirometry results are not often available in the medical record. This approach is analogous to that adopted by cardiovascular risk scores to predict events, but unlike cardiac events, respiratory events are not characterized by elevated levels of any biomarkers, and clinical diagnosis is the current criterion standard.

Conclusions

Defining airflow obstruction as an FEV1:FVC of less than 0.70 provided discrimination of COPD-related hospitalization and mortality that was not significantly different than or was more accurate than other fixed thresholds and the LLN. These results support use of an FEV1:FVC of less than 0.70 to identify individuals at risk of clinically significant COPD.

Back to top
Article Information

Corresponding Authors: Elizabeth C. Oelsner, MD, MPH, Division of General Medicine, Columbia University Medical Center, 630 W 168th St, Presbyterian Hospital Ninth Floor, Ste 105, New York, NY 10032 (eco7@cumc.columbia.edu); Surya P. Bhatt, MD, MSPH, Division of Pulmonary, Allergy, and Critical Care Medicine, University of Alabama at Birmingham, THT 422, 1720 Second Ave S, Birmingham, AL 35294 (sbhatt@uabmc.edu).

Accepted for Publication: May 23, 2019.

Author Contributions: Drs Oelsner and Bhatt had full access to all of the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis.

Concept and design: Bhatt, Oelsner.

Acquisition, analysis, or interpretation of data: All authors.

Drafting of the manuscript: Bhatt, Oelsner.

Critical revision of the manuscript for important intellectual content: All authors.

Statistical analysis: Balte, Schwartz, Oelsner.

Obtained funding: Oelsner.

Administrative, technical, or material support: O'Connor, Umans, White, Oelsner.

Supervision: Oelsner.

Conflict of Interest Disclosures: Dr Bhatt reports support from a National Institutes of Health (NIH) grant (K23 HL 133438) during the course of the study; receipt of consulting fees from Sunovion; and other (research funds to the institution) from Proterix Bio outside the submitted work. Dr Dransfield reports receipt of grants from NIH/National Heart, Lung, and Blood Institute (NHLBI) during the conduct of the study and from the Department of Defense and the American Lung Association outside the submitted work; contracted clinical trials from GlaxoSmithKline, Novartis, AstraZeneca, Yungjin, PneumRx/BTG and PulmonX; and consulting/personal fees from Boehringer Ingelheim, GlaxoSmithKline, PneumRx/BTG, Genentech, Boston Scientific, Quark Pharmaceuticals, and Mereo. Dr Couper reports receipt of grants from NHLBI and the COPD Foundation during the conduct of the study. Dr Kalhan reports receipt of grants from NHLBI during the conduct of the study; and outside the submitted work: grants and personal fees from Boehringer Ingelheim, AstraZeneca, and GlaxoSmithKline; grants from PneumRx/BTG, Spiration, and CVS Caremark; and personal fees from Aptus Health and Boston Scientific. Dr O’Connor reports receipt of grants from NIH during the conduct of the study and from Janssen Pharmaceuticals outside the submitted work; and personal/consulting fees from AstraZeneca. Dr Schwartz reports receipt of grants from NHLBI during the conduct of the study. Dr Balte reports receipt of grants from NHLBI during the conduct of the study. Dr Yende reports receipt of personal fees from Atox Bio and grants from Bristol-Myers Squibb outside the submitted work. Dr Umans reports receipt of grants from NIH/NHLBI outside the submitted work. Dr Oelsner reports receipt of grants from NIH/NHLBI during the conduct of the study. No other disclosures were reported.

Funding/Support: Dr Bhatt is supported by NIH grant K23 HL133438. Dr Oelsner is supported by NIH grants R21 HL129924 and K23 HL130627. The Atherosclerosis Risk in Communities (ARIC) study has been funded in whole or in part with federal funds from NIH, NHLBI, and the Department of Health and Human Services (contract numbers: HHSN268201700001I, HHSN268201700002I, HHSN268201700003I, HHSN268201700005I, and HHSN268201700004I). The Cardiovascular Health Study (CHS) was supported by contracts HHSN268201200036C, HHSN268200800007C, HHSN268201800001C, N01HC55222, N01HC85079, N01HC85080, N01HC85081, N01HC85082, N01HC85083, N01HC85086, and grants U01HL080295 and U01HL130114 from NHLBI, with an additional contribution from the National Institute of Neurological Disorders and Stroke. Additional support was provided by R01AG023629 from the National Institute on Aging (NIA). The Health, Aging and Body Composition (Health ABC) study was funded by NIA contracts N01-AG-6-2101, N01-AG-6-2103, N01-AG-6-2106, NIA grant R01-AG028050, National Institute of Nursing Research grant R01-NR012459, and supported in part by the intramural research program at NIA. the Multi-Ethnic Study of Atherosclerosis study was funded by NIH/NHLBI grants R01-HL-077612, R01-HL-093081, RC1-HL-100543, N01-HC-95159, N01-HC-95160, N01-HC-95161, N01-HC-95162, N01-HC-95163, N01-HC-95164, N01-HC-95165, N01-HC-95166, N01-HC-95167, N01-HC-95168 and N01-HC-95169.

Role of the Funder/Sponsor: The NIH had no role in the design and conduct of the study; collection, management, analysis, and interpretation of the data; preparation, review, or approval of the manuscript; and decision to submit the manuscript for publication.

Disclaimer: The content is solely the responsibility of the authors and does not necessarily represent the official views of NIH.

Additional Contributions: The authors thank the staff and participants of ARIC, CHS, Health ABC, and MESA studies for their important contributions.

Additional Information: A full list of principal CHS investigators and institutions can be found at CHS-NHLBI.org.

References
1.
Collaborators GBDCRD; GBD 2015 Chronic Respiratory Disease Collaborators.  Global, regional, and national deaths, prevalence, disability-adjusted life years, and years lived with disability for chronic obstructive pulmonary disease and asthma, 1990-2015: a systematic analysis for the Global Burden of Disease Study 2015.  Lancet Respir Med. 2017;5(9):691-706. doi:10.1016/S2213-2600(17)30293-XPubMedGoogle ScholarCrossref
2.
Mannino  DM, Homa  DM, Akinbami  LJ, Ford  ES, Redd  SC.  Chronic obstructive pulmonary disease surveillance—United States, 1971-2000.  MMWR Surveill Summ. 2002;51(6):1-16.PubMedGoogle Scholar
3.
Celli  BR, Halbert  RJ.  Point: should we abandon FEV1/FVC <0.70 to detect airway obstruction? no.  Chest. 2010;138(5):1037-1040. doi:10.1378/chest.10-2049PubMedGoogle ScholarCrossref
4.
Enright  P, Brusasco  V.  Counterpoint: should we abandon FEV1/FVC <0.70 to detect airway obstruction? yes.  Chest. 2010;138(5):1040-1042. doi:10.1378/chest.10-2052PubMedGoogle ScholarCrossref
5.
Vogelmeier  CF, Criner  GJ, Martinez  FJ,  et al.  Global Strategy for the Diagnosis, Management, and Prevention of Chronic Obstructive Lung Disease 2017 Report: GOLD executive summary.  Am J Respir Crit Care Med. 2017;195(5):557-582. doi:10.1164/rccm.201701-0218PPPubMedGoogle ScholarCrossref
6.
Qaseem  A, Wilt  TJ, Weinberger  SE,  et al; American College of Physicians; American College of Chest Physicians; American Thoracic Society; European Respiratory Society.  Diagnosis and management of stable chronic obstructive pulmonary disease: a clinical practice guideline update from the American College of Physicians, American College of Chest Physicians, American Thoracic Society, and European Respiratory Society.  Ann Intern Med. 2011;155(3):179-191. doi:10.7326/0003-4819-155-3-201108020-00008PubMedGoogle ScholarCrossref
7.
National Clinical Guideline Centre.  Chronic Obstructive Pulmonary Disease: Management of Chronic Obstructive Pulmonary Disease in Adults in Primary and Secondary Care. NICE Clinical Guidelines, No. 101. London, UK: Royal College of Physicians; 2010.
8.
Whelton  PK, Carey  RM, Aronow  WS,  et al.  2017 ACC/AHA/AAPA/ABC/ACPM/AGS/APhA/ASH/ASPC/NMA/PCNA guideline for the prevention, detection, evaluation, and management of high blood pressure in adults: a report of the American College of Cardiology/American Heart Association Task Force on Clinical Practice Guidelines.  J Am Coll Cardiol. 2018;71(19):e127-e248. doi:10.1016/j.jacc.2017.11.006PubMedGoogle ScholarCrossref
9.
American Diabetes Association.  2. Classification and diagnosis of diabetes: standards of medical care in diabetes-2018.  Diabetes Care. 2018;41(suppl 1):S13-S27. doi:10.2337/dc18-S002PubMedGoogle ScholarCrossref
10.
Tilert  T, Dillon  C, Paulose-Ram  R, Hnizdo  E, Doney  B.  Estimating the US prevalence of chronic obstructive pulmonary disease using pre- and post-bronchodilator spirometry: the National Health and Nutrition Examination Survey (NHANES) 2007-2010.  Respir Res. 2013;14:103. doi:10.1186/1465-9921-14-103PubMedGoogle ScholarCrossref
11.
Hankinson  JL, Odencrantz  JR, Fedan  KB.  Spirometric reference values from a sample of the general US population.  Am J Respir Crit Care Med. 1999;159(1):179-187. doi:10.1164/ajrccm.159.1.9712108PubMedGoogle ScholarCrossref
12.
Quanjer  PH, Stanojevic  S, Cole  TJ,  et al; ERS Global Lung Function Initiative.  Multi-ethnic reference values for spirometry for the 3-95-yr age range: the Global Lung Function 2012 equations.  Eur Respir J. 2012;40(6):1324-1343. doi:10.1183/09031936.00080312PubMedGoogle ScholarCrossref
13.
Oelsner  EC, Balte  PP, Cassano  PA,  et al.  Harmonization of respiratory data from 9 US population-based cohorts: the NHLBI Pooled Cohorts Study.  Am J Epidemiol. 2018;187(11):2265-2278. doi:10.1093/aje/kwy139PubMedGoogle ScholarCrossref
14.
ARIC Investigators.  The Atherosclerosis Risk in Communities (ARIC) Study: design and objectives.  Am J Epidemiol. 1989;129(4):687-702. doi:10.1093/oxfordjournals.aje.a115184PubMedGoogle ScholarCrossref
15.
Fried  LP, Borhani  NO, Enright  P,  et al.  The Cardiovascular Health Study: design and rationale.  Ann Epidemiol. 1991;1(3):263-276. doi:10.1016/1047-2797(91)90005-WPubMedGoogle ScholarCrossref
16.
Goodpaster  BH, Carlson  CL, Visser  M,  et al.  Attenuation of skeletal muscle and strength in the elderly: the Health ABC Study.  J Appl Physiol (1985). 2001;90(6):2157-2165. doi:10.1152/jappl.2001.90.6.2157PubMedGoogle ScholarCrossref
17.
Bild  DE, Bluemke  DA, Burke  GL,  et al.  Multi-Ethnic Study of Atherosclerosis: objectives and design.  Am J Epidemiol. 2002;156(9):871-881. doi:10.1093/aje/kwf113PubMedGoogle ScholarCrossref
18.
Miller  MR, Hankinson  J, Brusasco  V,  et al; ATS/ERS Task Force.  Standardisation of spirometry.  Eur Respir J. 2005;26(2):319-338. doi:10.1183/09031936.05.00034805PubMedGoogle ScholarCrossref
19.
Oelsner  EC, Loehr  LR, Henderson  AG,  et al.  Classifying chronic lower respiratory disease events in epidemiologic cohort studies.  Ann Am Thorac Soc. 2016;13(7):1057-1066. doi:10.1513/AnnalsATS.201601-063OCPubMedGoogle ScholarCrossref
20.
Kumar  R, Seibold  MA, Aldrich  MC,  et al.  Genetic ancestry in lung-function predictions.  N Engl J Med. 2010;363(4):321-330. doi:10.1056/NEJMoa0907897PubMedGoogle ScholarCrossref
21.
Rodriguez  J, Jiang  R, Johnson  WC, MacKenzie  BA, Smith  LJ, Barr  RG.  The association of pipe and cigar use with cotinine levels, lung function, and airflow obstruction: a cross-sectional study.  Ann Intern Med. 2010;152(4):201-210. doi:10.7326/0003-4819-152-4-201002160-00004PubMedGoogle ScholarCrossref
22.
Harrell  FE  Jr, Califf  RM, Pryor  DB, Lee  KL, Rosati  RA.  Evaluating the yield of medical tests.  JAMA. 1982;247(18):2543-2546. doi:10.1001/jama.1982.03320430047030PubMedGoogle ScholarCrossref
23.
Harrell  FE  Jr, Lee  KL, Mark  DB.  Multivariable prognostic models: issues in developing models, evaluating assumptions and adequacy, and measuring and reducing errors.  Stat Med. 1996;15(4):361-387. doi:10.1002/(SICI)1097-0258(19960229)15:4<361::AID-SIM168>3.0.CO;2-4PubMedGoogle ScholarCrossref
24.
Kang  L, Chen  W, Petrick  NA, Gallas  BD.  Comparing two correlated C indices with right-censored survival outcome: a one-shot nonparametric approach.  Stat Med. 2015;34(4):685-703. doi:10.1002/sim.6370PubMedGoogle ScholarCrossref
25.
Steyerberg  EW, Vickers  AJ, Cook  NR,  et al.  Assessing the performance of prediction models: a framework for traditional and novel measures.  Epidemiology. 2010;21(1):128-138. doi:10.1097/EDE.0b013e3181c30fb2PubMedGoogle ScholarCrossref
26.
Youden  WJ.  Index for rating diagnostic tests.  Cancer. 1950;3(1):32-35. doi:10.1002/1097-0142(1950)3:1<32::AID-CNCR2820030106>3.0.CO;2-3PubMedGoogle ScholarCrossref
27.
Li  DL, Shen  F, Yin  Y, Peng  JX, Chen  PY.  Weighted Youden index and its two-independent-sample comparison based on weighted sensitivity and specificity.  Chin Med J (Engl). 2013;126(6):1150-1154.PubMedGoogle Scholar
28.
Genant  HK, Cooper  C, Poor  G,  et al.  Interim report and recommendations of the World Health Organization Task-Force for Osteoporosis.  Osteoporos Int. 1999;10(4):259-264. doi:10.1007/s001980050224PubMedGoogle ScholarCrossref
29.
Bhatt  SP.  Diagnosis of chronic obstructive pulmonary disease: breathing new life into an old debate.  Ann Am Thorac Soc. 2018;15(2):163-165. doi:10.1513/AnnalsATS.201709-733EDPubMedGoogle ScholarCrossref
30.
Bhatt  SP, Sieren  JC, Dransfield  MT,  et al; COPDGene Investigators.  Comparison of spirometric thresholds in diagnosing smoking-related airflow obstruction.  Thorax. 2014;69(5):409-414. doi:10.1136/thoraxjnl-2012-202810PubMedGoogle ScholarCrossref
31.
García-Rio  F, Soriano  JB, Miravitlles  M,  et al.  Overdiagnosing subjects with COPD using the 0.7 fixed ratio: correlation with a poor health-related quality of life.  Chest. 2011;139(5):1072-1080. doi:10.1378/chest.10-1721PubMedGoogle ScholarCrossref
32.
Izquierdo Alonso  JL, De Lucas Ramos  P, Rodríguez Glez-Moro  JM; grupo de estudio CONSISTE.  The use of the lower limit of normal as a criterion for COPD excludes patients with increased morbidity and high consumption of health-care resources.  Arch Bronconeumol. 2012;48(7):223-228. doi:10.1016/j.arbres.2012.02.007PubMedGoogle ScholarCrossref
33.
Mannino  DM, Sonia Buist  A, Vollmer  WM.  Chronic obstructive pulmonary disease in the older adult: what defines abnormal lung function?  Thorax. 2007;62(3):237-241. doi:10.1136/thx.2006.068379PubMedGoogle ScholarCrossref
34.
Mannino  DM, Diaz-Guzman  E.  Interpreting lung function data using 80% predicted and fixed thresholds identifies patients at increased risk of mortality.  Chest. 2012;141(1):73-80. doi:10.1378/chest.11-0797PubMedGoogle ScholarCrossref
35.
Çolak  Y, Afzal  S, Nordestgaard  BG, Vestbo  J, Lange  P.  Young and middle-aged adults with airflow limitation according to lower limit of normal but not fixed ratio have high morbidity and poor survival: a population-based prospective cohort study.  Eur Respir J. 2018;51(3):1702681. doi:10.1183/13993003.02681-2017PubMedGoogle ScholarCrossref
36.
Woodruff  PG, Barr  RG, Bleecker  E,  et al; SPIROMICS Research Group.  Clinical significance of symptoms in smokers with preserved pulmonary function.  N Engl J Med. 2016;374(19):1811-1821. doi:10.1056/NEJMoa1505971PubMedGoogle ScholarCrossref
37.
Sood  A, Petersen  H, Qualls  C,  et al.  Spirometric variability in smokers: transitions in COPD diagnosis in a five-year longitudinal study.  Respir Res. 2016;17(1):147. doi:10.1186/s12931-016-0468-7PubMedGoogle ScholarCrossref
38.
Mannino  DM, Diaz-Guzman  E, Buist  S.  Pre- and post-bronchodilator lung function as predictors of mortality in the Lung Health Study.  Respir Res. 2011;12:136. doi:10.1186/1465-9921-12-136PubMedGoogle ScholarCrossref
39.
Fortis  S, Eberlein  M, Georgopoulos  D, Comellas  AP.  Predictive value of prebronchodilator and postbronchodilator spirometry for COPD features and outcomes.  BMJ Open Respir Res. 2017;4(1):e000213. doi:10.1136/bmjresp-2017-000213PubMedGoogle ScholarCrossref
40.
Huiart  L, Ernst  P, Suissa  S.  Cardiovascular morbidity and mortality in COPD.  Chest. 2005;128(4):2640-2646. doi:10.1378/chest.128.4.2640PubMedGoogle ScholarCrossref
×