Abstract
Rationale: The lambda-mu-sigma (LMS) method is a novel approach that defines the lower limit of normal (LLN) for the ratio of FEV1/FVC as the fifth percentile of the distribution of Z scores. The clinical validity of this threshold as a basis for establishing chronic obstructive pulmonary disease is unknown.
Objective: To evaluate the association between the LMS method of determining the LLN for the FEV1/FVC, set at successively higher thresholds, and clinically meaningful outcomes.
Methods: Using data from a nationally representative sample of 3,502 white Americans aged 40–80 years, we stratified the FEV1/FVC according to the LMS-LLN, with thresholds set at the 5th, 10th, 15th, 20th, and 25th percentiles (i.e., LMS-LLN5, LMS-LLN10, etc.). We then evaluated whether these thresholds were associated with an increased risk of death or prevalence of respiratory symptoms. Spirometry was not specifically completed after a bronchodilator.
Measurements and Main Results: Relative to an FEV1/FVC greater than or equal to LMS-LLN25 (reference group), the risk of death and the odds of having respiratory symptoms were elevated only in participants who had an FEV1/FVC less than LMS-LLN5, with an adjusted hazard ratio of 1.68 (95% confidence interval, 1.34–2.12) and an adjusted odds ratio of 2.46 (95% confidence interval, 2.01–3.02), respectively, representing 13.8% of the cohort. Results were similar for persons aged 40–64 years and those aged 65–80 years.
Conclusions: In white persons aged 40–80 years, an FEV1/FVC less than LMS-LLN5 identifies persons with an increased risk of death and prevalence of respiratory symptoms. These results support the use of the LMS-LLN5 threshold for establishing chronic obstructive pulmonary disease.
The lambda-mu-sigma (LMS) method is a novel approach that defines the lower limit of normal (LLN) for the ratio of FEV1/FVC as the fifth percentile of the distribution of Z scores. The clinical validity of this threshold as a basis for establishing chronic obstructive pulmonary disease (COPD) is unknown.
In persons aged 40–80 years, an FEV1/FVC less than LMS-LLN at the fifth percentile (LMS-LLN5) identifies those with an increased risk of death and prevalence of respiratory symptoms.
Establishing COPD based solely on an FEV1/FVC of less than 0.70 is seriously flawed (1, 2, 4). Because a fixed ratio does not account for normal age-related changes in airflow limitation, GOLD guidelines are likely to misdiagnose COPD, particularly in older persons (1, 2, 4). Likewise, establishing COPD based on the ATS/ERS-defined LLN can be problematic (2). Specifically, ATS/ERS guidelines define the LLN as the fifth percentile of the frequency distribution of reference values (ATS/ERS-LLN5), as derived by multiple regression in “healthy” never-smokers (6). However, multiple regression assumes that the relationship between the predictor variables (e.g., age, height, sex, and ethnicity) and spirometric measures is linear, and that reference values are normally distributed and have constant variability (2). These assumptions are incorrect, especially among older persons, because reference values for the FEV1/FVC range widely and are increasingly skewed at the extremes of age (2, 7).
Recently, Stanojevic and colleagues (2, 7) have proposed that the LLN should be calculated using the lambda-mu-sigma (LMS) method, an approach widely used to construct growth charts. These investigators have shown that the LMS method more accurately describes the relationship between spirometric lung function and anthropometric predictor variables across the lifespan (2, 7). Based on their results, the LMS-derived LLN for the FEV1/FVC is defined as the fifth percentile of the distribution of Z scores (LMS-LLN5) (2, 7). In the LMS method, the Z score accounts for: (1) the median (mu), representing how the spirometric variable changes with a predictor variable; (2) the coefficient of variation (sigma), which models the spread of spirometric reference values and adjusts for nonuniform dispersion; and (3) skewness (lambda), which models the departure of variables from normality using a box-Cox transformation (2, 7).
Although promising, because the LMS-LLN5 is based only on a statistical definition of “normal,” its clinical validity is uncertain, particularly in populations with a high prevalence of risk factors for COPD. For example, if a reference population of healthy never-smokers includes persons with a high frequency of exposures to air pollution, environmental tobacco smoke, or high-risk occupation, the LMS-LLN5 may be lower than the threshold that defines “biologically normal.” (8–10) We therefore propose that the LMS-LLN should be assessed empirically by evaluating the associations between successively higher thresholds and clinically meaningful outcomes, such as mortality and respiratory symptoms (5, 11–13).
In the present study, using data from a nationally representative sample of persons aged 40–80 years, we stratified the FEV1/FVC according to the LLN, set at successively higher thresholds, and calculated on the basis of LMS-derived Z scores (2, 7). We then evaluated whether these LMS-derived thresholds identified persons at an increased risk of death and prevalence of respiratory symptoms. As a secondary aim, we also compared prevalence rates for COPD, based on GOLD, ATS/ERS, and LMS-derived thresholds.
We used data from the Third National Health and Nutrition Examination Survey (NHANES) III, a nationally representative sample of community-living Americans assembled in 1988–1994, with mortality surveillance through December 31, 2000 (14, 15). For the present study, our source population included 3,502 participants, aged 40–80 years, who were white, had no self-reported asthma, and completed at least two ATS-acceptable spirometric maneuvers (14). As per current ATS recommendations, we did not exclude participants based on spirometric reproducibility criteria (16). We selected an age range of 40 years or older because COPD and its related mortality are unusual in younger persons (17). Our study population was limited to whites, as spirometric reference values for the LMS method are currently unavailable for minority groups (2, 7).
As described elsewhere, NHANES III recorded the presence of respiratory symptoms in the prior 12 months, defined as chronic cough or sputum production lasting for at least 3 consecutive months (i.e., “chronic bronchitis”), dyspnea on exertion, or wheezing or “whistling in the chest.” (14) Other clinical data included age, sex, ethnicity, height, body mass index (BMI; weight divided by height-squared, expressed as kg/cm2), self-reported chronic conditions and health status, as well as COPD risk factors, such as smoking history, urban residence (i.e., surrogate for air pollution), and high-risk occupation (i.e., based on exposure to airborne dust) (10, 14).
NHANES III also recorded all-cause mortality, which was ascertained from a public-use–linked mortality file that contains information based on the National Death Index, with follow-up through December 31, 2000 (15). Vital status was available on all participants (15).
NHANES III used a customized dry-rolling seal spirometer (14). After calibration, each participant performed five to eight FVC maneuvers, with the goal of meeting ATS criteria (14). For each study participant, as per current guidelines, the measured FEV1/FVC was calculated from the largest set of FEV1 and FVC values that were recorded in any of the spirometric maneuvers meeting ATS acceptability criteria (8, 16).
Based on the measured value, an LMS-derived Z score for the FEV1/FVC was then calculated for each participant as follows (2, 7): ([measured FEV1/FVC ÷ median FEV1/FVC]lambda − 1) ÷ (lambda × sigma). The prediction equations from Stanojevic and colleagues (2, 7) were used to calculate values for the median, lambda, and skewness, and the cubic splines for age were retrieved from tables at www.growinglungs.org.uk (7). These tables are based on four pooled reference samples, with ages ranging from 4 to 80 years. For the age group of 40–80 years, the focus of our study population, the reference sample was exclusively drawn from NHANES III (2, 7). In the United States, NHANES III is preferred for establishing the reference range, because it is the only available, nationally representative sample of the U.S. population that included spirometry (1, 2, 14). Based on Z scores, we then stratified our study population to successively higher thresholds for the LMS-derived LLN, namely at the 5th, 10th, 15th, 20th, and 25th percentiles (LLN5, LLN10, etc.). For example, Z scores that were less than −1.64 corresponded to an FEV1/FVC less than LMS-LLN5 (2, 7). We increased the criterion for “abnormal” to as high as the 25th percentile, identifying nearly 40% of our study participants, because prior work has shown that the prevalence of COPD among older NHANES III participants may be as high as 34.5% when using a GOLD fixed ratio of 0.70 (5).
We first summarized baseline characteristics as means of values accompanied by standard deviations or as counts accompanied by percentages, and we stratified the results according to age: 40–64 years and 65–80 years, denoting middle age and older age, respectively. Differences between middle age and older age were evaluated, using Chi-square for categorical variables or a (two-tailed) t test for continuous variables, with a P value less than 0.05 (two-sided) denoting statistical significance.
We next evaluated the association between the LMS-defined FEV1/FVC strata and death, using a single Cox regression model adjusted for age, height, sex, ethnicity, smoking history, BMI, number of chronic conditions, and health status. The LMS-defined FEV1/FVC strata were treated as nominal categories, with the reference group including participants with an FEV1/FVC of at least LMS-LLN25. Goodness of fit was assessed by model-fitting procedures and by the analysis of residuals. The proportional hazards assumption was tested by using interaction terms for the time-to-event outcome and each variable in the multivariable model. If significant at the P less than 0.05 level, after adjusting for the multiplicity of comparisons, these interaction terms were retained in the final model. Higher-order effects were tested for the continuous covariates, and were included in the final model if they met the forward selection criterion of P less than 0.20 (18). Scaled score residuals for each variable were calculated and plotted. In sensitivity analyses, observations with large residuals were removed from the data set; their removal made little change in reported results.
Similarly, we evaluated the associations between the LMS-defined FEV1/FVC strata and the presence of respiratory symptoms by calculating odds ratios (ORs) using a single logistic regression model. We subsequently stratified the results according to middle age and older age, respectively. To enhance clinical interpretability, we also evaluated the associations between the LMS-defined FEV1/FVC strata and the presence of respiratory symptoms limited to chronic bronchitis or exertional dyspnea. These symptoms represent the most frequent indications for therapeutic interventions in COPD (3).
Finally, we compared prevalence rates for COPD according to ATS/ERS, GOLD, and LMS-derived thresholds (1–3). This included determining prevalence rates for “misidentified” COPD that, based on the above risk analysis, was ultimately defined as participants who had an FEV1/FVC less than ATS/ERS-LLN5 or less than 0.70, but also had an FEV1/FVC of at least the value of LMS-LLN5.
SUDAAN version 10 (Research Triangle Institute, Research Triangle Park, NC) was used to estimate hazard ratios (HR; from Cox proportional hazards regression) and ORs (from logistic regression), with a P value less than 0.05 (two-sided) denoting statistical significance (19).
Table 1 shows the baseline characteristics of the study population, stratified by age. Overall, participants had a mean age of nearly 61 years, with about half being females. Compared with participants who were middle aged (40–64 yr), those who were older aged (65–80 yr) had less education and were less likely to be current smokers, but more likely to have fair-to-poor health status, chronic conditions, or respiratory symptoms. Over a follow-up period ranging between 6 and 12 years, 663 (18.9%) participants died, yielding an overall mortality rate of 22.5 per 1,000 person-years (95% confidence interval [CI], 20.8–24.2). For the middle-aged and older-aged groups, the mortality rates were 7.9 per 1,000 person-years (95% CI, 6.7–9.3) and 45.5 per 1,000 person-years (95% CI, 41.7–49.6), respectively.
Characteristic | All, 40–80 Yr (n = 3,502) | Middle Age, 40–64 Yr (n = 2,005) | Older Age, 65–80 Yr (n = 1,497) |
|---|---|---|---|
| Age, mean (SD), yr | 60.7 (12.0) | 51.9 (7.4) | 72.5 (4.5)* |
| Females, n (%) | 1,827 (52.2) | 1,051 (52.4) | 776 (51.8) |
| Education, mean (SD), yr | 12.2 (3.1) | 12.7 (2.8) | 11.4 (3.2)* |
| Urban residence†, n (%) | 1,325 (37.8) | 801 (40.0) | 524 (35.0) |
| High-risk occupation‡, n (%) | 491 (14.5) | 267 (13.7) | 224 (15.7) |
| Fair-to-poor health status, n (%) | 671 (19.2) | 301 (15.0) | 370 (24.8)* |
| Smoking status, n (%) * | |||
| Never | 1,429 (40.8) | 772 (38.5) | 657 (43.9) |
| Former | 1,337 (38.2) | 693 (34.6) | 644 (43.0) |
| Current | 736 (21.0) | 540 (26.9) | 196 (13.1) |
| Chronic conditions§, mean (SD), n | 0.69 (0.90) | 0.50 (0.75) | 0.94 (1.00)* |
| Respiratory symptoms¶, n (%) | 1,451 (41.5) | 757 (37.8) | 694 (46.5)* |
Table 2 shows the HR for all-cause mortality and ORs for having respiratory symptoms, stratified according to the LMS-LLN for the FEV1/FVC, among participants aged 40–80 years. In comparison to the reference group (i.e., FEV1/FVC ≥ LMS-LLN25), the risk of death was significantly elevated only among participants who had an FEV1/FVC of less than LMS-LLN5, with an adjusted HR of 1.68 (95% CI, 1.34–2.12). Similarly, the prevalence of respiratory symptoms was significantly elevated only among participants with an FEV1/FVC of less than LMS-LLN5, with an adjusted OR of 2.46 (95% CI, 2.01–3.02). These results did not change appreciably when respiratory symptoms were defined only as chronic bronchitis or exertional dyspnea (data available upon request). Comparable associations of FEV1/FVC strata with all-cause mortality and respiratory symptoms were also observed for participants aged 40–64 years (Table 3) and those aged 65–80 years (Table 4). In addition, the distribution of participants across the FEV1/FVC strata was comparable in each of the two age groups (Tables 3 and 4).
No. of Participants n (%) | No. of Deaths among Participants n (%) | No. of Participants with Respiratory Symptoms n (%) | HR for Mortality (95% CI) | OR for Respiratory Symptoms (95% CI) | |||||
|---|---|---|---|---|---|---|---|---|---|
| Measured FEV1/FVC | Unadjusted | Adjusted* | Unadjusted | Adjusted† | |||||
| All-cause mortality risk‡ | |||||||||
| ≥LMS-LLN25 | 2,123 (61.4) | 335 (15.8) | — | 1.00 | — | — | |||
| LMS-LLN20 ≤ to < LLN25 | 191 (5.5) | 38 (19.9) | — | 1.26 (0.85–1.88) | 1.24 (0.85–1.83) | — | — | ||
| LMS-LLN15 ≤ to < LLN20 | 210 (6.1) | 43 (20.5) | — | 1.28 (0.97–1.70) | 1.24 (0.92–1.66) | — | — | ||
| LMS-LLN10 ≤ to < LLN15 | 210 (6.1) | 34 (16.2) | — | 0.99 (0.76–1.29) | 1.13 (0.85–1.49) | — | — | ||
| LMS-LLN5 ≤ to < LLN10 | 248 (7.2) | 63 (25.4) | — | 1.59 (1.22–2.08) | 1.26 (0.95–1.66) | — | — | ||
| <LMS-LLN5 | 477 (13.8) | 138 (28.9) | — | 1.92 (1.52–2.42) | 1.68 (1.34–2.12) | — | — | ||
| Respiratory symptoms§ | |||||||||
| ≥LMS-LLN25 | 2,117 (61.3) | — | 801 (37.8) | — | — | 1.00 | |||
| LMS-LLN20 ≤ to < LLN25 | 191 (5.5) | — | 66 (34.6) | — | — | 0.87 (0.62–1.21) | 0.92 (0.64–1.32) | ||
| LMS-LLN15 ≤ to < LLN20 | 210 (6.1) | — | 84 (40.0) | — | — | 1.10 (0.83–1.44) | 1.20 (0.88–1.64) | ||
| LMS-LLN10 ≤ to < LLN15 | 210 (6.1) | — | 78 (37.1) | — | — | 0.97 (0.74–1.28) | 1.02 (0.74–1.39) | ||
| LMS-LLN5 ≤ to < LLN10 | 248 (7.2) | — | 111 (44.8) | — | — | 1.33 (1.05–1.69) | 1.12 (0.85–1.47) | ||
| <LMS-LLN5 | 477 (13.8) | — | 295 (61.8) | — | — | 2.66 (2.18–3.26) | 2.46 (2.01–3.02) | ||
No. of Participants n (%) | No. of Deaths among Participants n (%) | No. of Participants with Respiratory Symptoms n (%) | HR for Mortality (95% CI) | OR for Respiratory Symptoms (95% CI) | |||||
|---|---|---|---|---|---|---|---|---|---|
| Measured FEV1/FVC | Unadjusted | Adjusted* | Unadjusted | Adjusted† | |||||
| All-cause mortality risk‡ | |||||||||
| ≥LMS-LLN25 | 1,216 (61.5) | 65 (5.4) | — | 1.00 | — | — | |||
| LMS-LLN20 ≤ to < LLN25 | 100 (5.1) | 5 (5.0) | — | 0.91 (0.32–2.65) | 0.80 (0.28–2.36) | — | — | ||
| LMS-LLN15 ≤ to < LLN20 | 113 (5.7) | 10 (8.8) | — | 1.67 (0.94–2.96) | 1.43 (0.74–2.75) | — | — | ||
| LMS-LLN10 ≤ to < LLN15 | 137 (6.9) | 10 (7.3) | — | 1.34 (0.75–2.38) | 1.15 (0.68–1.96) | — | — | ||
| LMS-LLN5 ≤ to < LLN10 | 131 (6.6) | 11 (8.4) | — | 1.51 (0.90–2.53) | 1.34 (0.82–2.16) | — | — | ||
| <LMS-LLN5 | 281 (14.2) | 39 (13.9) | — | 2.66 (1.88–3.75) | 1.85 (1.30–2.63) | — | — | ||
| Respiratory symptoms§ | |||||||||
| ≥LMS-LLN25 | 1,215 (61.5) | — | 412 (33.9) | — | — | 1.00 | |||
| LMS-LLN20 ≤ to < LLN25 | 100 (5.1) | — | 32 (32.0) | — | — | 0.92 (0.59–1.42) | 0.98 (0.61–1.57) | ||
| LMS-LLN15 ≤ to < LLN20 | 113 (5.7) | — | 40 (35.4) | — | — | 1.07 (0.68–1.67) | 1.01 (0.60–1.69) | ||
| LMS-LLN10 ≤ to < LLN15 | 137 (6.9) | — | 47 (34.3) | — | — | 1.02 (0.67–1.54) | 0.92 (0.60–1.41) | ||
| LMS-LLN5 ≤ to < LLN10 | 131 (6.6) | — | 53 (40.5) | — | — | 1.32 (0.95–1.84) | 1.12 (0.79–1.58) | ||
| <LMS-LLN5 | 281 (14.2) | — | 165 (58.7) | — | — | 2.77 (2.00–3.84) | 2.22 (1.54–3.21) | ||
No. of Participants n (%) | No. of Deaths among Participants n (%) | No. of Participants with Respiratory Symptoms n (%) | HR for Mortality (95% CI) | OR for Respiratory Symptoms (95% CI) | |||||
|---|---|---|---|---|---|---|---|---|---|
| Measured FEV1/FVC | Unadjusted | Adjusted* | Unadjusted | Adjusted† | |||||
| All-cause mortality risk‡ | |||||||||
| ≥LMS-LLN25 | 907 (61.2) | 270 (29.8) | — | 1.00 | — | — | |||
| LMS-LLN20 ≤ to < LLN25 | 91 (6.1) | 33 (36.3) | — | 1.24 (0.84–1.84) | 1.33 (0.88–2.00) | — | — | ||
| LMS-LLN15 ≤ to < LLN20 | 97 (6.6) | 33 (34.0) | — | 1.08 (0.80–1.46) | 1.15 (0.85–1.55) | — | — | ||
| LMS-LLN10 ≤ to < LLN15 | 73 (4.9) | 24 (32.9) | — | 1.08 (0.74–1.56) | 1.13 (0.76–1.68) | — | — | ||
| LMS-LLN5 ≤ to < LLN10 | 117 (7.9) | 52 (44.4) | — | 1.48 (1.14–1.94) | 1.24 (0.90–1.70) | — | — | ||
| <LMS-LLN5 | 196 (13.2) | 99 (50.5) | — | 1.87 (1.42–2.45) | 1.63 (1.22–2.17) | — | — | ||
| Respiratory symptoms§ | |||||||||
| ≥LMS-LLN25 | 902 (61.1) | — | 389 (43.1) | — | — | 1.00 | |||
| LMS-LLN20 ≤ to < LLN25 | 91 (6.2) | — | 34 (37.4) | — | — | 0.79 (0.48–1.29) | 0.89 (0.51–1.58) | ||
| LMS-LLN15 ≤ to < LLN20 | 97 (6.6) | — | 44 (45.4) | — | — | 1.10 (0.73–1.63) | 1.43 (0.94–2.19) | ||
| LMS-LLN10 ≤ to < LLN15 | 73 (5.0) | — | 31 (42.5) | — | — | 0.97 (0.65–1.46) | 1.15 (0.77–1.72) | ||
| LMS-LLN5 ≤ to < LLN10 | 117 (7.9) | — | 58 (49.6) | — | — | 1.30 (0.82–2.04) | 1.14 (0.74–1.76) | ||
| <LMS-LLN5 | 196 (13.3) | — | 130 (66.3) | — | — | 2.60 (1.92–3.52) | 2.72 (1.91–3.89) | ||
Table 5 shows the prevalence of COPD, as defined by the ATS/ERS-LLN5, GOLD, and LMS-LLN5 thresholds for the FEV1/FVC. The percentage of participants aged 40–80 years who had ATS/ERS-defined COPD was 17.1%, including 15.6% in the middle-aged and 19.2% in the older-aged groups, while the percentage of participants aged 40–80 years who had GOLD-defined COPD was 27.0%, including 19.1% in the middle-aged and 37.7% in the older-aged groups. In contrast, the percentage of participants aged 40–80 years who had LMS-defined COPD was 13.8%, including 14.3% in the middle-age and 13.2% in the older-age groups.
No. (%) of Participants | |||||
|---|---|---|---|---|---|
| FEV1/FVC Threshold | All, 40–80 Yr (n = 3,502) | Middle Age, 40–64 Yr (n = 2,005) | Older-Age, 65–80 Yr (n = 1,497) | ||
| <ATS/ERS-LLN5 | 599 (17.1) | 312 (15.6) | 287 (19.2) | ||
| <0.70 (GOLD) | 947 (27.0) | 382 (19.1) | 565 (37.7) | ||
| <LMS-LLN5 | 484 (13.8) | 286 (14.3) | 198 (13.2) | ||
Table 6 shows the prevalence of misidentified COPD based on an FEV1/FVC of at least LMS-LLN5. ATS/ERS and GOLD thresholds “misidentified” COPD in 19.2 and 48.9% of participants aged 40–80 years, respectively, 8.3 and 25.1% of participants in the middle-aged group, respectively, and 31.0 and 63.2% of participants in the older-aged group, respectively.
No. (%) of Participants* | |||||
|---|---|---|---|---|---|
| All, 40–80 Yr | Middle Age, 40–64 Yr | Older Age, 65–80 Yr | |||
| <ATS/ERS-LLN5 and ≥LMS-LLN5 | 115/599 (19.2) | 26/312 (8.3) | 89/287 (31.0) | ||
| <0.70 (GOLD) and ≥LMS-LLN5 | 463/947 (48.9) | 96/382 (25.1) | 357/565 (63.2) | ||
Among persons aged 40–80 years, we found that an FEV1/FVC threshold for establishing COPD based on the LMS-LLN5 is associated with a statistically significant increase in the risk of death and likelihood of having respiratory symptoms. These results support a definition of COPD that is based on an LMS-LLN5 threshold for the FEV1/FVC.
The LMS method for calculating the LLN has a strong statistical rationale, because it accounts for normal age-related increases in airflow limitation, as well as the variability of the reference data and its departure from normality (2, 7). In support of the LMS method, we found that the distributions of NHANES III participants across the FEV1/FVC strata were comparable for the middle-aged and older-aged groups (Tables 3 and 4).
The results of our study also provide a strong clinical rationale for using an LMS-LLN threshold for the FEV1/FVC at the fifth percentile of the distribution of Z scores (i.e., LMS-LLN5) as a basis for establishing COPD in middle age and older age. In both age groups, only individuals with an FEV1/FVC less than LMS-LLN5 had an increased risk of all-cause mortality and prevalence of respiratory symptoms. All-cause mortality is an objective and definitive health outcome that is resistant to miscoding and has been the primary endpoint in landmark studies of oxygen therapy in COPD (11). In addition, respiratory symptoms are the most distressing feature of COPD, and can lead to disability and increased healthcare use (11, 12).
Our results suggest that diagnostic thresholds published by the ATS/ERS and GOLD commonly “misidentify” COPD, especially in older age. Specifically, we found that 31.0 and 63.2% of study participants aged 65–80 years were misidentified as having COPD, because they had an FEV1/FVC less than ATS/ERS-LLN5 or less than 0.70, respectively, but also a “normal” FEV1/FVC of at least LMS-LLN5. In comparison, participants aged 40–64 years had misidentification rates for COPD of 8.3 and 25.1% for the ATS/ERS and GOLD thresholds, respectively. These high rates of misidentification are likely due to methodological limitations of the ATS/ERS and GOLD approaches. For example, because the ATS/ERS threshold does not account for the variability of the reference data, including departure from normality (especially at the extremes of age) (2), it will overestimate the prevalence of COPD with advancing age. Similarly, because the GOLD threshold cannot distinguish normal age-related increases in airflow limitation from clinically significant pathology (1, 2), it will also overestimate the prevalence of COPD with advancing age.
Our results have important implications for clinical practice. Because a threshold based on the LMS-LLN5 is less likely to overdiagnose COPD, it could lead to a more targeted use of COPD-specific pharmacotherapies and, hence, a reduced frequency of medication-related adverse events (20, 21). Nonetheless, because pulmonary function, like many clinical phenomena, occurs along a continuum, we would suggest that the LMS-LLN5 should serve as a guide rather than a rigid “yes/no” threshold for the diagnosis of COPD (22). Individuals with an FEV1/FVC that lies just above the LMS-LLN5, for example, may still have COPD, particularly if they have significant risk factors for COPD and the likelihood of an alternative diagnosis is low. Importantly, our results may also be applicable to other respiratory diseases that are characterized by airflow limitation, such as asthma, although this will need to be formally evaluated in future studies.
We recognize potential limitations to our study. First, because cause of death in NHANES III was based only on information from death certificates (14, 15), we evaluated all-cause mortality as an outcome rather than COPD-specific mortality. Prior work has demonstrated that COPD is commonly underreported as a cause of death, even among patients with symptomatic COPD (23). Furthermore, COPD increases the risk of death from cardiovascular disease and lung cancer, and the number of deaths from these causes is much greater than those from respiratory disease among patients with COPD (23–25). In the Lung Health Study, for example, 81.8% of deaths in participants with COPD were due to cardiovascular disease or lung cancer (26). Nonetheless, our findings should be validated in cohorts that include adjudicated data on cause of death.
Second, we evaluated a diagnostic threshold based on the presence of respiratory symptoms, which are not necessarily specific to COPD (11, 12). Nonetheless, respiratory symptoms in COPD can be highly disabling and resource intensive (11, 12). For these reasons, and because there was a threshold relationship between the LMS-defined FEV1/FVC strata and respiratory symptoms, including those specific to chronic bronchitis and dyspnea, we conclude that it is valid to base a spirometric definition of COPD on these clinical features.
Third, because spirometry in NHANES III was not specifically obtained after a bronchodilator, we could not assess reversibility in airflow limitation, a recommended criterion for defining COPD (1). It is unlikely, however, that the absence of information on “reversibility” had a meaningful effect on our results, because persons with self-reported asthma were excluded from our analytical sample, and because prior work has shown that bronchodilator reversibility is neither a sufficient criterion to exclude COPD nor an independent predictor of mortality (27, 28). Nonetheless, we acknowledge that our study population may have included participants with asthma, either as a sole form of obstructive airways disease or concurrent with COPD.
Fourth, because specific LMS-derived Z scores for the FEV1/FVC have not yet been published for minorities (i.e., African Americans and Mexican Americans) (2, 7), our results are limited to a white population. Prior work has demonstrated racial differences in pulmonary function, in general, and COPD-related risk factors and susceptibility, in particular (29).
Fifth, because the sample size for each of the LMS-derived FEV1/FVC strata was modest, our analysis may have been underpowered to detect clinically meaningful effects at strata above the LLN5.
Finally, because a separate validation set was not included in the current analysis, our results will need to be replicated in other, ideally larger, cohorts that include population groups other than whites, so that the validity and generalizability of the LMS-LLN5 as a diagnostic threshold for COPD can be more firmly established.
In conclusion, among white persons aged 40–80 years, an FEV1/FVC threshold at the LMS-LLN5 conferred both an increase in the risk of death and likelihood of having respiratory symptoms. Although further validation is needed in other population groups, including minority representation, these results nonetheless provide strong support for the use of the LMS-LLN5 as a threshold for the diagnosis of COPD.
| 1. | Pellegrino R, Viegi G, Brusasco V, Crapo RO, Burgos F, Casaburi R, Coates A, van der Grinten CPM, Gustafsson P, Hankinson J, et al. Interpretative strategies for lung function tests. Eur Respir J 2005;26:948–968. |
| 2. | Stanojevic S, Wade A, Stocks J, Hankinson J, Coates AL, Pan H, Rosenthal M, Corey M, Lebecque P, Cole TJ. Reference ranges for spirometry across all ages. Am J Respir Crit Care Med 2008;177:253–260. |
| 3. | Rabe KF, Hurd S, Anzueto A, Barnes PJ, Buist SA, Calverley P, Fukuchi Y, Jenkins C, Rodriguez-Roisin R, van Weel C, et al.; Global Initiative for Chronic Obstructive Lung Disease. Global strategy for the diagnosis, management, and prevention of chronic obstructive pulmonary disease: GOLD executive summary. Am J Respir Crit Care Med 2007;176:532–555. |
| 4. | Hansen JF, Sun X-G, Wasserman K. Spirometric criteria for airway obstruction. Chest 2007;131:349–355. |
| 5. | Fragoso CAV, Concato J, McAvay G, Van Ness PH, Rochester CL, Yaggi HK, Gill TM. Defining chronic obstructive pulmonary disease in older persons. Respir Med 2009;103:1468–1476. |
| 6. | Hankinson JL, Odencrantz JR, Fedan KB. Spirometric reference values from a sample of the general US population. Am J Respir Crit Care Med 1999;159:179–187. |
| 7. | Joint website of Great Ormond Street Hospital for Children NHS Trust (GOSH) and UCL Institute of Child Health (ICH) [Internet]. Spirometry. London: GOSH and ICH [accessed 2009 April 24]. Available from: http://www.ich.ucl.ac.uk/ich/academicunits/growinglungs/Homepage |
| 8. | American Thoracic Society. Lung function testing: selection of reference values and interpretative strategies. American Thoracic Society. Am Rev Respir Dis 1991;144:1202–1218. |
| 9. | Pirkle JL, Bernert JT, Caudill SP, Sosnoff CS, Pechacek TF. Trends in the exposure of nonsmokers in the US population to secondhand smoke: 1988–2002. Environ Health Perspect 2006;114:853–858. |
| 10. | Celli BR, Halbert RJ, Nordyke RJ, Schau B. Airway obstruction in never smokers: results from the third national health and nutrition examination survey. Am J Med 2005;118:1364–1372. |
| 11. | Gross NJ. Chronic obstructive pulmonary disease outcome measurements. Proc Am Thorac Soc 2005;2:267–271. |
| 12. | Cherry DK, Burt CW, Woodwell DA. National ambulatory medical care survey: 1999 summary. Advanced Data from Vital and Health Statistics; no 322. Hyattsville, MD: National Center for Health Statistics. 2001. |
| 13. | Mannino DM, Buist AS, Vollmer WM. Chronic obstructive pulmonary disease in the older adult: what defines abnormal lung function? Thorax 2007;62:237–241. |
| 14. | US Department of Health and Human Services. Third National Health and Nutrition Examination Survey, 1988–94, NHANES III laboratory data file (CD-ROM); Hyattsville, MD: National Center for Health Statistics, Centers for Disease Control and Prevention; 1996. Available from National Technical Information Service, Springfield, VA. Public use data file documentation no. 76,200. |
| 15. | Wheatcroft G, Cox CS, Lochner KA. Comparative analysis of the NHANES III public-use and restricted-use linked mortality files [Internet]. Hyattsville, MD: National Center for Health Statistics [accessed January 12, 2010.]. Available from: http://www.cdc.gov/nchs/data_access/data_linkage/mortality/nhanes3_linkage.htm |
| 16. | Miller MR, Hankinson J, Brusasco V, Burgos F, Casaburi R, Coates A, Crapo RO, Enright P, van der Grinten CPM, Gustafsson P, et al. Standardisation of spirometry. Eur Respir J 2005;26:319–338. |
| 17. | Brown DW, Croft JB, Greenlund KJ, Giles WH. Deaths from chronic obstructive pulmonary disease: United States, 2000–2005. MMWR Morb Mortal Wkly Rep 2008;57:1229–1232. |
| 18. | Peduzzi P, Concato J, Feinstein AR, Holford TR. Importance of events per independent variable in proportional hazards regression analysis. II. Accuracy and precision of regression estimates. J Clin Epidemiol 1995;48:1503–1510. |
| 19. | Research Triangle Institute. SUDAAN language manual, release 9.0. Research Triangle Park, NC: Research Triangle Institute; 2004. |
| 20. | Singh S, Loke YK, Furberg CD. Inhaled anticholinergics and risk of major adverse cardiovascular events in patients with chronic obstructive pulmonary disease: a systematic review and meta-analysis. JAMA 2008;300:1439–1450. |
| 21. | Drummond MB, Dasenbrook EC, Pitz MW, Murphy DJ, Fan E. Inhaled corticosteroids in patients with stable chronic obstructive pulmonary disease: a systematic review and meta-analysis. JAMA 2008;300:2407–2416. |
| 22. | Feinstein AR. The inadequacy of binary models for the clinical reality of three-zone diagnostic decisions. J Clin Epidemiol 1990;43:109–113. |
| 23. | Huiart L, Ernst P, Suissa S. Cardiovascular morbidity and mortality in COPD. Chest 2005;128:2640–2646. |
| 24. | Nishimura K, Tsukino M. Clinical course and prognosis of patients with chronic obstructive pulmonary disease. Curr Opin Pulm Med 2000;6:127–132. |
| 25. | Anthonisen NR, Skeans MA, Wise RA, Manfreda J, Kanner RE, Connett JE, and the Lung Health Study Research Group. The effects of a smoking cessation intervention on 14.5-year mortality. Ann Intern Med 2005;142:233–239. |
| 26. | Anthonisen NR, Connett JE, Enright PL, Manfreda J. Hospitalizations and mortality in the lung health study. Am J Respir Crit Care Med 2002;166:333–339. |
| 27. | Vestbo J, Hansen EF. Airway hyperresponsiveness and COPD mortality. Thorax 2001;56:11–14. |
| 28. | Hansen EF, Vestbo J. Bronchodilator reversibility in COPD. Eur Respir J 2005;26:6–7. |
| 29. | Dransfield MT, Bailey WC. COPD: racial disparities in susceptibility, treatment, and outcomes. Clin Chest Med 2006;27:463–471. |
