Race, Lung Function, and Long-Term Mortality in the National Health and Nutrition Examination Survey III

To the Editor:

Interpretation of lung function relies on comparison with values measured in healthy reference populations, using equations that incorporate age, height, and sex (1). Traditionally, such equations have also been stratified by race, based on the observation that Black populations have consistently lower lung function than White populations (2, 3). A key question is whether such race-specific reference equations inappropriately normalize the lower lung function observed in Black populations and therefore underestimate the risk of poor clinical outcomes from lower lung function.

Methods

The NHANES III (National Health and Nutrition Examination Survey III) was conducted from 1988 to 1994 (4) and mortality assessed through December 31, 2015 (5). Individuals who self-identified as African American or Caucasian, with reproducible spirometry at baseline, complete covariate data, and linked mortality data, were included for a final data set of 12,770 individuals. The Global Lung Initiative (GLI) reference equations (6) were used to compare FEV₁ and FVC z-scores and percent predicted calculated using race-specific (African American, Caucasian) equations to values calculated using the multiracial, multiethnic “other” category, which represents an average of four race/ethnic groups: African American, Caucasian, North Asian, and South Asian. Statistical analysis was performed using STATA (version 15.1). Survey weighting, sampling units, and strata were used to account for the complex survey methods (4, 5). Survival was modeled using Cox proportional hazards, and covariates included race, age, sex, smoking status (never, former, or current), pack-years, and household income–poverty ratio. Interaction terms were created between race and spirometric indices to test effect modification. Marginal predicted survival was calculated with binning of calculated FEV₁ in units of 0.5 z-score, stratified by race. For illustration, predicted survival time was calculated using these marginal equations for a 55-year-old, female, never-smoker at twice the federal income–poverty ratio.

Results

Of 8,798 White participants, 2,884 died and of 3,972 Black participants, 1,121 died during the follow-up period. Age ranged from 18 to 79 years. Using survey weights, Black participants were younger (mean age 39 vs. 43 yr), were more likely female (55% vs. 51%), and were more active smokers (33% vs. 29%), with fewer pack-years (mean 15 vs. 23) among ever-smokers and lower household income to poverty ratio (2.1 vs. 3.3) compared with White participants.

On average, Black participants had lower FEV₁ (mean 2.96 vs. 3.31 L) and FVC (mean 3.65 vs. 4.21 L) compared with White participants. Race-specific equations resulted in higher FEV₁ z-score and percent predicted values for Black participants than a multiracial approach with mean increase 0.6 ± 0.1 z-score and 7.6 ± 1% predicted. Conversely, race-specific equations resulted in lower values for White participants with mean z-score change −0.5 ± 0.1 and percent predicted change −7.0 ± 1.2%.

Higher FEV₁ and FVC, whether normalized by race or not, were strongly associated with reduced mortality in multivariable models, with each z-score improvement associated with an ∼20% reduced hazard of death (Table 1). For a given age, sex, and income, predicted survival for FEV₁ z-scores based on multiracial equations was nearly identical between Black and White populations with no significant interaction between race and lung function (P > 0.05). Race-specific equations yield consistently lower predicted survival for Black participants at given FEV₁ z-scores (Figure 1). Furthermore, in multivariable models, the relationship between Black race and lower survival substantially diminished with use of multiracial equations (Table 1). Results were similar for models of FVC z-score (Table 1) and FEV₁ and FVC percent predicted (data not shown).

Table 1.

Association between Spirometry and Mortality Comparing Approaches to Interpretation of Lung Function (N = 12,770)

	FEV1						FVC
	Race-Specific*			Multiracial*			Race-Specific*			Multiracial*
	Hazard Ratio	95% CI	P Value	Hazard Ratio	95% CI	P Value	Hazard Ratio	95% CI	P Value	Hazard Ratio	95% CI	P Value
z-score †	0.79	0.76–0.82	<0.01	0.80	0.77–0.83	<0.01	0.80	0.76–0.84	<0.01	0.82	0.78–0.85	<0.01
Black race	1.27	1.13–1.43	<0.01	1.01	0.90–1.13	0.85	1.24	1.10–1.39	<0.01	0.98	0.87–1.10	0.69
Age, yr	1.10	1.10–1.10	<0.01	1.10	1.09–1.10	<0.01	1.10	1.10–1.11	<0.01	1.10	1.09–1.11	<0.01
Sex, F	0.72	0.66–0.79	<0.01	0.72	0.66–0.79	<0.01	0.72	0.66–0.79	<0.01	0.72	0.66–0.79	<0.01
Former smoker	1.10	0.97–1.25	0.13	1.10	0.97–1.25	0.13	1.14	1.00–1.30	0.04	1.14	1.00–1.30	0.04
Current smoker	1.67	1.47–1.90	<0.01	1.67	1.47–1.90	<0.01	1.80	1.58–2.05	<0.01	1.80	1.58–2.05	<0.01
Pack-years	1.00	1.00–1.01	<0.01	1.00	1.00–1.01	<0.01	1.01	1.00–1.01	<0.01	1.01	1.00–1.01	<0.01
Income to poverty ratio	0.90	0.87–0.92	<0.01	0.90	0.87–0.92	<0.01	0.90	0.87–0.92	<0.01	0.90	0.87–0.92	<0.01

Definition of abbreviations: CI = confidence interval; GLI = Global Lung Initiative.

^*Multivariable models are presented. Race-specific approach includes z-scores created using African American or Caucasian GLI reference equations. Multiracial approach includes z-scores created using GLI other category reference equations.

^† z-scores pertain to FEV₁ or FVC.

Figure 1.

Influence of race-specific compared with multiracial reference equations in interpretation of lung function and the association with survival. Histograms demonstrate the distribution of FEV₁ z-scores applying race-specific compared with multiracial, multiethnic Global Lung Initiative reference equations. The predicted survival for a 55-year-old, female never-smoker living at two times the federal income–poverty ratio is displayed for Black and White participants using each approach.

Discussion

Lung function has been shown repeatedly to be a strong predictor of morbidity and overall mortality in populations with and without diagnosed lung disease. The purpose of using population-based equations for interpreting lung function is to understand where a given individual’s lung function falls along a spectrum of healthy–unhealthy. Traditionally, such equations have incorporated race, normalizing population differences in lung function into what is considered healthy. However, this practice assumes both that race is a biologically relevant classification and that the lower lung function found on average in Black populations is not clinically meaningful. The availability of long-term data on mortality, a hard endpoint, in NHANES III provided a unique opportunity to address whether the association between lung function and mortality differs by race. In a representative sample of the U.S. population, our findings suggest that lower lung function found commonly among Black individuals has the same implications for all-cause mortality as similarly low lung function found in White individuals, extending prior work by Burney and Hooper (7).

If the lower average lung function found in Black individuals compared with White individuals were “normal,” we would expect that applying race-specific reference equations would yield similar risk of mortality across races for a given estimated lung function whereas a multiracial approach would overestimate mortality risk among Black participants. However, we found that a multiracial approach yields similar mortality risk between groups. This suggests that a race-specific approach reinforces a false assumption that lower lung function is “normal” among Black populations and does not have health implications.

Race is a sociopolitical construct and factors that contribute to the observed differences between Black and White individuals are incompletely understood and likely multifactorial (8–10). Using race-specific approaches to interpreting lung function may obscure the true causes of disparities and reduce the likelihood that modifiable risk factors for poor lung function are identified and acted upon. Using a universal, multiracial approach to interpreting lung function will result in many more Black patients being classified as having abnormal lung function, as on average, Black participants estimated z-score changed by 0.6 and percent predicted by 8%. Using a multiracial approach may have both positive and negative consequences for individual patients, including eligibility for jobs, benefits, medications, and life-saving procedures.

Strengths include the study population representative of the United States; application of GLI reference equations, which represents a global population sample; use of GLI “other” category that has been proposed as a universal, multiethnic, multiracial approach; and availability of linked mortality data. Limitations include that this analysis focused on Black race and did not focus on the implications of population-specific reference equations for other groups. Furthermore, the GLI “other” equations that have been proposed as a universal approach have been derived from four population groups that are not fully representative of the global population. If the GLI multiracial reference values are used for interpretation of lung function, further evidence will be needed to determine whether this leads to better health outcomes and decreased health disparity. Findings are limited to overall mortality, and future studies in disease-specific cohorts are needed to address respiratory morbidity and mortality. This work does not explain how lower lung function may result in higher mortality. It is possible that the factors that contribute to lower lung function in this population may be ultimate drivers of mortality or that lower lung function is a proxy for other mortality risks. However, these findings underscore the need to further understand the implications of incorporating race into our interpretation of lung function results in clinical, research, and population health arenas.