Worsened longitudinal visit-to-visit-FEV₁-variation and mortality in WTC exposed FDNY workers: a 23-year landmark analysis

Abstract

Introduction

The World Trade Center (WTC) collapse produced an inhalation disaster. Body Mass index (BMI) is a risk-factor for WTC-related lung injury. Visit-to-visit-forced-expiratory-volume-at-one-second-variation (FEV₁-variation) is a mortality risk-factor in WTC-exposed Fire Department of New York (FDNY) workers. We hypothesized: worsened-FEV₁-variation with greater time post-9/11/2001 will be a mortality risk-factor associated with cross sectional BMI and longitudinal weight-variation.

Methods

Annual WTC-monitoring measured lung function and weight. FEV₁-variation was quantified as individual-specific coefficient of variation (CV) and FEV₁-slope-SEM, among 9410 WTC-exposed workers who were alive at the 1/1/2019 Landmark and had ≥ 3 FEV₁-measurments early (9/12/2001 through 12/31/2009) and later (1/1/2010 through 12/31/2018). Weight-variation quantified as individual-specific weight-CV 9/12/2001 through 12/31/2018. Worsened-FEV₁-variation was defined as later/early FEV₁-variation ratio > 1. The outcome was all-cause mortality 1/1/2019 through 12/31/23. Cox regression assessed the 5-year mortality risk of Worsened-FEV₁-variation. Logistic regression assessed covariates associated with worsened-FEV₁-variation.

Results

Fourty-seven percent (4459/9410) of WTC-exposed FDNY-workers experienced worsened-FEV₁-variation. an exposure that was associated with 40% greater 5-year mortality risk (HR = 1.40; 95% CI 1.07, 1.89) in a Cox model adjusted for landmark, age, FEV₁, BMI, weight-variation, and smoking. Similarly, Worsened-FEV₁-variation defined by FEV₁-slope-SEM was a mortality risk factor (HR = 1.39; 95% CI 1.05, 1.83). Each 5 kg/m² increase in BMI and each 5%CV increased weight-variation was associated with 18% and 17% greater odds of worsened-FEV₁-variation (OR = 1.18; 95% CI 1.06, 1.27 and OR = 1.17; 95% CI 1.07, 1.27 respectively).

Interpretation

Worsened-FEV₁-variation with greater time post-9/11/2001 is a mortality risk-factor associated with obesity and weight-variation. The association of Worsened-FEV₁-variation and weight-variation may reflect concurrent metabolic and pulmonary instability that increase vulnerability to death.

Introduction

Terrorists attacked the World Trade Center (WTC) on September 11, 2001 (9/11) producing a particulate matter (WTC-PM) inhalation disaster. Fire Department of the City of New York (FDNY) rescue/recovery workers experienced markedly different exposure levels based on arrival time. Workers who arrived before noon on 9/11 (High-WTC-exposure) were enveloped in a dense dust cloud with extraordinarily high PM_2.5 concentrations in the mg/m³ range [1]. The intensity of this exposure overwhelmed normal upper-airway filtration mechanisms, allowing highly inflammatory super-coarse particles to penetrate the lower airways [2]. The dust was highly alkaline (pH 9.0–11.0) causing acute chemical injury to aerodigestive mucosa [3]. Those arriving after day 3 were intermittently but chronically exposed to resuspended WTC-PM. The combination of early arrival and prolonged exposure during rescue/recovery put workers at highest risk for non-resolving upper airway inflammation [4]. Initially adequate respiratory protection was rarely worn enhancing WTC-PM dust exposure exacerbating respiratory symptoms [5].

The FDNY WTC-medical monitoring program performed physical exams approximately every 12–18 months. Exams included spirometry, height, and weight. Because of ongoing medical treatment for WTC-related illnesses, there has been minimal longitudinal drop out [6–8]. Exposure and outcome metrics recorded by the FDNY WTC Health Program provide a rich longitudinal dataset ideal for investigating the pathophysiology of WTC-related lung injury. High-WTC-exposure increased the risk of lower respiratory tract injury [9] with incident asthma, COPD, and asthma-COPD overlap [10, 11] and cardiovascular diseases [12, 13]. Cardiometabolic risk factors, particularly body mass index (BMI), are well defined risk-factors for WTC related lung injury defined as incident FEV₁ below the lower limit of normal [14].

As the cohort has aged, mortality risk factors have become a research focus. We have observed that both lower cross-sectional FEV₁ and steeper FEV₁-decline are mortality risk factors in this cohort [15, 16]. Recently, we reported visit-to-visit-FEV₁ variation (FEV₁-variation) is another FEV₁ related mortality risk-factor [17]. We further observed that high WTC exposure and components of metabolic syndrome, including high baseline BMI, were associated with increased odds of high (top quartile) FEV₁-variation. In the current investigation, we test whether FEV₁ variation changes with increasing time since WTC exposure. We hypothesize that worsened-FEV₁-variation over time since 9/11 will be associated with BMI and weight variation and will represent an additional mortality risk factor.

Methods

Study population

The source population consisted of 11,474 WTC exposed FDNY rescue/recovery workers [17]. WTC exposed workers received medical monitoring examinations every 12–18 months. Worker signed informed consent; Montefiore Medical Center/Albert Einstein College of Medicine IRB approval (#07-09-320).

Data collection

Data was maintained in the FDNY WTC Health Program Data Center. WTC-monitoring exams include FEV₁, weight and height measurements. Mortality data for the FDNY cohort were ascertained through FDNY WTC records (9/12/2001-12/31/2023).

Spirometry

Screening spirometry without bronchodilator was performed according to contemporaneous ATS criteria: only high-quality, grade A/B measurements, were included (7). The cohort completed 113,628 WTC-monitoring exams 9/12/2001-12/31/2018 (12.1 ± 2.2 exams per person) with 111,231 (97.9%) being grade A/B. FEV₁ percent predicted values used race neutral GLI Global reference equations [18].

Study design

We conducted a Landmark mortality analysis of WTC-exposed FDNY rescue/recovery workers who were alive 1/1/2019 (Fig. 1), providing common start and equal follow-up times to reduce immortal time bias [19]. To assess FEV₁-variability trajectory, exposure was divided into two prespecified 9-year intervals; early (9/12/2001–12/31/2009) and later (1/12010–12/31/2018). Because estimation of within-person variation by linear regression requires more than three measurements, the analytic cohort of N = 9410 needed ≥3 FEV₁ measurements both early and later. The primary exposure was greater FEV₁-variation between early and later 9-year intervals, the outcome was 5-year all-cause mortality 1/1/2019-12/31/2023 (Fig. 1). Secondary exposures were landmark-BMI and longitudinal weight-variation (2001–2018). Covariates were landmark-FEV₁. Prespecified confounders were smoking (ever/never) and landmark-age. The study followed STROBE guidelines.

Fig. 1.

Landmark analysis study design. A 1/1/2019 landmark analysis of N = 9410 workers who were alive 12/31/2018 provided common start and equal follow-up times to reduce immortal time bias. Early/later FEV₁-variation change was the exposure and 5-year post-1/1/2019 mortality the outcome. Early-Interval FEV₁-variation was derived from N = 47,410 FEV₁ measurements (5.0 ± 1.4 per person). Later-Interval FEV₁-variation was derived from N = 63,784 FEV₁ measurements (6.4 ± 1.7 per person)

Statistical analysis

Visit-to-visit variation has been defined by both coefficient of variation (CV) and standard error of the slope (slope-SEM) derived from linear regression [17, 20–22]. We calculated both FEV₁-CV and FEV₁-slope-SEM. Doubling of FEV₁-variation was used because log₂ transformation improved FEV₁variation kurtosis (61.2 ± 0.05 to 5.8 ± 0.05 for FEV₁-slope-SEM and 11.5 ± 0.05 to 1.4 ± 0.05 for FEV₁-CV). Categorical Worsened-FEV₁-variation was defined as ratio > 1 of later-FEV₁-variation/early-FEV₁-variation. Continuous FEV₁-variation change was defined as log₂ of the later-FEV₁-variation/early-FEV₁-variation ratio. Logistic regression assessed worker characteristics associated with worsened-FEV₁-variation. The association between worsened-FEV₁-variation and all-cause mortality was modeled using Cox regression with equal person-time at risk for all study participants. There were no interactions between worsened-FEV₁-variation and age (p = 0.48) or weight variation (p = 0.98) but there was interaction between continuous worsening-FEV₁-variation and ICS/LABA treatment (p = 0.029).

Five sensitivity analyses evaluated the robustness of the association between worsened-FEV₁-variation and mortality: (1) Modeling FEV₁-variation as defined by either FEV₁-CV or FEV₁-slope-SEM. (2) Assessing worsened-FEV₁-variation in subgroup (N = 6434) without known pulmonary abnormalities. (3) Assessing worsened-FEV₁-variation in subgroup (N = 2935) with need for ICS/LABA treatment, having accelerated-FEV₁-decline or landmark FEV₁<80%. (4) Assessing worsening-FEV₁-variation as a continuous exposure. (5) Evaluating exposure dose response by categorizing FEV₁-variation into tertiles of FEV₁-CV or FEV₁ Slope-SEM. Data analysis was performed using SPSS.v30 and Prism.v10.

Results

Cohort characteristics and FEV₁-variation

The analytic cohort N = 9410 suffered 207 deaths over 46,487 person years yielding 4.45 deaths/1000 person years (Fig. 1). There were 111,231 FEV₁ measurements (11.8 ± 2.2 measurements per person) 9/12/2001–12/31/2018.

Mean FEV₁-variation was 5.87 ± 3.80 CV% in the early interval and 5.51 ± 3.22 CV% in the later interval, corresponding to a mean decrease in variation (improvement) between the two intervals of 0.36 ± 4.4 CV% (paired t test). Between 10th -90th percentile, early-to-later changes in FEV₁-variation ranged from a 4.85 CV% decrease (improvement) to a 3.96 CV% increase (worsened), an 8.8 CV% interdecile range. FEV₁-CV was not associated with FEV₁ measurements per person (β = 0.009 ± 0.016% per FEV₁ measurement p = 0.55).

We categorized individuals according to change in FEV₁-variation: 4459 (47.4%) demonstrated worsened-variation and 4951 (52.6%) improvement (Table 1). Workers with worsened-FEV₁-variation experienced 122 post-landmark deaths (2.73%, 5.5/1000 person-years) while workers with improved-FEV₁-variation experienced 85 post-landmark deaths (1.73%, 3.5/1000 person-years) (Table 1).

Table 1.

Characteristics of workers in Worsened- and Improve- FEV1-variation strata

Variable	FEV1-variation Strata		Unadjusted difference or (95% CI)*
	Worsened ¶ N = 4459 (47.4%)	Improved N = 4951 (52.6%)
Longitudinal FEV1-variation
Early FEV1-variation ml/year ∥	4.08 ± 2.05	7.48 ± 4.27	− 3.39 (− 3.53, − 3.26)
Later FEV1-variation ml/year††	6.82 ± 3.71	4.33 ± 2.08	2.49 (2.36, 2.61)
Later minus early FEV1-variation	2.73 ± 3.07	− 3.15 ± 3.50	5.88 (5.75, 6.01)
Demographics
Age on 9/11/2001	39.0 ± 7.6	38.6 ± 7.6	0.40 (0.095, 0.71)
Ever smoker**	1572 (35.2%)	1682 (33.9%)	1.2 (− 0.6, 3.2)
First post 9/11 -BMI §<30	2840 (63.7%)	3401 (68.7%)	Reference
First post 9/11 -BMI ≥ 30, < 35	1293 (29%)	1275 (25.8%)	1.21 (1.11,1.33)
First post 9/11 BMI ≥ 35	324 (7.3%)	274 (5.5%)	1.42 (1.20,1.68)
First post 9/11 BMI kg/m2	29.3 ± 4.0	28.7 ± 3.7	0.56 (0.41, 0.72)
BMI kg/m2 Closest to 1/1/2019	30.9 ± 5.1	30.1 ± 4.6	0.78 (0.58, 0.98)
Weight-slope lbs/year†	0.61 ± 1.4	0.52 ± 1.3	0.083 (0.028,0.14)
Weight-variation CV%‡	4.55 ± 2.56	4.35 ± 2.32	0.20 (0.10, 0.30)
FEV1 related risk-factors
First post 9/11 FEV1%predicted	99.8 ± 13.1	103.3 ± 15.6	− 3.5(− 4.0,− 2.9)
FEV1-slope ml/year 2001–2018	− 37.0 ± 22.0	− 32.9 ± 23.5	− 4.1 (− 5.1, − 3.2)
First post 9/11 FEV1/FVC < 70%	105 (2.3%)	116 (2.3%)	0.00 (− 0.01,0.01)
FEV1% Closest to 1/1/2019	95.7 ± 15.4	99.7 ± 14.8	− 4.0 (− 4.6, − 3.4)
Mortality
Alive 12/31/2018	4337 (97.3%)	4866 (98.3%)	Reference
Died 1/1/2019-12/31/2023	122 (2.73%)	85 (1.73%)	OR = 1.61 (1.22, 2.13)
Mortality Rate/1000-person years*	5.5	3.5	RR = 1.61 (1.22, 2.13)

^¶Worsened-FEV₁-variation as a categorical variable was defined as ratio > 1 of later-FEV₁-variation/early-FEV₁-variation. ∥ Forced expiratory volumes at one second variation in the interval 9/12/2001-12/31/2009 ^†† Forced expiratory volumes at one second variation of the interval 1/1/2010-12/31/2018 *Unadjusted odds or rate ratio > 1 represents greater risk with Worsened-FEV₁-variation. ^†Weight change (mean time effect on weight) and ^‡Weight-variation (standard error of the mean time effect). ** Reported current or former smoking on any of the 113,606 medical monitoring exams 9/12/2001–1/1/2019. ^§ Body Mass Index kg/m²

Despite having lower (better) FEV₁-variation in the early interval, the worsened-FEV₁-variation stratum had delayed FEV₁-variation increase. Early, FEV₁-variation was 3.39 CV% lower in the worsened-FEV₁-variation stratum than in the improved-FEV₁-variation stratum (95% CI = − 3.53 to − 3.26). The worsened stratum suffered a delayed increase from early to the later (Δ = 2.73 ± 3.07 CV% 95% CI = 2.64–2.82) (Table 1). Later, FEV₁-variation was 2.49 CV% higher in the worsened-FEV₁-variation stratum than in the improved-FEV₁-variation stratum (95% CI = 2.36–2.61).

First-post-9/11 (baseline) characteristics associated with worsened-FEV₁-variation.

We next assessed the association of worsened-FEV₁-variation with first post-9/11 risk factors. Compared with workers whose FEV₁-variation improved, workers suffering worsened-FEV₁-variation were slightly older (0.40 years 95% CI = 0.10–0.71) and had higher baseline-BMI and Landmark-BMI (0.56 kg/m², 95% CI = 0.41–0.72; and 0.78 kg/m² 95% CI = 0.58–0.98; respectively). Compared with workers with BMI < 30, workers with BMI ≥ 30, BMI ≤ 35 had 21% greater odds of worsened-FEV₁-variation (95% CI = 1.11–1.33) and workers with BMI > 35 had 42% higher odds of worsened-FEV₁-variation (95% CI = 1.20–1.68). Workers with worsened-FEV₁-variation had lower baseline FEV₁ (− 3.5% predicted lower; 95% CI=− 4.0 to − 2.9), but the prevalence of baseline FEV₁/FVC ratio < 70% was not different.

Worsened-FEV₁-variation predictors: multivariable models.

Adjusted logistic regression showed similar results as crude odds ratios (Fig. 2). Each 10-year increase in age was associated with 8% higher odds of worsened-FEV₁-variation (95% CI = 1.02–1.14) and each 10% predicted lower baseline FEV₁ was associated with 18% higher odds of worsened-FEV₁-variation (95% CI = 1.14–1.21). Each 5 kg/m² increase in baseline BMI is associated with18% higher odds of worsened-FEV₁-variation (95% CI = 1.12–1.24).

Fig. 2.

Two multivariable logistic regressions with worsened-FEV₁-variation as the outcome. Upper three rows, cross-sectional variables. Older age at study entry, lower first-post-9/11-FEV₁ and greater first-post-9/11-BMI are associated with increased odds of worsened-FEV₁-variation (OR 1.08 per 10 years; 95% CI 1.02–1.14, OR 1.18 per 10% predicted decline, 95% CI 1.14–1.21, OR 1.18 per 5 kg/m²; 95% CI 1.12–1.24, respectively). Lower two rows show an independent model with longitudinal variables. Visit-to-visit weight-variation is associated with increased odds of worsened-FEV₁-variation but weight change is not associated with worsened-FEV₁-variation (OR = 1.17 per 5 CV%; 95% CI = 1.07–1.27 and OR1.03 lbs/year; 95% CI 0.99–1.06, respectively)

We next evaluated if longitudinal risk factors were associated with worsened-FEV₁-variation. The interdecile range for FEV₁-slope was from − 66.3 to − 8.2 ml/year; weight-variation ranged 2.12–7.49 CV% and weight-change ranged from 0.92 lbs./year weight loss to 2.30 lbs./year weight gain. Compared to improved-FEV₁-variation, worsened-FEV₁-variation had greater weight-variation (Δ = 0.20 CV%; 95% CI = 0.10–0.30) and more weight gain (Δ = 0.083 lbs/year; 95% CI = 0.028–0.14) (Table 1). Each 5 CV% increase in weight-variation was associated with 17% higher odds of worsened-FEV₁-variation (95% CI = 1.07–1.27) but weight gain (Fig. 2 bottom rows) was not significantly associated with the odds of worsened-FEV₁-variation (OR = 1.03; 95% CI = 0.99–1.06). In linear regression, each 5% increase in weight-CV was associated with a 0.47 ± 0.93%, p < 0.001 increase of longitudinal FEV₁-CV change .

FEV₁-variation trajectory and mortality hazard.

Like the unadjusted rate ratio (RR = 1.61; 95% CI = 1.22–2.13 Table 1), a crude Cox model demonstrated 60% higher mortality hazard associated with Worsened-FEV₁-variation ( 95% CI = 1.21–2.11 Fig. 3A), Worsened-FEV₁-CV increased 5-year mortality hazard 40% (95% CI = 1.07–1.89) and Worsened-FEV₁-slope-SEM increased 5-year mortality hazard 39% (95% CI = 1.05–1.83 Fig. 3A bottom row) in separate fully adjusted cox models.

Fig. 3.

Mortality hazard of longitudinal change in FEV₁-visit-to-visit variation. Panel A Top row, the categorical variable Worsened-FEV₁-CV as time post 9/11 increases is a mortality risk-factor (HR = 1.40; 95% CI 1.07–1.89). Bottom row, Worsened-FEV₁-slope-SEM is also a mortality risk factor (HR = 1.39; 95% CI 1.05–1.83) when substituted for Worsened-FEV₁-CV in an independent cox model. Both cox models adjusted for cross-sectional FEV₁ per 10% predicted decrement closest to but before 1/1/2019, longitudinal weight variation, smoking, and age. Panel B The continuous variable Worsening-FEV₁-CV (log₂ ratio late/early FEV₁-variation) is a mortality risk-factor (HR = 1.29; 95% CI 1.07–1.55 per doubling of early/late ratio) in a restricted subgroup (N = 6434) without lung function abnormalities. This subgroup excluded participants with asthma and/or COPD, accelerated-FEV₁-decline, vital capacity < 80% predicted closest to but preceding the 1/1/2019 landmark. Panel C Worsening-FEV₁-CV is not mortality risk-factor (HR = 0.97; 95% CI 0.82–1.16) in a restricted subgroup (N = 2935) enriched of lung function abnormities. The subgroup only included participants with asthma or COPD or accelerated-FEV₁-decline or vital capacity < 80% predicted closest to but preceding the 1/1/2019 landmark

Findings were consistent across all sensitivity analyses Fig. 3B and C. To test if FEV₁-variation is a mortality risk factor in workers without lung function abnormalities we developed a restricted subgroup (N = 6434 Fig. 3B) that excluded workers treated for Asthma/COPD with ICS/LABA for 2 or more years (N = 1854), workers with accelerated-FEV₁-decline (N = 1164) and workers with landmark FVC < 80% (N = 1406). In this restricted subgroup, FEV₁-variation as a continuous variable is a mortality risk factor (HR = 1.29 per CV doubling; 95% CI = 1.07–1.55). Surprisingly FEV₁-variation was not a mortality risk factor (HR = 0.97 95% CI = 0.82–1.16) in the subgroup (N = 2935 Fig. 3C) enriched for lung function abnormalities.

We observed a trend toward interaction between worsened FEV1-variation and early ICS/LABA treatment (p = 0.098), and a significant interaction when worsening-FEV1-variation was modeled continuously (p = 0.029), suggesting effect modification of the association between improved-FEV₁-variation and mortality by ICS/LABA treatment. To determine whether effect modification was a direct ICS/LABA treatment effect, we performed multivariable logistic regression and found no association between early ICS/LABA treatment and improved-FEV₁-variation (OR = 0.91, 95% CI 0.80–1.04; p = 0.165) after adjustment for age, smoking status, BMI, first post-9/11-FEV₁, and early FEV₁-slope suggesting ICS/LABA does not appear to directly improve the variation phenotype.

To better characterize the dose–response relationship between change in FEV₁ variation and mortality, we categorized change in both FEV₁-CV and FEV₁ slope-SEM into tertiles representing worsened, stable, and improved variation. Mean (SD) %CV visit-to-visit-FEV₁-variation increased 3.68 ± 3.23% in the worsened stratum, decreased 0.20 ± 0.73% in the stable stratum, and decreased 4.55 ± 3.74% in the improved stratum. Compared with stable FEV₁-CV (Fig. 4A), worsened FEV₁-CV was associated with a 49% increased mortality hazard (HR = 1.49, 95% CI 1.06–2.09; p = 0.021), whereas improved FEV₁-CV was not associated with mortality (HR = 1.14, 95% CI 0.79–1.65; p = 0.472). Similarly, when visit-to-visit-FEV₁-variation was defined by SEM-slope, variation increased 17.7 ± 24.8 mL/year in the worsened stratum, declined 6.4 ± 4.4 mL/year in the stable stratum, and declined 37.7 ± 30.9 mL/year in the improved stratum. Compared with stable FEV₁ slope-SEM (Fig. 4B), worsened FEV₁ slope-SEM was associated with a 50% increased mortality hazard (HR = 1.50, 95% CI 1.08–2.09; p = 0.016), whereas improved FEV₁ slope-SEM was not associated with mortality (HR = 1.03, 95% CI 0.72–1.47; p = 0.869).

Fig. 4.

Mortality hazard of longitudinal change in categorical tertials (1) Improved (Green). (2) Stable (Blue) and (3) Worsened FEV₁-visit-to-visit variation (Red). Panel A Compared to stable-CV, worsened-CV had a 49% increased mortality (HR = 1.49, 95% CI = 1.06–2.09 p = 0.021) but improved FEV₁-CV was not significantly different from stable-FEV₁-CV (HR = 1.14, 95% CI = 0.79–1.65 p = 0.472) Panel B Compared to stable FEV₁-slope SEM worsened FEV₁-slope-SEM has a 50% increased mortality (HR = 1.50 95% CI = 1.08–2.09 p = 0.016) but improved FEV₁-slope-SEM was not significantly different from stable-FEV₁-slope-SEM (HR = 1.03, 95% CI = 0.72–1.47, p = 0.869)

Discussion

The trajectory of visit-to-visit FEV₁-variation (FEV₁-variation) over time following WTC exposure was heterogeneous. Approximately half of WTC-exposed FDNY workers experienced worsened-FEV₁-variation, defined by increased variation from the pre-January 1, 2010 to post-January 1, 2010 exposure windows. Pulmonary function instability emerging nine years after WTC exposure was associated with increased mortality risk and co-occurred with baseline obesity, greater visit-to-visit weight-variation (weight-variation), and steeper FEV₁ decline. Despite their collinearity, worsened-FEV₁-variation, weight-variation, and landmark adiposity remained independent predictors of mortality. The heterogeneity of the trajectory of FEV₁-variation parallels the heterogeneity of trajectory of respiratory symptoms after initiation of ICS/LABA treatment for asthma and or COPD [23, 24]. Multi-system instability, reflected by increased FEV₁-variation, adiposity, and weight-variation, may reduce physiologic reserve, and increase vulnerability to adverse health outcomes, including higher all-cause mortality. Further research is needed to determine whether instability in other treatable metabolic syndrome trait(s) are also associated with worsened-FEV₁-variation.

The association between FEV₁-variation and mortality is robust. In prior work we measured FEV₁ variation using a model-based estimate of FEV₁-slope standard error of the mean (FEV₁-slope-SEM) [17], whereas this study used coefficient of variation, a common measure of visit-to-visit fluctuation in cardio-metabolic risk factors [20, 21, 22], as the primary exposure and FEV₁-slope-SEM in sensitivity analysis. Each definition of visit-to-visit variation has advantages and disadvantages. FEV₁-CV does not use the number of FEV₁ measurements in its derivation but does not separate variation due to individuals’ average FEV₁-slope from visit-to visit-variation. FEV₁-slope-SEM explicitly separates variation due to longitudinal decline in FEV₁ from visit-to-visit variation around the average but uses the number of FEV₁ measurements per person in its derivation so can be biased by differences in the number of longitudinal FEV₁ measurements per person. The association of worsened-FEV₁-variation with mortality is replicated using both FEV₁-CV and FEV₁-slope-SEM definitions of FEV₁-variation. In a sensitivity analysis excluding workers with asthma and or COPD [23, 24], accelerated-FEV₁-decline [10, 16] and low vital capacity (FVC < 80% predicted), we observed worsened-FEV₁-variation remained significantly associated with mortality.

In a second sensitivity analysis of workers enriched for pulmonary abnormalities, worsened-FEV₁-variation was not significantly associated with mortality. We observed the association between worsened FEV₁ variation and mortality was absent among ICS/LABA treated workers but remained significant in ICS/LABA untreated workers, demonstrating effect modification. However, ICS/LABA therapy was not associated with improvement in FEV₁ variation in multivariable logistic regression using improved FEV₁-variation as the outcome. These findings do not support a simple mechanism in which ICS/LABA reduces mortality through improvement in FEV₁ variation. Absence, in this cohort, of an association between worsened-FEV₁-variation and mortality in asthma and/or COPD patients requiring prolonged ICS/LABA treatment argues against these diseases being the predominant explanation for the association between worsened-FEV₁-variation and mortality. Further research is needed to determine whether the observed effect modification reflects a direct treatment effect or residual confounding by indication, whereby ICS/LABA treatment identifies a subgroup with underlying immunologic or clinical characteristics associated with both treatment initiation and reduced mortality risk.

We also investigated the dose response to change in FEV₁-variation dividing the cohort into worsened, stable, and improved strata. Compared to stable-FEV₁-variation over time, worsened-FEV₁-variation has significantly greater mortality in both FEV₁-slope-SEM and FEV₁-CV definitions of FEV₁-variation. There is, however, no difference in mortality between stable and improved-FEV₁-variation stratum demonstrating the most severely impacted 33% is driving the mortality risk. Although continuous exposures often provide greater statistical efficiency, this was not observed in our analyses. For both the CV and slope-SEM definitions, continuous worsening in FEV₁ variation showed lower precision and weaker mortality discrimination than the dichotomous definition. These findings suggest that the relationship between worsening-FEV₁-variation and mortality may not be linear across the full distribution of FEV1-variation change. This pattern is consistent with a threshold effect and may explain why the dichotomous exposure demonstrated greater precision than the continuous exposure.

Worsened-FEV₁-variation could reflect progressive airway instability driven by recurrent systemic inflammation associated with metabolic syndrome [25, 26]. To test this possibility, we assessed the association of worsened-FEV₁-variation with both early obesity and with longitudinal weight-variation [22]. Both baseline obesity and greater longitudinal weight-variation were associated with greater odds of worsened-FEV₁-variation. Despite being correlated, worsened-FEV₁-variation and longitudinal weight-variation do not interact and remain significant mortality risk-factors when combined in a fully adjusted cox mortality model. Pulmonary and metabolic instability each carry prognostic information beyond their shared correlation. The associations between worsened-FEV₁-variation, greater adiposity, and increased 5-year all-cause mortality suggests that rising FEV₁-variation reflects underlying multisystem pathophysiologic processes that increase susceptibility to death. Further research is needed to test if additional components of metabolic syndrome are risk-factors for worsened-FEV₁-variation.

This analysis has limitations. Requiring three spirometries in two 9-year exposure windows likely introduced survivor bias but this is a healthy worker cohort with low mortality from 9/12/2001 to the landmark date [15, 16]. Unmeasured confounders associated with both worsened-FEV₁-variation and mortality may bias mortality models. The categorical exposure worsened-FEV₁-variation is necessarily arbitrary, but the continuous exposure of later FEV₁-variation is also a mortality risk suggesting little information is lost by stratification. Finally, given the unique nature of WTC-exposure in this healthy worker cohort, these observations require replication in independent occupational, community and disease specific cohorts.

In conclusion, FEV₁-variation is an understudied mortality risk-factor reflecting biologically important end-organ instability analogous to blood-pressure and/or weight-variation [20–22]. We interpret FEV₁ variation not as measurement noise but as a physiologic signal reflecting instability of pulmonary function, analogous to variability in blood pressure or body weight. Measures of longitudinal risk-factor instability contribute prognostic information beyond what is contained in cross-sectional and risk factor-slope measurements. An integrated systems approach that explicitly models risk-factor variability as an exposure may improve risk stratification in longitudinal cohorts. Regular assessment of FEV₁-variability may identify workers at increased risk, allowing early interventions to limit disease progression and improve health outcomes. Further research with analytic approaches capable of incorporating time-dependent covariates like BMI is needed [27]. Replication of this analysis in other cohorts is required to better understand longitudinal change in FEV₁-variation as a mortality risk-factor [28–31].

Author contributions

MDW, MW, HWC, DGG, LL and DJP conceived the research strategy. MW, AV and MDW performed data analysis. MW, MDW, AV and LL wrote the manuscript. DGG HWC DP edited the manuscript.

Funding

National Institute for Occupational Safety and Health grants R21OH012256 to DGG, MDW 1R21OH012958-01 TO MDW and DGG contracts 200-2017-93326, 200-2017-93426, 75D301-22-R-72,142, and 75D301-22-R-72,244 to DJP Programme of VLUHR to MHW.

Data availability

The data that support the findings of this study may be available upon request to , David J. Prezant.

Declarations

Competing interests

The authors declare no competing interests.