Abstract
Background.
Chronic rhinosinusitis (CRS) and asthma commonly co-occur. No studies have leveraged large samples needed to formally address whether preexisting CRS is associated with new onset asthma over time.
Methods.
We evaluated whether prevalent CRS (identified in two ways: validated text algorithm applied to sinus computerized tomography [CT] scan or two diagnoses) was associated with new onset adult asthma in the following year. We used electronic health record data from Geisinger in from 2008–2019. For each year we removed persons with any evidence of asthma through the end of the year, then identified those with new diagnosis of asthma in the following year. Complementary log-log regression was used to adjust for confounding variables (e.g., sociodemographic, contact with the health system, co-morbidities), and hazard ratios (HR) and 95% confidence intervals (CI) were calculated.
Results.
A total of 35,441 persons were diagnosed with new onset asthma and were compared to 890,956 persons who did not develop asthma. Persons with new onset asthma tended to be female (69.6%) and younger (mean [SD] age 45.9 [17.0] years). Both CRS definitions were associated (HR, 95% CI) with new onset asthma, with 2.21 (1.93, 2.54) and 1.48 (1.38, 1.59) for CRS based on sinus CT scan and two diagnoses, respectively. New onset asthma was uncommonly observed in persons with a history of sinus surgery.
Conclusion.
Prevalent CRS identified with two complementary approaches was associated with a diagnosis of new onset asthma in the following year. The findings may have clinical implications for the prevention of asthma.
Keywords: asthma etiology, epidemiology, risk factors, rhinosinusitis
Graphical Abstract
CRS and asthma commonly co-occur, but no studies have leveraged the very large sample sizes needed to address whether preexisting CRS is associated with new onset asthma over time.
In a longitudinal study of over 900,000 persons from 2008–2019, prevalent CRS (identified with sinus CT scan or diagnoses) was associated with 1.5–2.2-fold increased risk of the diagnosis of new onset asthma in the following year.
The findings may have implications for the prevention of adult-onset asthma.
Abbreviations: CRS, chronic rhinosinusitis; CT, computerized tomography.
INTRODUCTION
Upper and lower airways diseases commonly co-occur [1–6]. There is particular interest in whether the natural history of chronic rhinosinusitis (CRS) can lead to the development of asthma, but few studies have formally addressed this question [1]. The co-occurrence between CRS and asthma has been extensively explored. However, important specific questions remain unanswered that we address herein, such as whether there is a sub-phenotype of CRS that leads to the development of asthma, the role of allergic rhinitis in this natural history [7], the timing of onset of these related conditions, and the magnitude of the annual risk of new onset asthma in persons with CRS compared to those without.
Both CRS and asthma are common airway conditions that pose a significant burden on persons with these conditions [8]. The reported prevalence of asthma was 6.1% in males and 9.5% in females in 2020 [9]. Factors that predispose toward development of asthma in adulthood include sinusitis during childhood [1], allergic rhinitis, which itself commonly co-occurs with CRS [10], obesity [11, 12], female sex [13], psychosocial factors [14], bronchial hyperresponsiveness [15], active and passive smoking [16], and air pollution [17]. Understanding the risk for development of asthma in persons with CRS could motivate the development of management approaches to prevent this component of the natural history of CRS.
Both asthma and CRS are now recognized as heterogeneous inflammatory disorders with variable severity [18–20]. While CRS has historically been classified by the presence or absence of nasal polyps (CRS with [CRSwNP] and without [CRSsNP] nasal polyps), there is an increasing emphasis on CRS endotypes characterizing features of the inflammatory profile (e.g., types 1, 2, 3, or mixed inflammation) [19]. Tissue eosinophilia is a strong predictor of CRS severity [18, 21, 22]. The co-occurrence of CRS and asthma also carries implications for the severity of both conditions, such that CRS and its exacerbations are usually more severe in patients with asthma [23–29], and there is a high rate (80%) of CRS in patients with severe asthma [30]. The co-occurrence of these conditions is thought to be due to shared inflammation and contiguous airways [5].
No prior studies have been specifically designed to evaluate the risk of asthma over time in persons first diagnosed with CRS based on objective evidence, independent of sinus surgery [7] (we have reported on this association using self-reported information [31]). Such a study requires a large sample size given the relatively low annual incidence of asthma (e.g., estimated at 3.8 cases per 1000 person-years in one U.S. study [32]), and the requirement that prevalent asthma, and thus some prevalent CRS, must be excluded. Studies using longitudinal electronic health records (EHR) are well-suited to address this question.
We hypothesized that sinonasal inflammation was an independent risk factor for the development of asthma in people not previously noted to have asthma. We report here the risk of development of new onset asthma in persons with and without CRS. We used EHR data and radiologic findings on sinus computerized tomography (CT) scan to identify individuals with objective evidence of CRS. We also evaluated whether the association of CRS with new onset asthma was different by age, sex, allergic rhinitis status, history of eosinophilia, and sinus surgery status.
METHODS
Study Population and Design
We used EHR data from Geisinger and designed a longitudinal cohort study. Geisinger is an integrated health system serving over 2 million patients with many different health insurers in 40 counties in central and northern Pennsylvania. EHR data were used from 2005 to 2019 to identify persons without asthma every year from 2008 to 2018; these persons were at risk for development of new asthma in the following year, referred to as the risk set. New onset asthma was identified every year from 2009 to 2019. Persons had to be over 18 years of age in the year of entry into the cohort.
A total of 1,510,875 persons were eligible for the study before the application of inclusion and exclusion criteria. To ensure the presence of longitudinal data, we required that persons have contact with the health system in at least two years, no more than four years apart. More frequent contact was not required as younger, healthy adults are less likely to seek health care annually [33]. We excluded persons with common variable immunodeficiency, cystic fibrosis, primary ciliary dyskinesia, Munier-Kuhn syndrome (congenital tracheobronchomegaly), and allergic bronchopulmonary aspergillosis using diagnosis entries in the EHR. The final analysis included 926,397 unique persons.
Identification of CRS
CRS was identified in two ways, the first based on sinus CT scan text in radiology reports, and the second based on two CRS diagnosis entries on different days. The onset date of CRS for the former was the date of the sinus CT scan and for the latter was the date of the first diagnosis. This allowed us to create two ever vs. never CRS variables and the duration of each CRS definition through the end of the risk set year.
The EHR text algorithm that was applied to text of the sinus CT radiology report excluded “minimal” disease or opacification but included “mild” disease or opacification. The text algorithm included unilateral disease thought to be sinus inflammation but excluded unilateral sinus disease thought to be secondary to another process (e.g., papilloma, large retention cyst, antrochoanal polyp). Approximately 24% of positive sinus CT scans were based on the presence in the radiology report of evidence of past sinus surgery (Online Supplement Method-E1). The final text algorithm (Online Supplement Method-E2 and Table E1) was validated and found to have a positive predictive value (PPV) over 0.90 against chart review as the gold standard (Online Supplement Method-E3). The PPV was inadequate for valid identification of CRSwNP, so we could not distinguish CRSwNP from CRSsNP in the primary analysis.
Primary Health Outcome: Identification of New Onset Asthma
Asthma was identified with two diagnoses for asthma or one diagnosis plus one order for an asthma medication. The first evidence could occur no earlier than the year after the risk set, and the second at any time. This analysis required the removal of persons with prevalent asthma so that risk sets only included persons without any evidence of asthma through the relevant year. Since CRS and asthma commonly co-occur, this also removed many persons with prevalent CRS who had asthma.
Confounding Variables
EHR data were used to identify age (years [linear, squared, and cubic terms]); sex (female vs. male); race (non-White vs. White); ethnicity (Hispanic vs. non-Hispanic); Medical Assistance status (a surrogate for family socioeconomic status [34], ever vs. never used based on percent of time Medical Assistance was used); number of outpatient encounters through the risk set year (0 [i.e., contact with the healthcare system could be documented with medication orders or laboratory tests, not just encounters], 1 to < 3, 3 to < 6, 6 to < 11 [reference], 11 to < 18, 18 to < 30, 30+); body mass index (BMI, kg/m2, linear and squared); calendar year (indicator for each year, 2008 as reference); tobacco use (current, former, unknown vs. never); and co-morbid conditions based on two diagnoses for allergic rhinitis, gastroesophageal reflux disease (GERD), chronic obstructive pulmonary disease (COPD), obstructive sleep apnea (OSA), or one diagnosis for pneumonia. Eosinophilia in whole blood was defined as a count ever exceeding 500 cells/μL. History of sinus surgery was based on procedure codes (Current Procedural Terminology, CPT®) and text in the sinus CT scan report indicating post-surgical changes in paranasal sinuses.
Statistical Analysis
The primary goal of the analysis was to evaluate associations of prevalent CRS, identified in two ways, with new onset asthma in the following year (yes vs. no) while adjusting for potential confounding variables. Persons in the analysis with and without new onset asthma and by CRS and confounding variable status were first described. We used complementary log-log regression to adjust for confounding using the Stata cloglog command and generalized estimating equations with independence working model for correlation and robust standard errors derived clustered on person. Exponentiated coefficients from these models can be interpreted as hazard ratios [35]. Hazard ratios with 95% confidence intervals are reported.
Several pre-specified analyses to evaluate effect modification were completed by inclusion of cross-product terms between the CRS variables and 1) age, 2) sex (to evaluate sex differences in relation of CRS with new onset asthma), 3) history of allergic rhinitis or elevated eosinophils (to evaluate whether risk of asthma development was higher using two surrogates for type 2 inflammation), or 4) history of sinus surgery (to evaluate whether the risk of asthma development was lower among those with a history of sinus surgery). Sensitivity analyses were completed to evaluate assumptions and robustness of the primary findings (Online Supplement Method-E4).
RESULTS
Overview of Study Design, Data Features, and Study Subjects
A total of 43,552 (4.5%) persons with prevalent asthma were removed from the analysis; this included 2416 (5.6%) persons who had CRS identified by two diagnoses and 360 (0.8%) who ever had a positive sinus CT scan. Among those with prevalent asthma, the mean (SD) age was 45.2 (15.9) years; 67.0% were female; 4.8% Black race, 0.2% mixed race, 1.3% other races, and 0.4% were missing race information; and 4.8% were of Hispanic ethnicity.
Risk sets consisting of persons without prevalent asthma were assembled annually from 2008–2018 (Table 1, for ease of visualization only even years are tabulated) and then persons in the risk set who were diagnosed with new onset asthma were identified in the following year. In the year following the risk set, the number of persons with new onset asthma ranged from 2164 (0.39%) in 2019 (analyzed in relation to 2018 risk set) to 4164 (1.03%) in 2014 (analyzed in relation to 2013 risk set). The mean (SD, median) number of new onset asthma cases per year was 3221.9 (580.8, 3190). The mean (SD, median) annual cumulative incidence was 0.82 percent (0.21, 0.86). A total of 35,441 unique persons were diagnosed with new onset asthma during the study period and were compared to 890,956 unique persons who remained free of asthma (Table 2). On average, persons who developed asthma were more likely to be female, younger, of Black race, of Hispanic ethnicity, had ever received Medical Assistance, and had a higher BMI.
Table 1.
Selected characteristics of study subjects by selected calendar years, 2008–19 (only even years tabulated, all years in analysis). Persons with prevalent asthma were removed from each risk set, defined as persons at risk for new onset asthma in following year.
| Variable | Year of Risk Set (only even years are tabulated, every year was in analysis) | |||||
|---|---|---|---|---|---|---|
| 2008 | 2010 | 2012 | 2014 | 2016 | 2018 | |
| Risk Sets, New Onset Asthma, and CRS | ||||||
| Number in risk set, unique persons | 285,162 | 320,199 | 374,742 | 435,247 | 530,320 | 548,552 |
| Asthma, new diagnosis, in following year, n (%) | 3190 (1.1) | 2733 (0.9) | 3336 (0.9) | 3802 (0.9) | 3625 (0.7) | 2164 (0.4) |
| CRS, positive sinus CT scan, through risk set year, n (%) | 544 (0.2) | 992 (0.3) | 1371 (0.4) | 1648 (0.4) | 1911 (0.4) | 2074 (0.4) |
| CRS, two diagnoses, through risk set year, n (%) | 5055 (1.8) | 6748 (2.1) | 8436 (2.3) | 9967 (2.3) | 11,000 (2.1) | 11,323 (2.1) |
| Duration after positive sinus CT scan, years, mean (SD) | 1.6 (1.3) | 2.3 (1.6) | 3.2 (2.1) | 4.3 (2.6) | 5.2 (3.2) | 6.2 (3.7) |
| Sinus surgery, ever through risk set year, n (%) | 428 (0.15) | 730 (0.23) | 985 (0.26) | 1241 (0.29) | 1564 (0.29) | 1872 (0.34) |
| Sociodemographics | ||||||
| Age to Dec 31st of risk set year, years, mean (SD) | 47.7 (16.2) | 48.6 (16.6) | 49.4 (17.1) | 50.1 (17.5) | 51.1 (17.9) | 52.2 (18.2) |
| Sex, female, % | 57.9 | 57.3 | 57.0 | 56.8 | 56.4 | 56.3 |
| Race, White, % | 97.0 | 96.8 | 96.6 | 96.1 | 95.0 | 94.3 |
| Ethnicity, Hispanic, % | 1.8 | 2.2 | 2.5 | 2.8 | 3.2 | 3.7 |
| Medical Assistance, % time > 0, % | 10.9 | 13.3 | 14.6 | 14.6 | 14.1 | 14.6 |
| Habits | ||||||
| Tobacco use, % | ||||||
| Health care utilization | ||||||
| Duration from first contact through year, years, mean (SD) | 2.9 (1.2) | 4.1 (2.0) | 5.2 (2.8) | 6.2 (3.5) | 6.8 (4.3) | 7.9 (4.9) |
| OPT visits from first contact through year, mean (SD) | 10.0 (11.7) | 13.4 (165.2) | 15.8 (20.0) | 17.7 (23.5) | 18.2 (26.0) | 21.1 (29.4) |
| Co-morbid conditions * | ||||||
| Pneumonia, any, ever through risk set year, n (%) | 5951 (2.1) | 9304 (2.9) | 13,693 (3.7) | 17,518 (4.0) | 22,050 (4.2) | 26,810 (4.9) |
| COPD prevalence through risk set year, n (%) | 6488 (2.3) | 8382 (2.6) | 10,367 (2.8) | 13,007 (3.0) | 16,463 (3.1) | 19,849 (3.6) |
| AR prevalence through risk set year, n (%) | 23,004 (8.1) | 28,243 (8.8) | 32,591 (8.7) | 36,658 (8.4) | 39,676 (7.5) | 40,989 (7.4) |
| GERD prevalence through risk set year, n (%) | 6076 (2.1) | 8677 (2.7) | 19,202 (5.1) | 37,920 (8.7) | 58,351 (11.0) | 71,790 (13.1) |
| Eosinophils > 500/μL, ever through risk set year, n (%) | 10,970 (3.9) | 13,175 (4.1) | 17,101 (4.6) | 21,648 (5.0) | 27,604 (5.2) | 33,489 (601) |
| OSA prevalence through risk set year, n (%) | 7161 (2.5) | 11,121 (3.5) | 15,983 (4.3) | 20,546 (4.7) | 28,623 (5.4) | 34,185 (6.2) |
| Body mass index before calendar year, kg/m2, mean (SD) | 29.7 (7.1) | 29.7 (7.2) | 29.8 (7.2) | 29.8 (7.2) | 30.1 (7.3) | 30.2 (7.4) |
Abbreviations: AR = allergic rhinitis; COPD = chronic obstructive pulmonary disease; CRS = chronic rhinosinusitis; CT = computed tomography; GERD = gastroesophageal reflux disease; OPT = outpatient; OSA = obstructive sleep apnea
For co-morbid conditions, based on two diagnoses except for pneumonia for which one was used
Table 2.
Selected summary statistics, through the end of the risk set year, by new onset asthma status for 926,397 unique persons in analysis. All persons with prevalent asthma before the risk set year were removed from the analysis.
| Variable* | Status | p-value | |
|---|---|---|---|
| Persons with New Onset Asthma | Persons without Asthma | ||
| Number | 35,441 | 890,956 | |
| Sex, female, n (%) | 24,656 (69.6) | 485,381 (54.5) | < 0.001 |
| Age, years, mean (SD) | 45.9 (17.0) | 50.1 (18.7) | < 0.001 |
| Race, n (%) | < 0.001 | ||
| Hispanic ethnicity, n (%) | 1992 (5.6) | 33,061 (3.7) | < 0.001 |
| Tobacco use, n (%) | < 0.001 | ||
| Medical Assistance, ever, n (%) | 8002 (22.6) | 127,356 (14.3) | < 0.001 |
| CRS, prevalent to risk set year, n (%) | |||
| Pneumonia, n (%) | 645 (1.8) | 39,817 (4.5) | < 0.001 |
| Chronic obstructive pulmonary disease, n (%) | 779 (2.2) | 30.642 (3.4) | < 0.001 |
| Allergic rhinitis, n (%) | 3131 (8.8) | 55,358 (6.2) | < 0.001 |
| Gastroesophageal reflux disease, n (%) | 2754 (7.8) | 85,539 (9.6) | < 0.001 |
| Eosinophils > 500/μL, ever, n (%) | 1721 (4.9) | 48,830 (5.5) | < 0.001 |
| Obstructive sleep apnea, n (%) | 1709 (4.8) | 43,045 (4.8) | 0.94 |
| Body mass index, kg/m2, mean (SD) | 32.0 (8.6) | 29.8 (7.3) | < 0.001 |
Abbreviations: CRS = chronic rhinosinusitis; CT = computed tomography
All co-morbid conditions were identified with two diagnoses except pneumonia, which was identified with a single diagnosis. All conditions had to have at least the first diagnosis before the year of asthma onset.
Characteristics of Study Subjects with CRS
A total of 2897 persons ever had a positive sinus CT scan; these persons had a mean (SD) age (years) of 49.7 (15.7), ranging from 18.0 to 86.5 years; 49.2% were female; 97.1% and 1.5% were White and Black race, respectively; 2.0% were of Hispanic ethnicity; 16.1%, 33.1%, 47.3%, and 3.5% were current, former, never, and unknown users of tobacco; and 13.5% had ever been on Medical Assistance. These persons had a mean (SD) BMI of 30.3 (6.9) kg/m2; 9.7% had a history of pneumonia, 6.3% of COPD, 31.6% of allergic rhinitis, 23.8% of GERD, 13.7% of OSA, and 12.9% ever had a blood eosinophil level > 500 cells/μL.
A total of 15,717 persons were identified with CRS based on two diagnoses; these persons had a mean (SD) age (years) of 48.8 (16.1), ranging from 18.3 to 87.2 years; 57.0% were female; 97.5% and 1.5% were White and Black race, respectively; 1.9% were Hispanic ethnicity; 17.41%, 32.7%, 48.5%, and 1.4% were current, former, never, and unknown users of tobacco; and 17.3% had ever been on Medical Assistance. These persons had a mean (SD) BMI of 30.5 (7.1) kg/m2; 12.2% had a history of pneumonia, 7.9% of COPD, 40.9% of allergic rhinitis, 29.9% of GERD, 14.0% of OSA, and 12.3% ever had a blood eosinophil level > 500 cells/μL. The much larger number of persons identified by diagnoses than sinus CT scans is likely due to the common assignment of diagnoses by primary care providers so this measure may have higher sensitivity but lower specificity than the sinus CT scan definition.
Adjusted Associations with New Onset Asthma
Complementary log-log models identified several variables that were strongly and consistently associated with the diagnosis on new onset asthma (Table 3 and Figure 1). Persons who were female, Black race, Hispanic ethnicity, current users of tobacco, or had ever received Medical Assistance were more likely to be diagnosed with new onset asthma. Persons with allergic rhinitis, GERD, OSA, and a history of eosinophilia were also more likely to be diagnosed with new onset asthma, whereas persons with a history of pneumonia were less likely. Notably, both CRS variables were independently associated with new onset asthma, with associations stronger for the variable based on a positive sinus CT scan than for two CRS diagnoses. When sinus surgery was added to these models, associations were attenuated and sinus surgery itself was also associated with new onset asthma (Table 3 and Figure 1). When duration of CRS in years (quartiles) from the onset of CRS through the end of the risk set year were substituted for the binary CRS variables in fully adjusted models, hazard ratios were most elevated for shorter durations but had confidence intervals that excluded 1.0 for over six years (Table 4).
Table 3.
Fully adjusted associations* of selected variables with 35,441 cases of new onset asthma (vs. no new onset asthma), 2008–2018.
| Variable | Hazard Ratio (95% Confidence Interval) | |
|---|---|---|
| CRS sinus CT scan ever Model 1 |
CRS prevalence, two diagnoses Model 2 |
|
| Models without sinus surgery in the model | ||
| CRS, positive CT scan, ever vs. never | Figure 1, model 1a | |
| CRS, prevalent, two diagnoses vs. not | Figure 1, model 2a | |
| Female vs. male | 1.76 (1.72, 1.80) | 1.76 (1.72, 1.80) |
| Race | ||
| Hispanic vs. non-Hispanic ethnicity | 1.66 (1.58, 1.74) | 1.66 (1.58, 1.74) |
| Medical Assistance, > 0% time vs. never | 1.54 (1.50, 1.59) | 1.55 (1.50, 1.59) |
| Tobacco use | ||
| Pneumonia, ever vs. never | 0.61 (0.56, 0.66) | 0.60 (0.56, 0.65) |
| COPD, ever vs. never, two diagnoses | 1.06 (0.98, 1.15) | 1.06 (0.98, 1.14) |
| Allergic rhinitis, ever vs. never, two diagnoses | 1.48 (1.42, 1.54) | 1.45 (1.39, 1.51) |
| GERD, ever vs. never, two diagnoses | 1.41 (1.35, 1.47) | 1.40 (1.34, 1.46) |
| Eosinophils, ever > 500 cells/μL | 1.28 (1.22, 1.35) | 1.28 (1.22, 1.35) |
| OSA, ever vs. never, two diagnoses | 1.19 (1.13, 1.26) | 1.19 (1.13, 1.26) |
| Models with sinus surgery in the model | ||
| Model 2a | Model 2b | |
| CRS, positive CT scan, ever vs. never | Figure 1, model 1b | |
| CRS, prevalent, two diagnoses vs. not | Figure 1, model 2b | |
| Sinus surgery†, ever | Figure 1, model 1b | Figure 1, model 2b |
Abbreviations: COPD = chronic obstructive pulmonary disease; CRS = chronic rhinosinusitis; CT = computed tomography; GERD = gastroesophageal reflux disease; OSA = obstructive sleep apnea
All models also included age (linear, squared, and cubic terms), cumulative outpatient encounters through calendar year (0, 1 to < 3, 3 to < 6, 6 to < 11 [reference], 11 to < 18, 18 to < 30, 30+), body mass index (BMI, BMI2), and year (indicator for every year 2008 to 2018 vs. 2013 [reference]). All covariates were assessed through the end of the risk set year. Models 2a and 2b are the same as 1a and 1b except that sinus surgery was added to the two separate models.
Identified with procedure codes or with sinus CT scan text indicating prior sinus surgery
Figure 1.
Associations of two CRS variables with new onset asthma with and without adjustment for history of sinus surgery. Model 1 presents the CRS variable based on an abnormal sinus CT scan, without (Model 1a) and with adjustment for sinus surgery (Model 1b). Model 2 presents the CRS variable based on two CRS diagnoses on different days, without (Model 2a) and with adjustment for sinus surgery (Model 2b). These results are from the models in Table 3, and are adjusted for the same confounding variables.
Table 4.
Evaluation of associations of CRS duration quartiles (for each of two definitions of CRS) with new onset asthma from fully adjusted* final models.
| CRS duration variable | HR (95% CI) |
|---|---|
| CRS positive CT scan, duration from sinus CT to risk set year | |
| CRS prevalence, two diagnoses, duration from first diagnosis to risk set year |
Abbreviations: COPD = chronic obstructive pulmonary disease; CRS = chronic rhinosinusitis; CT = computed tomography; GERD = gastroesophageal reflux disease; OSA = obstructive sleep apnea
All models adjusted for age (linear, squared, and cubic terms); sex (female vs. male); cumulative outpatient encounters through calendar year (0, 1 to < 3, 3 to < 6, 6 to < 11 [reference], 11 to < 18, 18 to < 30, 30+); Medical Assistance (% time using > 0% vs. none); tobacco use (current, former, unknown vs. never); body mass index (kg/m2, BMI, BMI2), year (indicator for every year 2009 to 2018 vs. 2013 [reference]); race (Non-White vs. White); ethnicity (Hispanic vs. non-Hispanic); COPD, allergic rhinitis, GERD, and OSA (ever vs. never based on two diagnoses); pneumonia (ever vs. never based on one diagnosis); and history of blood eosinophilia > 500 cells/μL (ever vs. never). All covariates were assessed through the end of the risk set year.
There was no consistent evidence of effect modification of associations of CRS with new onset asthma by age, sex, history of allergic rhinitis, history of elevated eosinophils, or history of sinus surgery (results not shown). The analysis of effect modification by history of sinus surgery was limited by small sample size; across the 11 years, a mean (SD) of 12.1 (4.2) persons developed new onset asthma after sinus surgery.
Sensitivity Analyses
The analysis requiring a visit in the year prior to the risk set year, to require more time to identify prevalent conditions, revealed no substantive change in associations (results not shown). Analysis confined to 2012 and later (open dynamic cohort, so new patients could enter and leave after 2012) revealed no substantive change to associations (HR, 95% CI) for prevalent CRS based on sinus CT scan (2.15 [1.80, 2.55]) or two diagnoses (1.48 [1.35, 1.62]). Analysis of the closed cohort with entry only in 2012 revealed no substantive change to associations (HR, 95% CI) for prevalent CRS based on sinus CT scan (2.16 [1.77, 2.62]) or two diagnoses (1.49 [1.35, 1.65]). When the analysis was repeated with a cross-product between CRS and an indicator for asthma medication, there were no substantive changes in associations (results not shown).
The analysis to evaluate the proportional hazards assumption and differing measurement error across years, with CRS*age cross-products, showed a global test p-value for the group of cross-product terms all at once that was highly significant (p < 0.0004). This is evidence that the proportional hazards assumption was violated. The associations revealed that the strength of associations of CRS with new onset asthma declined across years, more evident for CRS diagnoses than for CRS based on the sinus CT scan (Online Supplement Table E2).
DISCUSSION
In a longitudinal analysis of prevalent CRS as a risk factor for new onset asthma, we found that two independent approaches for identification of CRS, one based on sinus CT scan reports and the other using two diagnosis entries, were both associated with a new diagnosis of asthma. Associations were robust to increasing covariate control and in several sensitivity analyses. Prevalent asthma was excluded in each of 11 years, 2008–2018, and a new diagnosis of asthma in the following year, 2009–2019, was identified with two EHR diagnoses or one diagnosis and a medication order. Annual asthma incidence was approximately 7 to 10 per 1000, higher than previous estimates [32], which have not been commonly reported from the U.S. Hazard ratios were higher for CRS based on a positive sinus CT scan, objective evidence of disease, and revealed that persons with a positive sinus CT scan in the past through the end of the risk set year were over 2.2 times more likely to be diagnosed with asthma in the following year. The sinus CT scan text algorithm was validated by chart review by two independent reviewers and had a PPV exceeding 0.90. CRS based on two diagnoses was not compared to a chart review. We suspect this has higher sensitivity but lower specificity than the sinus CT scan measure.
There is considerable heterogeneity in asthma; this implies that CRS is unlikely to be a risk factor for all phenotypes (and sub-endotypes) so the strong associations observed herein are likely stronger for certain asthma subgroups [18, 20]. Although there is growing evidence that eosinophils and type 2 inflammation are important to CRS and its links to asthma [18, 36], we found no evidence that the association of CRS with new onset asthma differed by history of eosinophilia or allergic rhinitis, which may not be surprising given the imprecision of these surrogates for the inflammatory heterogeneity of CRS [19]. In addition, there is some evidence that sinus surgery earlier in disease, versus later, can prevent the future development of co-morbidities from CRS but we found no evidence in a formal analysis that the associations of CRS with new onset asthma differed among those with and without sinus surgery. However, this analysis was limited by small sample sizes; fewer than 10 persons with past sinus surgery per year developed new onset asthma. The very small number of persons with past sinus surgery who were diagnosed with asthma in the following year is of itself quite interesting and may have clinical importance.
We found that the proportional hazards assumption was violated, with hazard ratios declining across years. However, we believe this is likely due to an artifact of the methodologic approach and not that it invalidates our primary findings. Both CRS by chart review and asthma were recorded as the first date after the appearance of the second piece of information, which could occur after the risk set year; the longer the observation time, the more likely to find new onset asthma. In later years, the two pieces of independent evidence to identify either condition can only occur in fewer years; for example, for the 2018 risk set year, this information can only be provided in the single year of 2019. This suggests that measurement error for prevalent CRS and new onset asthma was higher in later years. Similarly, because of the inclusion criterion of at least two encounters in different years no more than four years apart, prevalent asthma was more completely excluded in later years and in persons with longer observation times. False negative status for the conditions we studied were thus higher in earlier years. There may also be biologic explanations for the declining strength of associations over years, discussed below. Nevertheless, because the proportional hazards assumption was violated, we must be cautious about the interpretation of these findings.
Associations with CRS duration were stronger with shorter durations but persisted through over six years, providing some assurance about our findings. The duration analysis may suggest diagnosis bias because of stronger associations after shorter durations, but hazard ratios were elevated through over six years, arguing against a simple diagnostic bias for asthma after recent diagnosis of CRS accounting for associations. It is also possible that stronger associations with shorter durations may reflect an underlying biologic mechanism. Asthma is associated with more severe CRS and CRS exacerbations [26–29]. Patients with more severe CRS with more frequent exacerbations may seek earlier medical care and receive more intensive care so the strength of the CRS-asthma association would decline over time and with increasing duration of CRS. It may be more common for asthma to develop shortly after CRS and for inflammation in CRS to weaken over time with treatment. These associations could also be due, at least in part, to the previously discussed features regarding CRS and asthma ascertainment with two independent pieces of information over time.
The study had several limitations. We cannot reliably know nasal and sinus symptom status from EHR data, but it is a reasonable assumption that sinus CT scans showing radiologic inflammation were most commonly obtained in persons seeking care for relevant symptoms, so we assume that patients with a positive sinus CT scan had CRS. Our text algorithm was unable to separately identify CRSwNP and CRSsNP, and CRS surgery is not a good surrogate to identify polypoid disease because over 65% of endoscopic sinus surgery in the U.S. is for non-polypoid disease [37]. In EHR data, asthma diagnosis is likely to be variable between clinicians and may be a clinical impression or empirical diagnosis, thus posing the possibility of measurement error. This measurement error could bias away from the null in persons with CRS. It is possible that symptoms of CRS were most prominent at presentation, but asthma was also present and not the focus of diagnosis efforts at that time. This could explain the stronger associations for shorter durations between CRS and asthma, but concern is mitigated somewhat by associations out to six or more years duration between them. Requiring two diagnoses has the benefit of increasing specificity but may lead to data analytic artifacts for detecting associations in later years as noted above. The study sample consisted primarily of persons of White race and non-Hispanic ethnicity, potentially limiting the generalizability of the findings to other racial and ethnic groups.
The study had several strengths, including the required very large sample size (i.e., almost one million persons’ health records were analyzed), over 10 years of longitudinal data, and many new cases of asthma annually. Participants represented the general population in the region [38]. CRS was identified using two independent methods, including one that relied on a validated approach to sinus disease classification using sinus CT scan reports. We adjusted for many potential confounding variables, and the findings were robust in several sensitivity analyses evaluating the data structure, analytic assumptions, outcome definitions, and features of the data over time. Expected associations were present in the data (i.e., with race, ethnicity, Medical Assistance, allergic rhinitis, and eosinophilia), suggesting that the asthma identification was valid.
Prevalent CRS, including a definition based on objective information, was associated with the diagnosis of new onset asthma in the following year. New onset asthma was very uncommonly observed in persons with a history of sinus surgery. The findings may have clinical implications for the prevention of asthma.
Supplementary Material
Acknowledgments:
P01AI145818 (Chronic Rhinosinusitis Integrative Studies Program-2 [CRISP2], PI: R.P. Schleimer) from the National Institute of Allergy and Infectious Diseases. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.
Abbreviations:
- CRS
chronic rhinosinusitis
- EHR
electronic health record
References
- 1.Chang EH, et al. , Early life risk factors for chronic sinusitis: A longitudinal birth cohort study. J Allergy Clin Immunol, 2018. 141(4): p. 1291–1297 e2. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 2.Crimi C, et al. , Type 2-High Severe Asthma with and without Bronchiectasis: A Prospective Observational Multicentre Study. J Asthma Allergy, 2021. 14: p. 1441–1452. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 3.Frieri M, Asthma linked with rhinosinusitis: An extensive review. Allergy Rhinol (Providence), 2014. 5(1): p. 41–9. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 4.Lehrer E, et al. , Management of chronic rhinosinusitis in asthma patients: is there still a debate? Curr Allergy Asthma Rep, 2014. 14(6): p. 440. [DOI] [PubMed] [Google Scholar]
- 5.Bachert C, Patou J, and Van Cauwenberge P, The role of sinus disease in asthma. Curr Opin Allergy Clin Immunol, 2006. 6(1): p. 29–36. [DOI] [PubMed] [Google Scholar]
- 6.Schwartz BS, et al. , Strong and consistent associations of precedent chronic rhinosinusitis with risk of non-cystic fibrosis bronchiectasis. J Allergy Clin Immunol, 2022. 150(3): p. 701–708 e4. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 7.Benninger MS, et al. , Impact of medically recalcitrant chronic rhinosinusitis on incidence of asthma. Int Forum Allergy Rhinol, 2016. 6(2): p. 124–9. [DOI] [PubMed] [Google Scholar]
- 8.American Lung Association. Asthma Trends and Burden. 2023. [cited 2023 January 9, 2023]; Available from: https://www.lung.org/research/trends-in-lung-disease/asthma-trends-brief/trends-and-burden.
- 9.National Center for Environmental Health. 2020 National Health Interview Survey (NHIS) Data. NHIS Asthma Prevalence Data 2020 [cited 2023 January 9, 2023]; Available from: https://www.cdc.gov/asthma/nhis/2020/data.htm. [Google Scholar]
- 10.Guerra S, et al. , Rhinitis as an independent risk factor for adult-onset asthma. J Allergy Clin Immunol, 2002. 109(3): p. 419–25. [DOI] [PubMed] [Google Scholar]
- 11.Ronmark E, et al. , Obesity increases the risk of incident asthma among adults. Eur Respir J, 2005. 25(2): p. 282–8. [DOI] [PubMed] [Google Scholar]
- 12.Beuther DA and Sutherland ER, Overweight, obesity, and incident asthma: a meta-analysis of prospective epidemiologic studies. Am J Respir Crit Care Med, 2007. 175(7): p. 661–6. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 13.de Marco R, et al. , Differences in incidence of reported asthma related to age in men and women. A retrospective analysis of the data of the European Respiratory Health Survey. Am J Respir Crit Care Med, 2000. 162(1): p. 68–74. [DOI] [PubMed] [Google Scholar]
- 14.Brunner WM, et al. , Depression and risk of incident asthma in adults. The CARDIA study. Am J Respir Crit Care Med, 2014. 189(9): p. 1044–51. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 15.Anto JM, et al. , Risk factors of new-onset asthma in adults: a population-based international cohort study. Allergy, 2010. 65(8): p. 1021–30. [DOI] [PubMed] [Google Scholar]
- 16.Coogan PF, et al. , Active and passive smoking and the incidence of asthma in the Black Women’s Health Study. Am J Respir Crit Care Med, 2015. 191(2): p. 168–76. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 17.Young MT, et al. , Ambient air pollution exposure and incident adult asthma in a nationwide cohort of U.S. women. Am J Respir Crit Care Med, 2014. 190(8): p. 914–21. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 18.Brown HJ, et al. , Defining the Allergic Endotype of Chronic Rhinosinusitis by Structured Histopathology and Clinical Variables. J Allergy Clin Immunol Pract, 2021. 9(10): p. 3797–3804. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 19.Kato A, Schleimer RP, and Bleier BS, Mechanisms and pathogenesis of chronic rhinosinusitis. J Allergy Clin Immunol, 2022. 149(5): p. 1491–1503. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 20.Agache I, et al. , Advances and highlights in asthma in 2021. Allergy, 2021. 76(11): p. 3390–3407. [DOI] [PubMed] [Google Scholar]
- 21.Shah SA, Ishinaga H, and Takeuchi K, Pathogenesis of eosinophilic chronic rhinosinusitis. J Inflamm (Lond), 2016. 13: p. 11. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 22.Snidvongs K, et al. , Structured histopathology profiling of chronic rhinosinusitis in routine practice. Int Forum Allergy Rhinol, 2012. 2(5): p. 376–85. [DOI] [PubMed] [Google Scholar]
- 23.Fokkens WJ, et al. , European Position Paper on Rhinosinusitis and Nasal Polyps 2012. Rhinol Suppl, 2012. 23: p. 3 p preceding table of contents, 1–298. [PubMed] [Google Scholar]
- 24.Promsopa C, et al. , Prevalence of confirmed asthma varies in chronic rhinosinusitis subtypes. Int Forum Allergy Rhinol, 2016. 6(4): p. 373–7. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 25.Lin DC, et al. , Association between severity of asthma and degree of chronic rhinosinusitis. Am J Rhinol Allergy, 2011. 25(4): p. 205–8. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 26.Kountakis SE, et al. , Molecular and cellular staging for the severity of chronic rhinosinusitis. Laryngoscope, 2004. 114(11): p. 1895–905. [DOI] [PubMed] [Google Scholar]
- 27.Batra PS, Tong L, and Citardi MJ, Analysis of comorbidities and objective parameters in refractory chronic rhinosinusitis. Laryngoscope, 2013. 123 Suppl 7: p. S1–11. [DOI] [PubMed] [Google Scholar]
- 28.Kwah JH, et al. , Clinical factors associated with acute exacerbations of chronic rhinosinusitis. J Allergy Clin Immunol, 2020. 145(6): p. 1598–1605. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 29.Wu D, Bleier B, and Wei Y, Definition and characteristics of acute exacerbation in adult patients with chronic rhinosinusitis: a systematic review. J Otolaryngol Head Neck Surg, 2020. 49(1): p. 62. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 30.Bresciani M, et al. , Rhinosinusitis in severe asthma. J Allergy Clin Immunol, 2001. 107(1): p. 73–80. [DOI] [PubMed] [Google Scholar]
- 31.Hirsch AG, et al. , Risk of new-onset and prevalent disease in chronic rhinosinusitis: A prospective cohort study. Int Forum Allergy Rhinol, 2023. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 32.Winer RA, et al. , Asthma incidence among children and adults: findings from the Behavioral Risk Factor Surveillance system asthma call-back survey--United States, 2006–2008. J Asthma, 2012. 49(1): p. 16–22. [DOI] [PubMed] [Google Scholar]
- 33.Villarroel MA, Blackwell DL, and Jen A. Secondary Tables of Summary Health Statistics for U.S. Adults: 2018 National Health Interview Survey 2018. Tables of Summary Health Statistics for U.S. Adults: National Health Interview Survey 2018 [cited 2023; Available from: https://www.cdc.gov/nchs/nhis/shs/tables.htm. [Google Scholar]
- 34.Casey JA, et al. , Measures of SES for Electronic Health Record-based Research. Am J Prev Med, 2018. 54(3): p. 430–439. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 35.Prentice RL and Gloeckler LA, Regression analysis of grouped survival data with application to breast cancer data. Biometrics, 1978. 34(1): p. 57–67. [PubMed] [Google Scholar]
- 36.Bachert C, et al. , Efficacy and safety of benralizumab in chronic rhinosinusitis with nasal polyps: A randomized, placebo-controlled trial. J Allergy Clin Immunol, 2022. 149(4): p. 1309–1317 e12. [DOI] [PubMed] [Google Scholar]
- 37.Ference EH, et al. , How often is sinus surgery performed for chronic rhinosinusitis with versus without nasal polyps? Am J Rhinol Allergy, 2018. 32(1): p. 34–39. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 38.Casey JA, et al. , Greenness and Birth Outcomes in a Range of Pennsylvania Communities. Int J Environ Res Public Health, 2016. 13(3). [DOI] [PMC free article] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.