Factors Associated with Persistence of Severe Asthma from Late Adolescence to Early Adulthood

Abstract

Rationale: Asthma in children generally starts as being mild but may progress to being severe and may stay severe for unknown reasons.

Objectives: To identify factors in childhood that predict the persistence of severe asthma in late adolescence and early adulthood.

Methods: The CAMP (Childhood Asthma Management Program) is, to our knowledge, the largest and longest asthma trial to date; it includes 1,041 children aged 5–12 years with mild to moderate asthma. We evaluated 682 program participants with analyzable data in late adolescence (age, 17–19 yr) and early adulthood (age, 21–23 yr).

Measurements and Main Results: To best capture the cases of severe asthma, a status of severe asthma was defined by using criteria from the American Thoracic Society and the National Asthma Education and Prevention Program. Logistic regression with stepwise elimination was used to analyze the clinical features, biomarkers, and lung function that are predictive of the persistence of severe asthma. In late adolescence and early adulthood, 12% and 19% of the participants had severe asthma, respectively; only 6% at both time points had severe cases. For every 5% decrease in the postbronchodilator FEV₁/FVC ratio in childhood, the odds of the persistence of severe asthma increased by 2.36-fold (95% confidence interval [CI], 1.70–3.28; P < 0.0001); for participants with maternal smoking during pregnancy, the odds of the persistence of severe asthma increased by 3.17-fold (95% CI, 1.18–8.53; P = 0.02). A reduced-growth lung function trajectory was significantly associated with the persistence of severe asthma compared with a normal-growth lung function trajectory.

Conclusions: Lung function and maternal smoking during pregnancy were significant predictors of severe asthma from late adolescence to early adulthood. Early interventions to preserve lung function may prevent disease progression.

At a Glance Commentary

Scientific Knowledge on the Subject

Asthma in children typically starts out as being mild, but, for reasons to be determined, some children’s asthma evolves to have a more severe course. Childhood predictors of severe persistent asthma are limited, especially during the critical period from late adolescence to early adulthood, when many patients improve.

What This Study Adds to the Field

We analyzed the CAMP (Childhood Asthma Management Program) trial—the largest and longest asthma trial in children—to identify factors in childhood that predict the persistence of severe asthma. Among the more than 22 relevant factors, a low postbronchodilator FEV₁/FVC ratio and maternal smoking during pregnancy were found to be significant childhood predictors of having severe asthma that persists from late adolescence to early adulthood. Early lung function decline likely drives severity. Interventions to preserve lung function may prevent disease progression.

Asthma is the most common chronic disease, affecting over 339 million adults and children worldwide (1). In 2016, the prevalence of asthma in the United States was 8.3%, affecting over 20 million adults and 6 million children (2). Poorly controlled asthma leads to major functional disability, a major financial burden, and a reduced quality of life (3, 4). Therefore, predicting the clinical trajectory of childhood asthma has important prognostic implications.

Asthma in children is often mild (2), but unfortunately, for reasons to be determined, some children’s asthma evolves to a more severe disease course. The prevalence of severe asthma in childhood and adolescence is poorly understood, with estimates ranging from 5% to 10% (5–7). The estimates for severe asthma are quite variable and difficult to assess because of the variety of definitions and populations studied (7). Available definitions have been developed primarily for clinical management strategies (8–11). The National Asthma Education and Prevention Program Expert Panel Report 3 (NAEPP EPR 3) guidelines defined severe asthma in patients who are untreated on the basis of impairment (symptoms, use of a rescue inhaler, and lung function) and risk (exacerbations and use of oral corticosteroids) (8). The American Thoracic Society (ATS) defined severe asthma as asthma requiring high-dose inhaled corticosteroids (ICSs) (or systemic corticosteroids) in combination with two or more of the following minor clinical criteria: a second long-term (controller) medication, symptoms, lung function, and exacerbations (9). A primary aim of the ATS definition is to differentiate difficult-to-treat asthma from therapy-resistant asthma. For this paper, we used a combination of these criteria to identify participants who had concerning features related to symptom burdens, lung function, severe exacerbations, and medication requirements.

Cohort studies have focused on predictors of the persistence of asthma from childhood to adolescence or adulthood, with remission being the main focus (usually as a secondary outcome) (12–14), but predictors of persistent severe asthma have been limited. The identification of predictive factors that determine a severe asthma status in childhood may lead to novel interventions to target these patients early in order to improve their overall prognosis.

Started in 1993, the CAMP (Childhood Asthma Management Program) trial is, to our knowledge, the largest and longest asthma clinical trial and follow-up study in children with mild to moderate asthma to date; the first 4.3 years consisted of a randomized, double-blind trial with budesonide, nedocromil, and placebo, which was followed by an additional 14 years of observational study that ended in 2013 (15). We previously characterized the natural history of asthma from childhood to adolescence in the CAMP cohort and found remission of asthma in adolescence to be infrequent and not significantly affected by antiinflammatory controller therapy (16). Wang and colleagues (17) recently found a higher rate of remitting asthma in older adolescents followed in CAMP, although a different set of criteria was applied. Important observations that identified various patterns in the lung growth trajectory from the CAMP data were also reported (18). A significant reduction in lung growth was found in 50% of participants by the time they reached late adolescence, and of those, half had already demonstrated evidence of early decline. Whether such a phenotype of reduced lung growth is associated with clinical features of severe disease is unclear.

From this large CAMP longitudinal cohort, we proposed a definition of severe asthma on the basis of a combination of ATS and NAEPP EPR 3 criteria. In this analysis, we evaluated the natural history of asthma severity from school age to late adolescence and into early adulthood and identified factors associated with the persistence of a severe phenotype. We also determined whether individuals who incurred compromised lung growth experienced the persistence of severe disease. Some of the results of this study have been previously reported in the form of an abstract (19).

Methods

Entry criteria in CAMP included being aged 5–12 years and having a history of mild to moderate asthma and a positive methacholine challenge result. The details of CAMP have been previously described elsewhere (16, 20). Six hundred eighty-two CAMP participants who had available data from late adolescence to early adulthood were studied (Table 1). The study visit used to determine asthma activity groups or phenotypes in late adolescence was the first study visit at the age of 18 years. The study visit used to determine asthma activity groups or phenotypes in early adulthood was the first study visit at the age of 22 years. To ensure that there was no loss to follow-up between late adolescence and early adulthood, the chosen participants were required to have consistent study visits at least once yearly between late adolescence and early adulthood (Figure 1). In order to meet the criterion of consistent study visits, there were 46 participants for whom the study visits chosen in early adolescence and late adulthood started at the ages of 19 years and 23 years, respectively. For remitting asthma, we also considered any asthma activity at all study visits over the preceding year from the chosen study visit. We also considered all study visits over the preceding year in our calculation of hospitalizations, emergency department (ED) visits, and prednisone courses per year.

Table 1.

Characteristics of Selected Participants at Baseline (Childhood [Aged 5–12 yr]) (N = 682)

Age, yr	8.7 (2.0)
Sex, M	59.8
Ethnicity
White	67.7
Black	15.2
Hispanic	9.2
Other	7.8
Body mass index percentile	66th (28th)
Duration of asthma, yr	4.7 (2.6)
Age at symptom onset, yr	3.2 (2.5)
Age at M.D. diagnosis, yr	4.0 (2.6)
At least one positive skin test	87.4
Sensitized and exposed* to indoor allergens	81.7
Eczema	28.9
Hay fever	52.6
Food allergy	18.0
Parental asthma	43.1
Treatment group assignment in CAMP†
Budesonide	29.8
Nedocromil	29.6
Placebo	40.6

Definition of abbreviations: CAMP = Childhood Asthma Management Program; M.D. = Doctor of Medicine.

Data are presented as a percentage or as the mean (SD).

^*The criteria for significant allergen exposure are as follows: cat allergen concentration >8,000 ng/g or cat allowed inside the home; dog allergen concentration >10,000 ng/g or dog allowed inside the home; mite allergen concentration >2,000 ng/g; cockroach allergen concentration >2 U/g or cockroaches in the home; and >25,000 mold colonies or mold on any surface of the home.

^†Refers to original treatment group assignment from the Childhood Asthma Management Program Research Group (see Reference 15).

Figure 1.

Study participant flowchart. CAMP = Childhood Asthma Management Program.

Severe Asthma

The ATS definition of severe asthma was used for participants receiving high-dose ICSs (9), and the NAEPP EPR 3 definition of severe asthma was used for participants not receiving high-dose ICSs (10). The ATS definition of severe refractory asthma requires meeting at least one major criterion (being treated with high-dose ICSs or receiving oral corticosteroids for ⩾50% of the year) and at least two minor criteria (using a daily controller in addition to ICSs, having daily or near-daily asthma symptoms requiring rescue, having a percent-predicted FEV₁ <80%, having one or more urgent care visits for asthma per year, having three or more oral steroid “bursts” per year, having prompt deterioration with a ⩽25% decrease in the oral corticosteroid or ICS dose, and having a near-fatal asthma event in the past). Severe asthma as determined by the NAEPP EPR 3 guidelines for participants not receiving high-dose ICSs included having any of the following: ⩾4 nighttime awakenings, ⩾8 instances of symptoms interfering with activity or requiring albuterol in the last 7 days, an FEV₁/FVC ratio of <0.70, or a percent-predicted FEV₁ <60%. Because the NAEPP EPR 3 guidelines do not provide specific criteria for the risk assessment of severe asthma, participants who were not receiving high-dose ICSs with a hospitalization or >2 ED visits or oral corticosteroid courses for asthma in the past year were also considered to have severe asthma (8). “Remitting asthma” included participants who had none of the following recorded at the selected study visit or any visit over the preceding year: contact with a local medical provider, school absence, ED visit, or hospitalization due to worsening asthma; the presence of wheezing or any exercise-related symptoms; or the use of rescue or controller medications. “Intermittent asthma” included participants who were not using a controller, had no oral corticosteroid use, had no hospitalizations or emergency visits, had no nighttime awakenings, and had fewer than three instances of symptoms interfering with activity or requiring albuterol in the last 7 days. “Persistent asthma” included participants who did not fit the criteria for remitting, severe, or intermittent asthma. A Sankey diagram was produced by using SankeyMATIC (http://sankeymatic.com) to depict the prevalence of these groups over time (Figure 2).

Figure 2.

Sankey diagram of asthma severity changes, demonstrating the proportion of patients with each category of asthma severity over time: at baseline in childhood, late adolescence, and early adulthood. Asthma severity was dynamic from childhood to early adulthood. The data are presented as the mean (SD). ^†“Stayed Severe” refers to patients with a persistence of severe asthma from late adolescence to early adulthood.

We adjusted for the original eight CAMP covariates: age at randomization, sex, ethnicity (binary of White or other), treatment group, clinic of enrollment, duration of asthma, severity of asthma at baseline (mild or moderate persistent asthma as assessed by a physician), and any aeroallergen skin prick test reactivity (20). More than 22 relevant factors were evaluated for determining the persistence of severe asthma from late adolescence to early adulthood, including: secondhand smoke exposure, maternal smoking during pregnancy, body mass index percentile, hay fever, eczema, food allergy, parental asthma, being sensitized and exposed to any perennial aeroallergen, bronchial hyperreactivity, prebronchodilator FEV₁/FVC ratio, postbronchodilator FEV₁/FVC ratio, bronchial hyperresponsiveness to methacholine challenge (log₁₀ of the provocative concentration resulting in a 20% fall in the FEV₁ [PC₂₀]), log₁₀ of the serum total IgE, and log₁₀ of the serum absolute eosinophil count.

Statistical Analyses

The probability that a participant experienced the persistence of severe asthma from late adolescence to early adulthood was analyzed by using a logistic regression model. Logistic regression modeling included all relevant variables described earlier. Stepwise selection was used to select explanatory variables for the final model, with a significance level of 0.2 being used for entry into the model and a significance level of 0.1 being used for staying in the model. A forest plot was created by using univariate logistic regression for exploratory purposes and to illustrate unadjusted odds ratios (95% confidence intervals [CIs]) of the persistence of severe asthma from late adolescence to early adulthood for candidate variables. Characteristics among groups were compared from baseline to late adolescence and from baseline to early adulthood by using repeated-measure analyses such as McNemar’s test, Bhapkar’s test, and linear mixed models. The odds ratios (95% CIs) of persistent severe asthma for participants with different lung function trajectory patterns over time, adjusted for the main eight covariates, were computed by using logistic regression. Briefly, these lung function trajectories identified in another CAMP publication were as follows: “normal growth” (FEV₁ growth curve almost always at or above the 25th percentile) versus “reduced growth” (FEV₁ growth curve almost always below the 25th percentile) and “early decline” (an earlier-than-expected decrease in the FEV₁) versus “no decline” (18). A Bonferroni correction was employed to account for multiple comparisons such that the significance-threshold P value for age-group comparisons in Table 2 was set to 0.05/3 = 0.0167. A Benjamini-Hochberg false discovery rate adjustment was calculated separately for the “Late Adolescence,” “Early Adulthood,” and “Stayed Severe” columns, with thresholds for significance of <0.0228, <0.0131, and <0.0153, respectively, being used (Table 3). Otherwise, statistical significance was set at P < 0.05. SAS statistical software (version 9.4; SAS Institute Inc.) and SPSS statistical software (version 26; IBM) were used for analyses.

Table 2.

Comparison of Clinical Features, Biomarkers, and Lung Function over Time in Participants with Mild to Moderate Asthma (N = 682)

	Childhood*	Late Adolescence†	Early Adulthood‡	P Value for All Groups§	P Value for Childhood vs. Late Adolescence§	P Value for Childhood vs. Early Adulthood§	P Value for Late Adolescence vs. Early Adulthood§
Asthma severity
Remitted	—	24 (3.5)	35 (5.1)	—	—	—	0.101
Intermittent	329 (48.2)	326 (47.8)	319 (46.8)	0.207	0.160	0.077	0.609
Persistent	353 (51.8)	251 (36.8)	197 (28.9)	<0.0001	<0.0001	<0.0001	<0.0001
Severe	—	81 (11.9)	131 (19.2)	—	—	—	<0.0001
Lung function
Pre-BD FEV1% predicted	94.2 ± 13.8	98.0 ± 13.3	95.3 ± 13.4	<0.0001	<0.0001	0.040	<0.0001
Post-BD FEV1% predicted	107.7 ± 13.4	105.0 ± 11.8	101.3 ± 12.6	<0.0001	<0.0001	<0.0001	<0.0001
Pre-BD FVC% predicted	104.6 ± 13.1	110.0 ± 12.1	108.5 ± 12.1	<0.0001	<0.0001	<0.0001	0.0002
Post-BD FVC% predicted	104.0 ± 12.9	110.7 ± 12.0	108.9 ± 11.9	<0.0001	<0.0001	<0.0001	<0.0001
Pre-BD FEV1/FVC, %	79.4 ± 8.1	78.1 ± 9.2	76.2 ± 8.9	<0.0001	<0.0001	<0.0001	<0.0001
Post-BD FEV1/FVC, %	85.2 ± 6.1	83.3 ± 7.9	80.7 ± 7.9	<0.0001	<0.0001	<0.0001	<0.0001
Bronchodilator reversibility, %	11.2 ± 10.6	7.7 ± 7.1	6.8 ± 6.0	<0.0001	<0.0001	<0.0001	0.015
Methacholine FEV1 PC20‖¶	1.1 ± 3.2	4.4 ± 5.5	6.7 ± 5.0	<0.0001	<0.0001	<0.0001	0.0003
Asthma risk
Prednisone courses/yr	1.1 ± 2.0	0.1 ± 0.5	0.1 ± 0.4	<0.0001	<0.0001	<0.0001	0.534
ED visits/yr	0.6 ± 1.3	0.1 ± 0.4	0.1 ± 0.5	<0.0001	<0.0001	<0.0001	0.213
Hospitalizations/yr	0.1 ± 0.2	0.0 ± 0.1	0.0 ± 0.1	<0.0001	<0.0001	<0.0001	0.847
Biomarkers
Serum IgE, ng/ml¶	467 ± 4.7	400 ± 4.2	390 ± 4.3	0.004	0.305	0.001	0.020
Eosinophil count, cells/μl¶	302 ± 4.1	186 ± 2.8	162 ± 3.9	<0.0001	<0.0001	<0.0001	0.225
Medications
No medication, %	—	41	44	—	—	—	0.187
Rescue medications only, %	—	20	25	—	—	—	0.028
Any daily controller therapy, %	—	39	31	—	—	—	<0.0001
ICS monotherapy, %	—	5	4	—	—	—	0.564
ICS + LABA, %	—	4	7	—	—	—	0.003
ICS daily dose**	—	322 ± 243	358 ± 253	—	—	—	0.145

Definition of abbreviations: BD = bronchodilator; ED = emergency department; ICS = inhaled corticosteroid; LABA = long-acting β-agonist; PC₂₀ = provocative concentration resulting in a 20% fall in the FEV₁.

Data are presented as n (%) or mean ± SD unless otherwise indicated. Tests were conducted by using linear mixed models for continuous variables. Bhapkar’s test was used for intermittent/persistent asthma severity to test differences among all age groups. Tests between pairs of age groups were completed by using McNemar’s test for all categorical variables.

^*Age: range, 5–12 years; mean ± SD, 8.7 ± 2.0 years.

^†Age: range, 17–19 years; mean ± SD, 18.3 ± 0.4 years.

^‡Age: range, 21–23 years; mean ± SD, 22.4 ± 0.4 years.

^§P value derived from one-way ANOVA with repeated measures. Boldface P values are <0.0167 to account for pairwise comparisons (0.05/3 = 0.0167).

^‖PC₂₀ represents the geometric mean of the concentration of methacholine needed to produce a 20% fall in the FEV₁ from baseline.

^¶Values in table represent geometric mean ± SD; significance was calculated using log₁₀ values.

^**Fluticasone equivalent of ICS total dose per day in micrograms.

Table 3.

Characteristics of the Selected Participants Who Had Severe, Persistent, Intermittent, and Remitting Asthma

	Late Adolescence: Ages 17–19 yr					Early Adulthood: Ages 21–23 yr					Stayed Severe: Ages 21–23 yr*
	Remitting	Intermittent	Persistent	Severe	P Value	Remitting	Intermittent	Persistent	Severe	P Value	Severe	P Value
Participants	24 (4)	326 (48)	251 (37)	81 (12)	—	35 (5)	319 (47)	197 (29)	131 (19)	—	42 (6)	—
Baseline characteristics
Age, yr	18.2 (0.3)	18.3 (0.3)	18.3 (0.4)	18.3 (0.4)	0.376	22.3 (0.3)	22.4 (0.3)	22.4 (0.4)	22.4 (0.4)	0.552	22.5 (0.4)	0.180
Sex, M	54	61	57	64	0.546	60	61	50	72	0.002	79	0.011
Ethnicity
White	54	69	67	70	0.700	69	63	70	75	0.242	71	0.822
Black	25	15	16	12	0.700	14	18	12	15	0.242	17	0.822
Hispanic	4	10	10	9	0.700	6	11	10	5	0.242	5	0.822
Other	17	7	8	9	0.700	11	8	9	6	0.242	7	0.822
BMI percentile	0.71 (0.29)	0.69 (0.26)	0.66 (0.27)	0.69 (0.28)	0.309	0.74 (0.26)	0.73 (0.28)	0.64 (0.34)	0.67 (0.29)	0.401	0.81 (0.20)	0.133
Duration of asthma, yr	5.07 (2.43)	4.46 (2.73)	4.80 (2.45)	5.07 (2.76)	0.106	5.79 (2.25)	4.38 (2.69)	4.71 (2.45)	5.06 (2.74)	0.003	5.42 (2.64)	0.051
Age at symptom onset, yr	2.64 (2.72)	3.39 (2.72)	3.14 (2.20)	2.60 (2.38)	0.023	1.89 (1.93)	3.48 (2.71)	3.05 (2.33)	2.98 (2.28)	0.002	2.06 (1.82)	0.002
Age at M.D. diagnosis, yr	3.37 (2.96)	4.23 (2.76)	3.92 (2.35)	3.40 (2.32)	0.041	2.86 (2.30)	4.34 (2.78)	3.84 (2.34)	3.68 (2.35)	0.004	2.64 (1.72)	0.093
At least one positive SPT	83	85	93	94	0.014	86	87	93	88	0.162	91	1.000
Sensitized and exposed†	58	69	79	83	0.003	71	72	77	76	0.614	74	0.829
Eczema	25	29	32	22	0.386	23	27	34	27	0.279	24	0.454
Hay fever	42	53	56	47	0.304	54	54	56	47	0.474	33	0.009
Food allergy	13	18	19	17	0.894	14	16	22	18	0.293	17	0.812
Parental asthma	46	40	45	48	0.416	46	40	49	43	0.227	50	0.352
Treatment group in CAMP‡
Budesonide	38	30	29	30	0.873	40	30	29	29	0.517	38	0.248
Nedocromil	17	29	32	30	0.873	37	30	30	28	0.517	19	0.248
Placebo	46	41	39	41	0.873	23	41	41	44	0.517	43	0.248
Lung function
Pre-BD FEV1% predicted	103.7 (13.7)	101.9 (11.2)	95.6 (12.9)	88.2 (15.5)	<0.0001	98.9 (11.7)	99.1 (11.0)	94.5 (14.8)	86.7 (12.8)	<0.0001	81.7 (14.7)	<0.0001
Post-BD FEV1% predicted	107.1 (13.3)	107.5 (10.6)	103.5 (11.7)	99.1 (14.0)	<0.0001	103.7 (11.3)	104.0 (10.9)	101.2 (13.3)	94.6 (12.9)	<0.0001	90.2 (15.7)	0.010
Pre-BD FVC% predicted	106.5 (13.6)	109.3 (11.6)	110.0 (12.0)	113.6 (13.8)	0.063	104.6 (11.3)	107.6 (11.4)	109.3 (12.2)	110.8 (13.2)	0.009	111.9 (14.2)	0.053
Post-BD FVC% predicted	106.0 (13.0)	109.6 (11.4)	111.1 (11.8)	115.0 (13.7)	0.005	104.9 (11.4)	107.7 (11.2)	109.9 (12.0)	111.7 (13.0)	0.001	113.1 (14.3)	0.015
Pre-BD FEV1/FVC ratio	85.6 (7.2)	81.6 (6.6)	76.3 (9.0)	67.8 (10.1)	<0.0001	81.9 (6.4)	80.0 (5.9)	75.0 (9.3)	67.9 (8.2)	<0.0001	62.9 (6.6)	<0.0001
Post-BD FEV1/FVC ratio	88.8 (5.7)	85.9 (6.0)	81.9 (7.9)	75.4 (8.5)	<0.0001	85.5 (5.5)	83.8 (5.2)	79.9 (8.3)	73.4 (7.8)	<0.0001	68.7 (6.8)	<0.0001
Bronchodilator rev., %	3.45 (2.61)	5.75 (4.33)	8.84 (7.44)	13.60 (10.79)	<0.0001	5.00 (3.46)	5.16 (3.88)	7.76 (7.65)	9.38 (6.70)	<0.0001	10.65 (6.85)	<0.0001
Methacholine FEV1 PC20§‖	1.14 (0.61)	0.81 (0.70)	0.43 (0.73)	0.39 (0.75)	<0.0001	1.33 (0.43)	0.91 (0.68)	0.61 (0.69)	0.73 (0.72)	0.002	0.31 (0.76)	0.042
Asthma risk
Prednisone courses/yr	—	—	0.30 (0.65)	0.27 (0.84)	0.096	—	—	0.14 (0.44)	0.33 (0.79)	0.033	0.31 (0.72)	0.001
ED visits/yr	—	—	0.10 (0.51)	0.25 (0.51)	0.0002	—	—	0.13 (0.53)	0.44 (0.83)	<0.0001	0.50 (1.15)	<0.0001
Hospitalizations/yr	—	—	0.01 (0.09)	0.00 (0.00)	0.421	—	—	0.01 (0.07)	0.02 (0.12)	0.343	0.05 (0.22)	<0.0001
Biomarkers
Serum IgE, ng/ml‖	282 (2.9)	277 (4.2)	541 (4.2)	650 (3.7)	<0.0001	176 (4.6)	393 (4.5)	516 (3.0)	338 (5.6)	0.370	498 (10.1)	0.592
Eosinophil count, cells/μl‖	187 (1.6)	172 (2.8)	212 (2.9)	164 (3.3)	0.167	70 (5.8)	139 (4.5)	231 (3.4)	221 (1.7)	0.013	314 (2.4)	0.362
Medications
No medication	100	64	2	54	<0.0001	100	61	3	50	<0.0001	43	0.863
Rescue medications only	0	36	2	21	<0.0001	0	39	2	31	<0.0001	36	0.085
Any daily controller therapy	—	—	97	25	<0.0001	—	—	95	19	<0.0001	21	0.157
ICS monotherapy	—	—	13	0	<0.0001	—	—	15	0	<0.0001	0	0.248
ICS + LABA	—	—	10	0	0.001	—	—	24	2	<0.0001	0	0.064
ICS daily dose¶	—	—	289 (223)	625 (214)	<0.0001	—	—	301 (223)	618 (218)	<0.0001	642 (226)	0.0007

Definition of abbreviations: BD = bronchodilator; BMI = body mass index; CAMP = Childhood Asthma Management Program; ED = emergency department; FDR = false discovery rate; ICS = inhaled corticosteroid; LABA = long-acting β-agonist; M.D. = Doctor of Medicine; PC₂₀ = provocative concentration resulting in a 20% fall in the FEV₁; rev. = reversibility; SPT = skin prick test.

Data are presented as a percentage or as the mean (SD) unless otherwise indicated. Chi-square and Fisher exact tests were used for categorical variables. Kruskal-Wallis tests were used for continuous variables. Bold P values represent significance as determined by using a Benjamini-Hochberg FDR adjustment calculated separately for the “Late Adolescence,” “Early Adulthood,” and “Stayed Severe” columns; the thresholds for significance were <0.0228, <0.0131, and <0.0153, respectively.

^*The “Stayed Severe” column refers to patients who had a persistence of severe asthma from late adolescence to early adulthood.

^†“Sensitized and exposed” refers to indoor allergens. The criteria for significant allergen exposure are as follows: cat allergen concentration >8,000 ng/g or cat allowed inside the home; dog allergen concentration >10,000 ng/g or dog allowed inside the home; mite allergen concentration >2,000 ng/g; cockroach allergen concentration >2 U/g or cockroaches in the home; and >25,000 mold colonies or mold on any surface of the home.

^‡Refers to original treatment group assignment from the Childhood Asthma Management Program Research Group (see Reference 15).

^§PC₂₀ represents the geometric mean of the concentration of methacholine needed to produce a 20% fall in the FEV₁ from baseline.

^‖Values in table represent geometric mean (SD).

^¶Fluticasone equivalent of ICS total dose in micrograms.

Results

Asthma Severity in Childhood, Late Adolescence, and Early Adulthood

The initial CAMP study enrolled only participants with mild and moderate asthma in childhood, defined by the presence of symptoms, the use of an inhaled bronchodilator at least twice weekly, or the use of daily medication for asthma (20). Overall, severe asthma was observed in 81 (12%) participants in late adolescence and 131 (19%) participants in early adulthood. In late adolescence, 77 (11%) participants were receiving high-dose ICSs and 20 (3%) participants met ATS criteria for severe asthma, whereas 605 (89%) participants were not receiving high-dose ICSs and 61 (9%) participants met NAEPP EPR 3 criteria for severe asthma. In early adulthood, 59 (9%) participants were receiving high-dose ICSs and 25 (4%) participants met ATS criteria for severe asthma, whereas 623 (91%) participants were not receiving high-dose ICSs and 106 (16%) participants met NAEPP EPR 3 criteria for severe asthma.

Asthma severity fluctuated from childhood to early adulthood (Figure 2). About half of the participants with severe asthma in late adolescence continued to experience asthma severity into early adulthood (42 participants, or 6% of all participants). From late adolescence to early adulthood, the severity classification improved in 123 participants (18%), the severity classification did not change in 414 participants (61%), and the severity classification worsened in 145 participants (21%). Remitting asthma represented the smallest proportion, consisting of 24 participants (3%) in late adolescence, which increased over time to 35 participants in early adulthood (5%).

Comparison of Relevant Features in Childhood, Late Adolescence, and Early Adulthood

We compared clinical features, biomarkers, and lung function in childhood, late adolescence, and early adulthood (Table 2).

The postbronchodilator percent-predicted FEV₁ and pre- and postbronchodilator FEV₁/FVC ratio significantly decreased from childhood to late adolescence and from late adolescence to early adulthood (P < 0.0001). Prebronchodilator FEV₁ values increased from childhood to late adolescence (P < 0.0001) and remained stable into early adulthood (P = 0.040). Bronchodilator reversibility decreased from childhood to late adolescence (P < 0.0001) and decreased from childhood to early adulthood (P = 0.015). Bronchial hyperresponsiveness substantially decreased (increased geometric mean of the methacholine FEV₁ log₁₀ of the PC₂₀) from childhood to late adolescence (P < 0.0001) and substantially decreased from childhood to early adulthood (P = 0.0003). The numbers of prednisone courses, ED visits, and hospitalizations were highest in childhood (P < 0.0001) but remained stable between late adolescence and early adulthood. Eosinophil counts decreased from childhood to late adolescence (P < 0.0001) but did not decrease from late adolescence to early adulthood (P = 0.225). Serum IgE amounts only decreased from childhood to early adulthood (P = 0.001). Only 39% and 31% of the participants were receiving daily controller therapy in late adolescence and in early adulthood, respectively.

We also compared clinical features, biomarkers, and lung function by severity status (Table 3). Participants with severe asthma in late adolescence and early adulthood had decreased pre- and postbronchodilator FEV₁ (P < 0.0001) values, decreased postbronchodilator FVC values, decreased pre- and postbronchodilator FEV₁/FVC ratios (P < 0.0001), and increased bronchodilator reversibility (P < 0.0001) compared with the other groups without severe asthma (Table 3). Bronchial hyperresponsiveness (a lower methacholine PC₂₀) was higher in those with persistent and severe asthma (P < 0.0001) (Table 3; see Figure E1 in the online supplement). The numbers of ED visits per year were higher in participants with severe asthma (P = 0.0002, P < 0.0001). Participants with persistent and severe asthma had higher serum IgE in late adolescence and higher eosinophil counts in early adulthood. Participants with severe and persistent asthma in late adolescence were more likely to have aeroallergen skin test positivity and to be sensitized and exposed to perennial allergens in childhood. Participants with severe asthma in early adulthood were predominantly male. Participants with persistent asthma were much more likely to be using a daily controller than were those with severe asthma (P < 0.0001). However, those with severe asthma who were using a controller were receiving more than twice as many total ICS doses per day (P < 0.0001), and participants with a persistence of severe asthma had the highest ICS dose per day (P = 0.0007).

Factors in Early Childhood Predictive of Severe Asthma Later in Life

By using univariate logistic regression models, significant variables in early childhood that were associated with the persistence of severe asthma from late adolescence to early adulthood were identified. These included a younger age (odds for 1-yr increase, 0.84; 95% CI, 0.72–0.99; P = 0.0350), male sex (odds ratio, 2.59; 95% CI, 1.22–5.51; P = 0.0133), no reported hay fever (odds ratio, 0.42; 95% CI, 0.22–0.82; P = 0.0107), a low prebronchodilator FEV₁/FVC ratio (odds ratio for an increase of 5%, 0.58; 95% CI, 0.48–0.71; P < 0.0001) and a low postbronchodilator FEV₁/FVC ratio (odds ratio for an increase of 5%, 0.49; 95% CI, 0.38–0.63; P < 0.0001), and a physician diagnosis of moderate (vs. mild) asthma during the screening period (odds ratio, 0.50; 95% CI, 0.26–0.95; P = 0.0347) (Figure 3).

Figure 3.

Forest plot of relevant childhood factors and the persistence of severe asthma from late adolescence to early adulthood. By using univariate logistic regression models, significant predictors of severe asthma from late adolescence to early adulthood were found to be as follows: male sex (odds ratio, 2.59; 95% confidence interval [CI], 1.22–5.51; P = 0.0133), hay fever (odds ratio, 0.42; 95% CI, 0.22–0.82; P = 0.0107), prebronchodilator FEV₁/FVC ratio (odds ratio for increase of 5%, 0.58; 95% CI, 0.48–0.71; P < 0.0001), postbronchodilator FEV₁/FVC ratio (odds ratio for increase of 5%, 0.49; 95% CI, 0.38–0.63; P < 0.0001), M.D. baseline diagnosis of mild versus moderate asthma (0.50; 95% CI, 0.26–0.95; P = 0.0347), and age at randomization (odds for 1-yr increase, 0.84; 95% CI, 0.72–0.99; P = 0.0350). The variables that were analyzed but are not included in the figure include the enrollment clinic. M.D. = Doctor of Medicine; PC₂₀ = provocative concentration resulting in a 20% decrease in the FEV₁.

After running stepwise selection on all candidate variables with a forced adjustment of the eight covariates, the postbronchodilator FEV₁/FVC ratio remained a significant predictor in our multivariate model: for every 5% decrease in the postbronchodilator FEV₁/FVC ratio in childhood, the odds of the persistence of severe asthma increased by 2.36-fold (95% CI, 1.70–3.28; P < 0.0001). The other significant predictor identified by our multivariate model was maternal smoking during pregnancy, which increased the odds of the persistence of severe asthma by 3.17-fold (95% CI, 1.18–8.53; P = 0.02). The following predictors were also identified by our multivariate model but did not reach significance: parental asthma and hay fever (Table 4).

Table 4.

Baseline Predictors of the Persistence of Severe Asthma from Late Adolescence to Early Adulthood (N = 682)

Characteristic	Persistence of Severe Asthma [OR (95% CI)]	P Value
Postbronchodilator FEV1/FVC ratio	0.84 (0.79–0.90)	<0.0001
Maternal smoking during pregnancy	3.17 (1.18–8.53)	0.02
Parental asthma	2.09 (0.97–4.50)	0.06
Hay fever	0.50 (0.22–1.12)	0.09

Definition of abbreviations: CI = confidence interval; OR = odds ratio; PC₂₀ = provocative concentration resulting in a 20% fall in the FEV₁.

Multinomial logistic regression includes all predictors that resulted from our model by using stepwise selection while adjusting for the following covariates: age at randomization, sex, ethnicity (White or other), treatment group, clinic of enrollment, duration of asthma, severity at baseline (mild or moderate persistent asthma as assessed by a physician), and any environmental skin prick test reactivity. Candidate variables included the following: secondhand smoke exposure, maternal smoking during pregnancy, body mass index percentile, hay fever, eczema, food allergy, parental asthma, sensitized and exposed to any environmental allergens, bronchial hyperreactivity, prebronchodilator FEV₁/FVC ratio, postbronchodilator FEV₁/FVC ratio, bronchial hyperresponsiveness to methacholine challenge (log₁₀ of PC₂₀), serum total IgE (log₁₀), and serum absolute eosinophil count (log₁₀).

Association of Severe Asthma and Lung Function Trajectory Patterns

In late adolescence, 78% of participants with severe asthma had reduced growth, and 42% of those participants had reduced growth and early decline (Table 5). In early adulthood, 70% of participants with severe asthma had reduced lung growth, and 53% of those participants had reduced growth and early decline. In those whose asthma remained severe, 96% had reduced growth; of these, 40% had reduced growth and early decline. Both in late adolescence and early adulthood, more participants with persistent and severe asthma had reduced growth than normal growth (Figure 4; see Figure E2 in the online supplement); in contrast, there were more participants with remitting and intermittent asthma who had a normal growth pattern.

Table 5.

Distribution of Lung Function Trajectory Patterns in Patients with Severe Asthma

Lung Function Trajectory	Total Number*	Late Adolescence† [n (%)]	Early Adulthood‡ [n (%)]	Stayed Severe§ [n (%)]
Normal growth	21	6 (11)	14 (14)	1 (4)
Normal growth with early decline	23	6 (11)	17 (17)	0 (0)
Reduced growth	74	25 (45)	34 (33)	15 (58)
Reduced growth with early decline	66	18 (33)	38 (37)	10 (38)

Lung function trajectory patterns were as follows: normal growth (FEV₁ growth curve almost always at or above the 25th percentile) versus reduced growth (FEV₁ growth curve almost always below the 25th percentile) and early decline (earlier-than-expected decrease in FEV₁) versus no decline (see Reference 18).

^*N = 184 with available data.

^†Age: 17–19 years; n = 55.

^‡Age: 21–23 years; n = 103.

^§The “Stayed Severe” column refers to patients aged 21–23 years (n = 26) who had a persistence of severe asthma from late adolescence to early adulthood.

Figure 4.

Asthma severity by lung function trajectory pattern. The distribution of participants with remitting, intermittent, persistent, and severe asthma is different between participants with normal lung growth patterns and those with reduced lung growth patterns. Lung function trajectory patterns were as follows: “normal growth” (FEV₁ growth curve almost always at or above the 25th percentile) versus “reduced growth” (FEV₁ growth curve almost always below the 25th percentile) and “early decline” (earlier-than-expected decrease in the FEV₁) versus “no decline” (see Reference 18).

We found a strong association between a reduced-growth lung function trajectory and the persistence of severe asthma by using a logistic regression analysis with adjustment for the eight main clinical and demographic covariates. Compared with normal growth, the odds of severe asthma persisting from late adolescence to early adulthood were significant for both reduced growth without early decline (odds ratio, 16.38; 95% CI, 2.03–132.30; P = 0.0087) and reduced growth with early decline (odds ratio, 13.13; 95% CI, 1.56–110.61; P = 0.0179) but were not significant for normal growth with early decline.

Discussion

Reports of longitudinal cohorts tracking asthma severity over time and featuring multiple characterization elements are limited (17, 21–23). The variability of severity over time supports asthma’s heterogeneity, and here, we show that asthma has a dynamic course from childhood to early adulthood. Asthma remission in late adolescence and early adulthood has been a major focus of research (17), but during this critical period—when many adolescents and young adults believe that their asthma has gone away—there is still a significant minority who suffer from severe disease. We found that 12–19% of participants who had mild to moderate asthma in early childhood and were followed over 10–18 years had features consistent with severe asthma between 17 and 23 years of age. Six percent had severe asthma both during adolescence and during early adulthood. In our univariate analysis of over 20 relevant factors in early childhood, the following factors were predictors at baseline: younger age, male sex, the absence of hay fever, a low pre- and postbronchodilator FEV₁/FVC ratio, and asthma severity at baseline. Long-term treatment with an ICS, the mainstay therapy for persistent asthma, was not a significant factor associated with asthma severity. Maternal smoking during pregnancy and a low postbronchodilator FEV₁/FVC ratio were found to be predictors of the persistence of severe asthma from late adolescence to early adulthood in our multivariate model. We also determined that reduced growth from childhood to early adulthood was associated with severe asthma.

Asthma activity over time was dynamic (Figure 2). Although most participants with asthma improved from late adolescence to early adulthood, approximately half did not change severity status, and one-fifth worsened. On the basis of our findings, clinicians should consider asthma severity to be more fluid than static, especially when counseling participants about their long-term prognosis.

We found that as many as 12–19% of participants had features of severe asthma between 17 and 23 years of age on the basis of our definition, but likely only 6% had severe asthma that persisted from adolescence to early adulthood. This is consistent with the published estimates of 5–10% (5, 6), which lends support to our selection of elements used to identify these individuals. Those with severe asthma also had reduced lung function, had increased bronchodilator reversibility, had more ED visits, and were on higher ICS doses, further supporting our definition (Table 3). Baseline characteristics were generally similar among severity groups. Individuals with severe asthma in early adulthood and those with persistence of severe asthma from late adolescence to early adulthood were predominantly male. In our univariate analysis, male sex was also a predictive factor for the persistence of severe asthma. A large proportion of our participants were undertreated and were classified as having severe asthma on the basis of lung function, which may explain the male predominance. Although adult women have a higher prevalence of asthma and more frequent use of asthma-related health care, they also have better lung function when compared with adult men (24). Furthermore, McGeachie and colleagues (18) previously found male sex to be a predictor of lung function decline in the CAMP cohort. It is unclear what drives lower lung function in males, but hormones may play a role (25).

Studies regarding factors collected in early childhood that influence or predict severe asthma status in adulthood are limited. In this study, which spans late adolescence and early adulthood, we identified early childhood evidence that a low postbronchodilator FEV₁/FVC ratio and maternal smoking during pregnancy were significant childhood predictors of severe asthma. The postbronchodilator FEV₁/FVC ratio is a noninvasive marker of pathologic airway remodeling in asthma (26, 27). Here, we demonstrate that for every 5% decrease in the postbronchodilator FEV₁/FVC ratio, participants were 2.36-fold more likely to have a persistently severe phenotype during a stage of life when asthma usually improves.

Lung function in childhood to adolescence and adulthood has been used as a surrogate of airway remodeling (18, 28, 29). Pathologic airway remodeling in asthma may lead to irreversible airflow obstruction and poor outcomes (30, 31). Unfortunately, little is known about the mechanisms that induce pathologic airway remodeling (32, 33). Recent studies have focused on early-life influences during asthma inception that may lead to irreversible lung function deficits, such as the microbial environment of the airway and gut or patterns of weight gain (34). Those in our cohort were mostly atopic at baseline and at the time of assessment in late adolescence and early adulthood. Those with severe and persistent asthma had higher serum IgE concentrations in late adolescence and higher eosinophil numbers in early adulthood than participants with intermittent and remitting asthma. Although biological therapies targeting type 2 inflammation have shown promise in improving asthma outcomes, less is known about their effect on long-term airway remodeling. Some studies have hinted at a reversal of airway remodeling, particularly with omalizumab, which has been the most studied (35). More research is needed to fully address whether—and to what extent—biological therapy of asthma can affect long-term airway remodeling.

Participants whose mothers smoked during pregnancy had a 3.17-fold–higher likelihood of having persistently severe asthma from late adolescence to early adulthood. Maternal smoking during pregnancy has been associated with an increased incidence of asthma (36) and decreased lung function in childhood (36, 37). Here, we suggest that maternal smoking may also be an important prognostic factor for the persistence of severe asthma. Smoking exposure in utero may start pathologic lung remodeling earlier, and some evidence suggests that this may be due to epigenetic modifications (38). Although smoking prevention in pregnancy to prevent asthma is clearly desired, we add that individuals with asthma whose mothers smoked during pregnancy may also need closer monitoring.

Among participants with severe asthma in late adolescence and early adulthood, over two-thirds had a reduced-growth trajectory pattern (Table 5, Figure 4, Figure E2). Furthermore, among participants whose asthma remained severe between late adolescence and early adulthood, almost all (96%) had reduced lung growth. Participants with a trajectory pattern of reduced growth and participants with a trajectory pattern of reduced growth with early decline were 16.38 times more likely (95% CI, 2.03–132.30; P = 0.0087) and 13.13 times more likely (95% CI, 1.56–110.61), respectively, to have persistence of severe asthma than participants with normal growth. Patients with normal growth but early decline did not have an increased likelihood of experiencing a persistence of severe asthma. Severity is likely driven by lung function, and an earlier start to the decline (reduced growth vs. normal growth) was more predictive than the trajectory pattern (early decline vs. no decline) for the persistence of severe asthma. A meta-analysis found irreversible airflow obstruction to be much more common in severe asthma (30). These participants with severe asthma may represent those who progress further and have worse outcomes over time. McGeachie and colleagues (18) evaluated the predictors of lung function decline in the CAMP study. They reported the following predictors of early decliners compared with those who did not decline: lower baseline FEV₁ values, a smaller bronchodilator response, airway hyperresponsiveness at baseline, maternal smoking during gestation, and male sex. We found that many of these factors also predict the persistence of severe asthma. Taken together, it may be prudent to check lung function as early as possible and to monitor those with asthma more closely if they start with low function in childhood.

An earlier age at randomization, a lower prebronchodilator FEV₁/FVC ratio at baseline, and a physician diagnosis of moderate versus mild asthma at baseline were significant predictors of the persistence of severe asthma in our univariate model. The results of the univariate analysis suggest that features of persistent asthma in younger patients may lead to earlier lung function impairment and worse outcomes. Parental asthma in childhood was identified as a predictor of the persistence of severe asthma in our multivariate model, although this was not quite significant (P = 0.06). Parental asthma is a well-established predictor of asthma diagnosis and is commonly used in predictive measures such as the Asthma Predictive Index (39). Here, we postulate that patients with a genetic predisposition toward asthma may have an earlier onset of lung function decline, which may affect the persistence of a severe phenotype. Collinearity with the asthma onset–related covariates mentioned earlier may explain why the prebronchodilator FEV₁/FVC ratio and parental asthma were not identified and did not reach significance, respectively, in our multivariate model.

Hay fever as a protective factor for the persistence of severe asthma from late adolescence to early adulthood was significant only in our univariate model. Hay fever in our study represents seasonal rhinitis symptoms, as it was evaluated in terms of whether the child had any eye itchiness, runny nose symptoms, or sneezing that recurred over several weeks in a particular season at any time for at least 2 consecutive years. Substantial evidence exists that allergic rhinitis leads to an increased risk of asthma (40) and worse asthma outcomes (41). An adjustment for aeroallergen sensitization in the multivariate model likely led to lack of significance in our multivariate model (P = 0.09). The reporting of hay fever in the child may simply represent a more observant or involved parent, which may explain its protective quality in the univariate analysis.

Treatment with an ICS for at least 4 years during the CAMP study did not affect the development of more severe disease later on in life. Treatment with ICS therapy also did not modify the lung function trajectory (18) or the incidence of asthma remission in early adolescence (16). However, shared decision-making discussions with the families of children with persistent asthma should include an emphasis on the fact that the therapy available for asthma at this time is aimed at reducing the immediate asthma burden (such as symptom improvement or preventing exacerbations) and that whether long-term goals such as reducing airway remodeling or preventing a worsening of asthma later on in life will be achieved is not predictable.

Our study had several strengths. The CAMP cohort was well characterized from beginning to end, with a majority having long-term follow-up for an average of 14 years. The detailed longitudinal follow-up reduced observational bias and allowed control for many relevant covariates from childhood to adolescence and early adulthood.

There were limitations to our study. This is a post hoc analysis, and not all measurements were performed at every visit. We were only able to evaluate 682 participants because they had data available in late adolescence and early adulthood. This number matches CAMP studies that have analyzed data from the original cohort and have evaluated the long-term lung function trajectory by using outcome measures evaluated in early adulthood (17). In that study, the characteristics of the 684 participants with at least one spirometric measurement at the age of 23 years or older—as compared with the remaining 357 participants—were analyzed. Multiple logistic regression analyses that compared baseline characteristics of completers (n = 681) with noncompleters (n = 351) were performed. Participants with data at the age of 23 years or older (completers) were older (P < 0.0001) and were more likely to be prepubertal at baseline (P = 0.008) than the remaining participants (n = 355). Otherwise, they were similar with respect to sex, race or ethnic group, asthma severity, the interval between asthma diagnosis and enrollment, the atopy status, and the assigned treatment group. Methacholine challenge results were only available for 437 (64%) and 326 (48%) participants in the late adolescence and early adulthood groups, respectively. A definition of remitting asthma on the basis of methacholine challenge results would be limited. More notable changes and measurements of exposure in the environment, such as acquiring new pets, home water damage, or moving to another city, were not quantified in the observational phase in CAMP. Results may have been different if other criteria for severe asthma had been used. We used the ATS definition of severe asthma for participants receiving a high dose of ICSs to best capture therapy-resistant severe asthma (9). The ATS definition of severe asthma was reevaluated in a combined effort with the European Respiratory Society in 2014 (10). We did not use the European Respiratory Society–ATS criteria because they use subjective questionnaires not collected in our cohort and were published after the CAMP cohort observational period. Furthermore, to capture undertreated asthma—a common problem in adolescent populations (42) and pediatric populations (43–45)—and to match criteria used in our studied cohort previously (16), we used the NAEPP EPR 3 definition of severe asthma with refined risk criteria for patients not receiving a high dose of ICSs (8).

Previous CAMP studies focused on remission rather than on severe asthma. Our remission rate of 3–5% was similar to previous estimates (16, 17). The more recent study by Wang and colleagues (17) found a higher remission rate of 15% in adulthood (ages of 18–23 yr) when compared with our study, which found a 12% remission rate in early adulthood (ages of 21–23 yr). However, this is likely due to school absenteeism or the lack of accessing medical care in the criteria from Wang and colleagues, which were included in the present study to better match the previous CAMP publications (16).

In conclusion, patients who had lower postbronchodilator lung function and had experienced maternal smoking during pregnancy were more likely to have persistence of severe asthma through the critical period of late adolescence to early adulthood. Early lung function decline likely drives severity. These findings should be validated in other longitudinal cohorts from childhood to early adulthood. Immunologic changes or lung remodeling may begin earlier than the onset of asthma symptoms. Future studies should focus on interventions to preserve lung function that may prevent and reverse disease progression.