Age of Pseudomonas aeruginosa Acquisition and Subsequent Severity of Cystic Fibrosis Lung Disease

Summary

Rationale: Pseudomonas aeruginosa (Pa) is associated with poor pulmonary outcomes in cystic fibrosis (CF), but the association between age of Pa infection and severity of subsequent lung disease has not been thoroughly investigated. Objective: Our goal was to determine the association between age of Pa acquisition and subsequent severity of CF lung disease. Methods: Case–control study using CF Foundation Registry data of 629 ΔF508 homozygotes with severe and mild lung disease (FEV₁ in the lowest and highest quartile of birth cohort, respectively). Multivariate logistic regression was performed to determine the association between age of Pa acquisition and lung disease severity. Results: Earlier age of Pa infection was strongly associated with increased odds of severe lung disease. For first and persistent Pa, adjusted odds ratios for severe lung disease were 6.5 (95% CI 3.1, 13.7; P < 0.0001) and 11.2 (5.4, 23.1; P < 0.0001), respectively, for subjects with infection before age 5 versus at ≥10 years; the association was stronger in females than males. Conclusions: Earlier Pa infection, particularly before 5 years of age, is strongly associated with severe CF lung disease later in life. This study is not designed to determine causality; Pa infection may be causing lung injury, or may be a marker of ongoing inflammation and lung damage in young children with CF.

INTRODUCTION

Pseudomonas aeruginosa (Pa) is a significant pathogen in cystic fibrosis (CF) lung disease^1–3; its prevalence in respiratory cultures increases with age, reaching approximately 80% by adulthood.⁴ Factors associated with acquisition of Pa in patients with CF include female gender, pancreatic exocrine insufficiency, viral infections (particularly respiratory syncitial virus), socioeconomic status, and previous infection with Staphylococcus aureus (S. aureus).^5–9 Infection with Pa, particularly the mucoid strain,¹⁰ has been associated with decline in lung function and/or more severe lung disease, and earlier infection has been associated with lower forced expiratory volume in 1 sec (FEV₁) and higher risk of death in childhood.^4,10–18 While there is considerable evidence supporting an association between Pa infection and more severe lung disease, previous studies have had several limitations, including small sample size, limited follow-up time, inclusion of subjects with varied CFTR genotype, and a multivariable analysis not designed to specifically assess the relationship between Pa infection and lung disease.

Our objective was to determine the association between age of Pa infection in childhood and severity of CF lung disease later in life in a large population of ΔF508 homozygotes with an extended duration of respiratory culture data. Some of the results of this study have been previously reported in the form of an abstract.¹⁹

METHODS

Objective and Design

We performed a case–control study of subjects with CF enrolled in the University of North Carolina and Case Western Reserve University Gene Modifier Study (GMS),²⁰ which was designed to study CF modifier genes employing an extremes-of-phenotype design, with subjects with severe and mild lung disease serving as cases and controls, respectively. The University of North Carolina at Chapel Hill Institutional Review Board approved the current study; the GMS study was approved by the human studies or ethical review board at each participating institution, and informed consent was obtained from all subjects prior to enrollment.

Population

Patients enrolled in the GMS were ΔF508 homozygotes and had either (a) lung function (FEV₁) in the lowest quartile for birth cohort and were 8–25 years old or (b) lung function in the highest quartile and were ≥age 15. Siblings and subjects post lung transplantation or lobectomy were excluded from the GMS.²⁰ To avoid confounding by changes in treatment practices over time, we included only “young mild” (enrolled at age 15–28, n = 403) and severe subjects (enrolled at age 8–25, n = 406) who were enrolled in the GMS between January, 2000 and March, 2007.²⁰ Our analysis was restricted to the 629 subjects for whom CF Foundation Patient Registry (CFFPR) data was available and health insurance status was known (Fig. 1).

Fig. 1.

Subject enrollment criteria and diagram of study population. Subjects were included in our analysis if they (a) met Gene Modifier Study (GMS) enrollment criteria as either severe or young mild subjects, (b) had CF Foundation Patient Registry (CFFPR) data available, and (c^*) health insurance status at the time of CFFPR enrollment was not missing or classified as “other,” “none,” or “unknown.” †One mild subject was excluded for lack of TGF-β genotype data.

Data

Data was obtained from the GMS [from which disease severity, age of GMS enrollment, and transforming growth factor-β (TGF-β) codon 10 genotype were assigned] and the CFFPR (all other data, including standardized clinical, demographic, and respiratory culture data).^20,21

Variable Definitions

Age of Pa infection (primary independent variable) was defined in two ways: (1) age of first Pa-positive culture, and (2) age of persistent Pa, defined as the age of the first of three consecutive years with Pa-positive cultures in at least 2 of the 3 years. Both variables were coded in 5 year blocks: Pa infection at age 0–4 years, 5–9 years, and the referent group consisting of those with infection at ≥10 years, or who never met Pa-infection criteria. Our outcome of interest was severity of CF lung disease in childhood and young adulthood (age 8–28) coded as severe or mild (Fig. 1).²⁰

Potential confounders included age of first respiratory culture data, years of culture data, birth year, age of CF diagnosis, TGF-β codon 10 genotype, gender, symptoms at diagnosis (respiratory and gastrointestinal vs. other), S. aureus infection (coded 3 ways: age of first S. aureus infection, age of persistent S. aureus infection, and S. aureus before or after first Pa), health insurance status (public vs. private), and CF center at time of CFFPR enrollment. Because of the potential for confounding by differences in management practices, CF centers (46 in total) were classified based on the centers’ proportion of severe and mild subjects enrolled in the GMS [>60% severe; >60% mild; or even (referent)]. All time and age variables were coded in years.

Analysis

Bivariate analyses were performed using Student’s t-tests and χ²-tests, as appropriate, with a P-value < 0.05 considered significant.²² Formal analysis was performed using multivariate logistic regression including potential confounders (listed above), identified a priori based on literature review, in the initial model.²² Analysis of the association of age of Pa infection with severity of disease was performed using age of Pa infection as both a continuous and categorical variable. For categorical coding, both 3 and 5 year blocks were tested, with cutoff for the referent group ranging from 10 to 15 years for age of Pa infection. The most precise classification (based on confidence limit ratio) was chosen for use in the final analysis (data not shown).

When testing for effect measure modification, a change in estimate (odds ratio; OR) of >10% was considered significant. Backwards elimination was employed to determine variables (potential confounders) included in the final model, with a change in estimate of >10% considered significant. The same rule was applied to sensitivity analyses to determine if changes in variable definition for CF center and symptoms at presentation affected strength of the observed association between age of Pa and lung disease severity. All analyses were performed using Stata^® version 10 (www.stata.com).

RESULTS

Study Population and Demographics

Final analysis included 310 subjects with severe disease and 319 with mild disease (Fig. 1). Summary data for severe and mild subject groups are shown in Table 1. There was no significant difference in gender distribution between the two groups. Subjects with severe lung disease were born later, diagnosed at a younger age, had fewer years of culture data, and enrolled in the GMS at a younger age than those with mild disease, likely all secondary to study design (enrollment ages).

TABLE 1.

Demographics and Culture Data

	Severe (n = 310) (SD)	Mild (n = 319) (SD)
Demographics
Female	49%	46%
Year of birth	1987 (5.1)1	1982 (4.3)
Age (years) at CF diagnosis	1.3 (2.6)2	2.1 (4.1)
Age (years) at GMS enrollment	16.4 (4.5)1	20.8 (3.9)
Years of culture data	16.0 (4.4)1	18.2 (3.8)
Age of transition3 from OP to sputum cultures	7.6 (4.0)1	11.7 (5.6)
Respiratory culture data
Age (years) first Pa-positive culture	5.8 (4.6)1	10.8 (5.7)
Age (years) persistent Pa-positive cultures	6.6 (4.7)1	12.3 (5.5)
Years from first culture data to first Pa-positive culture	2.8 (3.4)1	5.6 (5.0)
Never Pa positive by “first culture” definition	8 (3%)	17 (5%)
Never Pa positive by “persistent” definition	25 (8%)1	48 (15%)

¹P < 0.0001 (Student’s t-test).

²P < 0.05 (Student’s t-test).

³Transition age was defined as the age at the time of the first of three consecutive years with ≥1 sputum culture reported in at least 2 of the 3 years.

Age of Pseudomonas aeruginosa and Severity of Lung Disease

Mean Age of Pa Infection (Bivariate Analysis)

Infection with Pa occurred approximately 5 years earlier in severe subjects than mild subjects, whether comparing age of initial or persistent Pa infection (P < 0.0001 for both; Table 1). Mild subjects were also more likely than severe subjects to never have persistent Pa infection (15% and 8% of subjects, respectively; P < 0.0001). The average time from first culture data to first Pa-positive culture was 2.8 versus 5.6 years in the severe versus mild group (P < 0.0001; Table 1, Fig. 2). Figure 2 plots the cumulative percentage of subjects with acquisition Pa by age, highlighting the pattern of earlier Pa infection in subjects with severe lung disease. Mean age of transition from oropharyngeal (OP) to sputum cultures—defined as the age at the time of the first of three consecutive years with ≥1 sputum culture reported in at least 2 of the 3 years—was significantly younger in severe than in mild subjects (7.6 vs. 11.7 years of age, respectively; P < 0.0001).

Fig. 2.

Cumulative percentage of subjects ever reported positive for Pseudomonas aeruginosa (Pa) by age and lung disease severity. Subjects were considered Pa positive at a given age if they had a history of at least one culture positive for Pa by that age.

Age of Pa Infection as a Categorical Variable (Multivariate Analysis)

In developing a multivariate model of the effect of age of Pa acquisition on lung disease severity, we evaluated different classifications of the primary independent variable (age of Pa infection). We compared age of Pa infection as a continuous variable and as a categorical variable in 3 or 5 year blocks with the referent group composed of subjects with age of Pa infection at ≥10, 12, or 15 years. Similar patterns of association between age of Pa infection and severity of lung disease were seen regardless of the classification scheme employed (data not shown). Our final classification of Pa infection in 5 year blocks, with subjects with Pa infection at ≥10 years of age (or who were never Pa positive) as the referent group was chosen to obtain the most precise (smallest confidence limit ratio) estimate of association, and to allow for nonlinearity (compared to classification of age of Pa infection as a continuous variable) of the relationship between age of Pa infection and disease severity. Because a small percentage of subjects never met criteria for first or persistent Pa infection, these subjects were included in the referent group (see Table 1).

When age of Pa infection was treated as a categorical variable in 5 year blocks, there was a strong association between age of first Pa infection and severity of lung disease later in life both in crude analysis and when adjusted for years of culture data, age of first culture, and health insurance status, which served as our final model (Table 2). Subjects with a first Pa-positive culture at 0–4 years of age had 5.9 (95% CI 3.2, 10.9; P < 0.0001) times the odds of severe disease compared to the referent group (those with first Pa at ≥10 years, or never Pa positive); subjects with their first Pa-positive culture at 5–9 years had 3.5 (95% CI 2.1, 5.8; P < 0.0001) times the odds of severe disease. An even stronger association was seen when comparing age of persistent Pa infection. Subjects with persistent Pa at 0–4 years of age had 11 (95% CI 5.5, 22.1; P < 0.0001) times the odds of severe disease as the referent group; those with persistent Pa at 5–9 years had 4.3 (95% CI 2.7, 6.8; P < 0.0001) times the odds of severe disease. S. aureus infection did not have a significant impact on the observed association between age of Pa and severity of CF lung disease (data not shown). Alternate definitions (sensitivity analyses) for CF center and symptoms at diagnosis did not have a significant impact on the association of interest.

TABLE 2.

Association Between Age of Pseudomonas aeruginosa Infection and Severity of Lung Disease

	Odds ratio (95% CI) for severe lung disease
	Crude	Adjusted1
Age of first Pseudomonas aeruginosa
0–4 years	9.6 (6.1, 15.0)2	5.9 (3.2, 10.9)2
5–9 years	3.2 (2.2, 4.8)2	3.5 (2.1, 5.8)2
10 years or older (or never Pa positive)	1 (referent)	1 (referent)
Age of persistent Pseudomonas aeruginosa
0–4 years	18.6 (10.3, 33.9)2	11 (5.5, 22.1)2
5–9 years	4.3 (3.0, 6.4)2	4.3 (2.7, 6.8)2
10 years or older (or never persistent Pa)	1 (referent)	1 (referent)

¹Adjusted for years of culture data, age of first culture data, CF center, and type of health insurance.

²P < 0.0001, compared to referent group.

Age of Pa Infection by Culture Type

A similar pattern of association between age of Pa infection and severity of CF lung disease was seen when analysis was restricted by culture type (years with only OP culture data and years with at least one sputum culture reported; data not shown). Specifically, subjects with severe disease had earlier onset (approximately 5–6 years) of Pa infection when analysis was restricted to either years with only OP culture data or years with at least one sputum culture reported (sputum culture data only or OP and sputum culture data available).

Impact of Gender on Association (Effect Measure Modification)

Because female gender has been associated with poorer pulmonary outcomes in CF,^{5–7,12,14,21,23} we assessed for a possible interaction effect (effect measure modification) of gender on the association between age of Pa infection and severity of lung disease (Table 3). The association between age of persistent Pa infection and severity of lung diseasewas significantly stronger in females than in males, despite the fact that there was no significant difference in mean age of diagnosis between females and males [1.75 (±3.4) and 1.73 (±3.5) years of age, respectively; P = 0.9]. The adjusted ORs for severe lung disease for subjects with persistent Pa between 0 and 4 years of age and between 5 and 9 years of age were approximately 9 times higher in females than in males, both compared to the same-sex referent group (P ≤ 0.003). A similar pattern was seen when comparing gender differences in the association between age at first Pa infection and lung disease severity (Table 3).

TABLE 3.

Association Between Age of Pseudomonas aeruginosa Infection and Severity of Lung Disease by Gender

	Adjusted* odds ratio (95% CI) for severe lung disease
	Females (n = 301)	Males (n = 328)
Age of first Pseudomonas aeruginosa
0–4 years	9.6 (4.1, 22)1,2 (n = 91)	4.0 (1.8, 8.8)1 (n = 88)
5–9 years	7.8 (3.7, 16.3)1,3 (n = 99)	1.8 (0.9, 3.4) (n = 105)
10 years or older (or never Pa positive)	1 (referent)	1 (referent)
Age of persistent Pseudomonas aeruginosa
0–4 years	39.7 (12, 131.8)1,3 (n = 54)	4.4 (1.9, 10.3)1 (n = 65)
5–9 years	11.7 (5.9, 23.5)1,3 (n = 100)	1.7 (0.9, 3.3) (n = 96)
10 years or older (or never Pa positive)	1 (referent)	1 (referent)

^*Adjusted for years of culture data, age of first culture data, CF center, and type of health insurance.

¹P ≤ 0.001, compared to referent group.

²P = 0.1, comparing females to males by age group.

³P ≤ 0.003, comparing females to males by age group.

DISCUSSION

This case–control study of 629 ΔF508 homozygotes collected for the GMS shows a strong association between earlier age of acquisition of Pa and severity of CF lung disease later in life, even when adjusting for potential confounders. Subjects first culture-positive for Pa before 5 years of age had 6.5 times the odds of severe lung disease later in life compared to those who were first positive at 10 years or older (or who were never Pa positive); this association was even more pronounced for persistent Pa, where the odds of severe disease were 11.2 times higher in subjects with persistent Pa at 0–4 years of age compared to the referent group. There appeared to be a stronger association between age of Pa infection and likelihood of severe lung disease in females compared to males.

Though several previous studies have linked respiratory infection with Pa to poor pulmonary outcomes in CF, this study is one of the first to focus on specific age of infection, particularly in early childhood, in a large study population with an extended time span of respiratory culture data. We have minimized the impact of CFTR genetics by restricting subjects to ΔF508 homozygotes. We also minimized potential confounding by birth cohort (from differences in treatment practices) by limiting our analysis to severe subjects and those “young milds” enrolled in the GMS between 15 and 28 years of age.²⁰ The majority of our subjects were first infected with Pa prior to the common use of inhaled tobramycin in pediatric CF clinics and in eradication protocols,²⁴ further lessening the impact of changes in treatment practices on our results.

In order to address the variability of respiratory culture data, we tested two definitions of Pa infection—first Pa and persistent Pa. Because existing classification systems for Pa infection rely on multiple points of culture data in a single year,^25,26 we created a definition of persistent Pa infection to accommodate annualized data; specifically, age of persistent Pa was defined as the age of the first of three consecutive years with Pa-positive cultures in at least 2 of the 3 years. The observed association between age of Pa and severity of CF lung disease was present when either definition of Pa infection was used in the analysis, though a stronger association was seen when comparing age of persistent Pa and lung disease severity. Our multiple sensitivity analyses on characterization of age of Pa infection, including continuous coding, varying size and number of age blocks for categorical coding, and comparison of age of Pa infection by years with only OP culture data versus years with sputum culture data available, all yielded similar patterns of association, strengthening the likelihood that our results reveal a true association. Shorter time from first culture to first Pa-positive culture among subjects with severe versus mild lung disease (Table 1) indicates that the earlier age of Pa infection observed in the severe group is not simply due to earlier age of available culture data, but instead reflects a true predilection for earlier Pa acquisition. Sensitivity analyses of variable definitions for confounders did not have a significant impact on our results. The observed association implies that presence of Pa under age 10, and particularly before age 5, is strongly tied to poorer pulmonary outcomes in subjects with CF, but does not infer causality.

We saw an interaction between gender and age of Pa infection, with a stronger association between age of Pa infection and likelihood of severe lung disease in females compared to males (Table 3). This is consistent with previous reports of female gender as a risk factor for worse pulmonary outcomes in CF.^{5–7,12,14,23} These results suggest that females may deal more poorly with Pa infection, or that females with worse lung disease may be more prone to Pa infection than males with the same degree of disease.²⁷ The interaction between gender and age of Pa infection has yet to be fully determined, and bears further investigation in future studies.

Our findings are consistent with previous studies in subjects with CF which have shown Pa infection to be associated with increased endobronchial inflammation, poorer clinical status, more rapid progression of lung disease, and increased morbidity and mortality in young children with CF.^{2,3,11–13,18} Previous studies have also raised the question of whether Pa infection is a cause of decline in lung function or a marker of more severe disease.¹¹ Female gender has also been shown in multiple studies to be an independent risk factor for more severe CF lung disease, similar to the findings in our study.^{5,14,23,28–30} Newer studies have suggested that Pa may be acquired in the preschool age range in the majority of patients with CF^1,31; this may be due to changes in culture practices as the CF community moved toward adopting eradication protocols.

We recognize our study has certain limitations, the most important of which is that we cannot determine causality based on our study design. Another limitation is the extremes of phenotype model, which limits our ability to generalize our results to the CF population as a whole. Our reliance on annualized data, made necessary by the later (1998) adoption of quarterly data into the CFFPR, means we cannot determine the density of Pa-positive cultures within a particular year, nor can we distinguish chronic infection from consecutive “new” episodes of Pa infection; however, we employed two distinct definitions of Pa infection with similar results. Staggered entry and exit times for CFFPR data have the potential to create bias (potentially leading to earlier age of Pa infection in those subjects with data at an earlier age), though the shorter time to Pa positivity seen in subjects with severe disease suggests that the observed earlier age of Pa infection in this group is not simply due to CFFPR data availability at a younger age. Staggered entry data in the CFFPR also limited our ability to determine and control for nutritional status early in life. Our database lacked strong indicators of socioeconomic status such as parental income, though we did employ health insurance status at the time of CFFPR enrollment as a proxy.⁹ We were unable to adjust for methicillin-resistant S. aureus or analyze the impact of mucoid versus non-mucoid Pa because these data were not collected until much later in our CFFPR dataset. We did not control for race/ethnicity, as less than 3% (n = 15) of our subjects were identified as non-Caucasian. False positives that are unrelated to the severity of disease, false negatives, and errors in reporting or recording of Pa status are a possibility in this study. However, such non-differential measurement errors would serve to attenuate the observed association between age of Pa infection and severity of CF lung disease in relation to the true population association, again suggesting that our reported association is a true one. Because the majority of our subjects acquired Pa prior to the routine use of inhaled tobramycin, our findings may not directly apply to current patients with CF. Despite these limitations, we believe that the robust association seen within our population between age of Pa infection and severity of later CF lung disease describes a true relationship, particularly since the association holds through multiple sensitivity analyses.

In summary, earlier age of infection with Pa in our population was strongly associated with greater likelihood of severe lung disease later in life, most particularly in those subjects who acquired Pa before age 5; the observed association was stronger in females than in males. Pa infection may be a cause of more severe CF lung disease, but may also be a marker of some other processes determining lung function in children with CF.

ABBREVIATIONS

CF

Cystic fibrosis

CTFR

CF Transmembrane conductance regulator

Pa

Pseudomonas aeruginosa

S. aureus

Staphylococcus aureus

FEV₁

Forced expiratory volume in 1 sec

GMS

Gene Modifier Study

CFFPR

CF Foundation Patient Registry

TGF-β

Transforming growth factor-β

OR

Odds ratio

OP

Oropharyngeal