Epidemiologic Study of Cystic Fibrosis (ESCF): 25 Years of Observational Research

SUMMARY

The Epidemiologic Study of Cystic Fibrosis (ESCF) was a prospective observational study of over 32,000 people with cystic fibrosis (CF) from 250 clinical care sites in North America from 1994 to 2005. Begun as a pharmacovigilance study in connection with the approval of dornase alfa in 1993, ESCF was open to all people with CF treated at any participating site in the United States or Canada. In addition to obtaining safety and effectiveness data on dornase alfa, ESCF collected encounter-based data to characterize the natural history and management of CF with a special focus on lung disease. During the study, 32,178 patients reported at least one encounter, contributing 869,136 encounters, 622,592 pulmonary function tests, 432,896 cultures, and 118,563 pulmonary exacerbations treated with intravenous antibiotics. Although ESCF data collection concluded in 2005, through a collaboration with the U.S. Cystic Fibrosis Foundation Patient Registry, additional follow-up data through 2017 was available for two-thirds of patients. This allowed for updating of CF genotype and survival information. Fifty-six peer-reviewed publications (cited over 3600 times) resulted from this study. In this manuscript we summarize the published ESCF manuscripts in thematic groups with key study findings and brief comments, and speculate on how ESCF findings will inform future data registries and patient care practices.

INTRODUCTION

The Epidemiologic Study of Cystic Fibrosis (ESCF) was an observational study begun in 1993 and ultimately spanning more than 10 years of data collection. There have been 56 peer-reviewed publications [1-56] that have been cited over 3600 times (mean 65, median 43). ESCF was initiated in response to an FDA request for post marketing pharmacovigilance on the newly approved therapy dornase alfa (Pulmozyme®) and was sponsored by Genentech, Inc. The study was designed to obtain safety and effectiveness data on dornase alfa and to use encounter-based data collection to characterize the natural history and management of CF with a special focus on lung disease [1].

Although begun in connection with the approval of dornase alfa, ESCF was open to all patients with CF in the United States or Canada; by 1996, 20,000 patients were enrolled. A total of 250 sites and 28,909 patients had 675,117 encounters recorded in the first phase of ESCF, which used paper case report forms (CRFs). In 2003, data collection was changed from paper-based to electronic and sites and patients were re-enrolled. There were 184 sites and 17,963 patients who contributed 194,019 encounters using the electronic CRF. Data collection concluded 12/21/2005. Throughout the study a total of 32,178 patients reported at least one encounter, contributing 869,136 encounters, 622,592 pulmonary function tests, 432,896 cultures, and 118,563 pulmonary exacerbations treated with intravenous (IV) antibiotics (IV exacerbations).

Although ESCF data collection concluded in 2005, through a collaboration with the U.S. Cystic Fibrosis Foundation Patient Registry (CFFPR) additional follow-up data has been available for the two-thirds of patients who provided CFFPR patient identification numbers [57]. This allowed for updating of CF genotype and survival information.

In this manuscript we summarize the published ESCF manuscripts in thematic groups with key study findings and brief comments. We then speculate on how ESCF findings will inform future data registries and patient care practices.

PATIENT STATUS, PRACTICE PATTERNS, AND CHANGE OVER TIME

Over the 12 years of ESCF, there were striking changes in therapeutic practice patterns (including several new therapies) and improved patient outcomes that were captured in the encounter-based data collection.

Practice patterns [2,3].

The long-term nature of ESCF allowed for both cross-sectional characterization and longitudinal evaluation of patient status, practice patterns, and their relationships. Adherence to the 1990 CF practice guidelines [58] varied: 57.5% had ≥4 visits (but only 27.4% had ≥4 routine visits), 75.8% had ≥2 spirometries, 79.3% had ≥1 respiratory culture, and 68.3% had ≥1 radiograph. Use of therapies showed indication bias (sicker patients were more likely to be treated).

Center-based analysis of practice patterns [7].

Variation in care patterns across CF centers provides an opportunity to address indication bias by comparing outcomes at different sites. Practice patterns in 1995-1996 were compared between top and bottom quartile sites based on median lung function as percent predicted forced expiratory volume in 1 second (ppFEV₁) by age group (6-12, 13-17, and 18+ years). Patients at top quartile sites had more frequent monitoring of clinical status, spirometry, and respiratory cultures and generally received more interventions, especially IV antibiotic treatments.

Practice patterns for young children [12].

A similar approach was used to compare care patterns for young children (age 0 to 3 years) based on each site’s ppFEV₁ for 6- to 12-year-olds. Top quartile sites had more CF infants diagnosed by family history or newborn screening and correspondingly fewer infants diagnosed because of symptoms. Infants at top quartile sites had more visits; more respiratory tract cultures; and more frequent use of IV antibiotics, oral corticosteroids, mast cell stabilizers, and mucolytics; but they received less chest physiotherapy, inhaled bronchodilators, oral nutritional supplements, and pancreatic enzymes.

Trends in inhaled antibiotic use [15].

An increase in chronic inhaled antibiotic use from 1996 through 2005 was accompanied by a decrease in acute use to treat exacerbations.

Trends in pulmonary function and respiratory signs and symptoms [16].

From 1995-2005, pulmonary function improved and the percentages of patients reporting cough or sputum production or having crackles or wheeze decreased. Changes in pulmonary function were not consistently mirrored by changes in symptoms.

Trends in use of routine therapies [23].

From 1995 to 2005, use of airway clearance, inhaled bronchodilators, dornase alfa, inhaled corticosteroids, inhaled antibiotics, oral nutritional supplements, and insulin/oral hypoglycemic agents increased substantially, while use of mast cell stabilizers and oral bronchodilators decreased. Three therapies not tracked in 1995, oral macrolide antibiotics, leukotriene inhibitors/antagonists, and inhaled hypertonic saline, had notable use by 2005.

Choice of dornase alfa regimen [52].

Dornase alfa was approved with two regimens: 2.5 mg inhaled once daily (QD) or twice daily (BID), with little guidance as to when to use each regimen. Patients beginning BID regimens 1994-2005 had worse lung function and more frequent IV-treated exacerbations than those beginning QD regimens. At the time of regimen switch, patients switching from QD to BID had more exacerbations and more signs and symptoms, whereas patients switching from BID to QD did not. Switching from QD to BID was associated with increased exacerbations; deterioration in lung function did not appear to be a driver for this switch. Patients switching from BID to QD were clinically stable, on average, suggesting that treatment burden and cost may have been drivers of the decision to switch regimens.

Socioeconomic status (SES) and treatment [19,27].

Starting in 2003, median household income by zip code, maternal educational attainment, and state insurance coverage were collected. There were no differences in chronic therapy use related to SES. Similarly, SES had little effect on treatment for exacerbation with any antibiotic. However, IV antibiotics were prescribed more frequently for patients with lower SES.

Comment:

Through quality improvement programs and studies that have shown the effectiveness of a range of therapies, there has been an increase in the use of therapies and a concomitant improvement in health outcomes. The disparities in health outcomes associated with low SES need further study.

RISK FACTORS IN EARLY CHILDHOOD

Early diagnosis and management of children with CF can have a profound effect on their health later in life. Beyond newborn screening, recognition of risk factors that manifest in early childhood may improve outcomes.

Risk factors for poor lung function at age 6 years [8].

Growth/nutritional indexes (weight for age (WFA), height for age (HFA), and percent ideal body weight) and pulmonary signs and symptoms (cough, sputum, clubbing, crackles), and presence of Pseudomonas aeruginosa (Pa) at age 3 were strongly associated with ppFEV₁ at age 6 years (Table 1). Analysis of forced vital capacity (FVC) and FEV₁/FVC ratios indicated that deficits in lung function at age 6 years were largely due to smaller lungs and not obstructive in nature.

Table 1:

Signs and symptoms of lung disease at age 3 years and pp FEV₁ at age 6 years (mean ± SD)

Characteristic	ppFEV1 With	ppFEV1 Without	P value
Cough	95 ± 21	100 ± 19	< 0.001
Sputum	93 ± 22	98 ± 20	0.001
Clubbing	94 ± 22	97 ± 20	0.035
Crackles	88 ± 22	97 ± 20	< 0.001
Pa	92 ± 21	100 ± 19	< 0.001

Pa, Pseudomonas aeruginosa. Adapted from [8] Table 4 p. 628

Risk factors for persistence of respiratory signs and symptoms [31].

Persistence of signs and symptoms (cough, sputum, clubbing, crackles, and wheeze) was studied over time in patients <4 years of age at enrollment. Earlier onset was associated with pancreatic enzyme use, presence of Pa, and prior diagnosis of asthma. Signs and symptoms of lung disease begin early and especially for cough and crackles a majority of those eventually affected have symptoms by age 2 years (Figure 1).

Figure 1.

FIGURE 1Percentage of patients reaching onset of persistent sign/symptom by age. Adapted with permission from Pediatric Pulmonology

Wheeze in early childhood [42].

Any history of wheeze in the first 6 years of life (61% of patients) was associated with lower ppFEV₁ at age 6 to 8 years (Table 2).

Table 2:

Wheezing phenotype and ppFEV₁ at age 6y to <8y (adjusted mean ± SE)

Wheezing	ppFEV1 at age 6y to <8y
No Wheeze (no wheeze by age 6y)	104.0 ± 0.8
Transient wheeze (before age 3y but not after age 3y)	97.6 ± 0.9
Late Wheeze (after age 3y)	100.0 ± 1.6
Persistent Wheeze (before age 3y and after age 3y)	95.7 ± 1.0

Adapted from [42] Table 2 p. 748

BMI and BMI percentile in children [48].

Percent ideal body weight (%IBW), BMI, and BMI percentile have been used as composite measures of nutritional status, although %IBW has been discredited. However, BMI and BMI percentile fail to identify a substantial proportion of children 2-18 years with CF who are stunted or have potentially poor nutritional status as measured by WFA and/or HFA percentile.

Risk factors for early mortality [50].

CFFPR mortality data in 2013 linked to ESCF data 1994-2005 showed that 5.7% of patients died before age 18 years. Mortality risk factors included sign and symptoms, demographics, Pa, and WFA (Figure 2). Addition of ppFEV₁ to the model had minimal effect on the other factors.

Figure 2.

Risk factors for mortality before age 18 years. Hazard ratios for two survival models: one that does not include ppFEV1category as a predictor (open circles) and one that includes ppFEV1category as a predictor (filled circles). Adapted with permission from Pediatric Pulmonology.50CFFPR, Cystic Fibrosis Foundation Patient Registry; ppFEV1, percent predicted forced expiratory volume in 1 s

Comments:

These findings suggest that aggressive intervention early in life aimed at growth/nutrition status, and pulmonary signs/symptoms may lead to better pulmonary function later in childhood. The BMI finding points out the importance of not relying on a single measure for assessing growth and nutritional status. The predictive power of clubbing and crackles at age 3 to predict mortality even after including FEV₁ as a predictor strongly suggest that they are important clinical findings that should inform care and be routinely collected in patient registries and observational studies of therapeutic effectiveness.

AIRWAY INFECTION AND LUNG DISEASE PROGRESSION

CF is characterized by complex chronic airway infections that contribute to inflammation and lung disease progression but demonstrating relationships between bacterial infections and outcomes is difficult, requiring at minimum comprehensive microbiologic information and extended observation periods.

Microbiology procedures [6].

In a site survey of pulmonary specimen collection and microbiology laboratory procedures, differences in culture frequency and conditions contribute to differences in species prevalence across sites.

Allergic bronchopulmonary aspergillosis (ABPA) [5].

Only 2% of patients had ABPA in 1994-1996 with substantial regional differences. In 2018, prevalence was 5.4% [57].

Methicillin resistant S. aureus (MRSA) [13].

Among the 7.5% of patients who had S. aureus but no other organisms detected in 2001, the 11% with MRSA had mean ppFEV₁ lower by about 9 points than those with methicillin-sensitive S. aureus (MSSA). Likelihood of hospitalization and treatment with oral, inhaled, and IV antibiotics were all significantly higher in MRSA patients.

MRSA and change in FEV₁ [18].

Compared to patients who remained MRSA-negative, the 12% of patients ≥6 years old with incident MRSA detection 2001-2003 had lower baseline ppFEV₁, received more antibiotic and other therapies, and had a higher rate of ppFEV₁ decline both before and after the incident culture. However, incident MRSA detection was not associated with a statistically significant change in rate of ppFEV₁ decline.

Multiple antibiotic-resistant Pa (MARPA) and change in FEV₁ [32].

Incident MARPA was defined as Pa (1) resistant to gentamicin and either tobramycin or amikacin, and (2) resistant to ≥1 antipseudomonal beta lactam after ≥2 years of Pa-positive but MARPA-negative respiratory cultures. The 25.5% MARPA-incident patients 1996-2003 had lower baseline ppFEV₁ and received more oral and inhaled antibiotic therapies. Decline in ppFEV₁ the two years before MARPA detection was similar to the decline two years after detection.

Treatment of newly detected Pa [36].

The EPIC (Early Pseudomonas Infection Control) study used ESCF data as historical controls. Protocol-based therapy for newly detected Pa resulted in lower Pa recurrence rates but comparable hospitalization rates compared to historical controls.

Comment:

These findings support the recommendation for a minimum of four respiratory cultures per year. MRSA and MARPA are more likely to be a marker of more severe disease and more intensive therapy, and less likely to be contributing independently to more rapid lung function decline.

DIAGNOSIS OF PULMONARY EXACERBATIONS

Although pulmonary exacerbations are important clinical events, there is no consensus as to what clinical presentations constitute exacerbation. Epidemiologic definitions of exacerbation often rely on antibiotic treatment and specifically IV treatment.

Clinical characteristics associated with exacerbations [9].

The four clinical characteristics most associated with exacerbation treatment (subsequently termed “Rabin criteria”) differed by age group (Table 3). Although Rabin criteria were likely to be present when exacerbations were diagnosed, the converse was not the case: presence of Rabin criteria was not a useful predictor of exacerbation diagnosis.

Table 3:

Clinical characteristics most associated with exacerbation treatment by age group

Age <6 years	Age 6-12 years	Age 13-17 years	Age 18+ years
New crackles	Decline in ppFEV1	Decline in ppFEV1	Decline in ppFEV1
Increased cough	Increased cough	Increased cough	New crackles
Decline in WFA	New crackles	New crackles	Hemoptysis
Increased sputum	New Pa	Hemoptysis	Increase cough

Adapted from [9] Table 3 p. 403

Acute FEV₁ drop [40]

Nearly one-third of children 6-12 years and one-fourth of adolescents 13-17 years with acute relative FEV₁ drop ≥10% were neither hospitalized nor treated with a new antibiotic course, even though their mean drop was near 20%. (Figure 3A) Those with the best lung function were one-sixth as likely to receive IV antibiotics as those with the worst lung function, suggesting a missed opportunity to preserve lung function and possibly contributing to the more rapid decline observed in children with high FEV₁.

Figure 3.

Treatment of PEx. Panel A. Percentage of children and adolescents treated with antibiotics after acute≥10% relative FEV1 drop stratified by ppFEV1 decile and age group. Adapted with permission from the Journal of Pediatrics.40Panel B. Area-proportional diagram of antibiotic treatments of clinician-identified exacerbations by route of administration. Adapted with permission from Pediatric Pulmonology.35Panel C. Mean antibiotic treatment events per patient‐year by IV route (upper panel) and by any route (lower panel). Adapted with permission from the Journal of cystic fibrosis.37Panel D. Documentation and antibiotic (ABX) treatment of PEx identified by Rabin Criteria or by≥15ppFEV1 drop. Odds ratios of treatment are shown for top quartile sites (Q1) versus other quartiles (Q2–Q4). Adapted with permission from Pediatric Pulmonology.44Panel E. Percentage of young children treated with antibiotics by number of Rabin Criteria present. Adapted with permission from Pediatric Pulmonology.34Panel F. Odds ratios for recovery of≥90% of baseline ppFEV1 after PEx for patients receiving in patient (filled circles) or outpatient (open circles) treatment as a function of ppFEV1 drop at diagnosis (upper panel) and Baseline ppFEV1(lower panel). Adapted with permission from the Annals of the American Thoracic Society.49IV, intravenous; PEx, pulmonary exacerbation; ppFEV1, percent predicted forced expiratory volume in 1 s

Acute FEV₁ drop events and clinician-diagnosed exacerbation events [55]

Among ≥10% relative FEV₁ drop events, 71% resulted in antibiotic treatment, of which two-thirds were IV. Of the clinician-diagnosed exacerbations, 32% had a drop ≥10%; 37% were treated intravenously. That is, a ≥10% relative FEV₁ drop was not always considered an exacerbation and a lack of drop did not rule out an exacerbation diagnosis.

Comment:

The lack of treatment of substantial decline events was surprising and led to further investigations (see below).

TREATMENT OF PULMONARY EXACERBATIONS

Although ESCF collected information on treated exacerbations from inception, beginning in 2003 clinicians identified exacerbations which might or might not have been treated.

Exacerbation treatment and change in lung function [35].

Clinician-identified exacerbations were treated with oral (73%), inhaled (24%), and/or IV (39%) antibiotics (Figure 3B), implying that limiting analyses to IV exacerbations excludes over 60% of exacerbations. Mean ppFEV₁ acutely improved 5.1 following IV and 2.0 following non-IV treatment, but regardless of route ppFEV₁ declined 3.8 from the previous year as of 180 days after the exacerbation. Lung function loss was similar for patients with only one exacerbation and those with none.

Trends in exacerbation treatment [37].

Overall treatment incidence rate and IV incidence fell between 1995 and 2005, while non-IV incidence increased in children ≤12 years and decreased in older patients. (Figure 3C) Treatment thresholds (Rabin scores [9]) lowered over time, except for IV treatment in older patients. The overall decrease in treatment implies that lower treatment thresholds were more than offset by reduced treatment incidence (presumably resulting from improved health).

Exacerbation treatment by site [44].

High quartile sites (based on median FEV₁) were more likely to treat exacerbations identified by Rabin criteria or by drop of ≥15 ppFEV₁. (Figure 3D)

Exacerbation treatment outcomes in young children [34].

The number of Rabin criteria present in children <6 years old during a six-month evaluation predicted a higher hospitalization rate over the following year as well as lower WFA and lower ppFEV₁ at age 7 years. Antibiotic treatment within 7 days before and 28 days after the encounter was associated with a lower proportion of children with crackles, cough, and Pa at follow-up (Figure 3E).

FEV₁ recovery after exacerbation treatment [49].

All exacerbation treatments (hospitalizations, home IV antibiotics, new inhaled antibiotics, new oral quinolones, or other oral antibiotics) were associated with an increased likelihood of recovery to at least 90% of baseline lung function compared with no treatment, and inpatient treatment was better than outpatient treatment (Figure 3F). These findings were similar irrespective of baseline lung function.

Inpatient versus outpatient exacerbation treatment [53].

Employing three alternative techniques to control for indication bias (including instrumental variables), multiple response criteria, and multiple measures of inpatient versus outpatient administration, inpatient IV antibiotic treatment was consistently associated with higher likelihood of FEV₁ recovery. Treatment response was not related to duration of IV therapy.

Evaluation of lung function recovery after exacerbation treatment [56].

Complete ppFEV₁ recovery of baseline at follow-up was dependent on baseline definition, timing of assessment, and the ppFEV₁ difference between baseline and treatment start. Lung function recovery as return to baseline is prone to artifact because of inherent FEV₁ variability, sampling methodologies, and underlying disease progression.

Comment:

Research is needed to better understand how clinicians identify exacerbations and when and how they choose to treat (inpatient IV, outpatient IV, inhaled, or oral antibiotics). Special attention is needed for exacerbations treated only with oral or inhaled antibiotics in the outpatient setting. Acute drops of ppFEV₁ are frequently not treated despite the evidence that treatment improves lung function irrespective of baseline level.

PREDICTING FUTURE LUNG FUNCTION

Chronic progressive lung disease is the most frequent cause of mortality in patients with CF.

Risk factors for lung function decline in children and adolescents [14] and in adults [33].

For children and adolescents, high baseline ppFEV₁ and persistent crackles were significant risk factors for three-year ppFEV₁ decline, and risk factors in some age groups female sex, presence of Pa, low WFA, sputum, wheezing, sinusitis, IV exacerbations, elevated liver function tests, and pancreatic insufficiency (Figure 4A). For adults 18-24 years, Burkholderia cepacia, pancreatic enzyme use, MARPA, and female sex predicted greater decline; low baseline ppFEV₁ and sinusitis predicted less decline. For the ≥25y group, only pancreatic enzyme use predicted greater decline; low baseline ppFEV₁ and sinusitis predicted less decline.

FIGURE 4.

Lung function decline by age and baseline ppFEV1.Panel A. Mean rates of ppFEV1 decline by age group. Overall rates (open circles) and rates stratified by baseline ppFEV1 category (filled circles) adjusted for other risk factors are shown. Adapted with permission from the Journal of Pediatrics and the Journal of CysticFibrosis.14,33Panel B. Average changes in ppFEV1 around age 18years by ppFEV1 category adjusted for other risk factors. Adapted with permission from Pediatric Pulmonology.30ppFEV1, percent predicted forced expiratory volume in 1 s

Pulmonary outcome prediction (POP) tools [22].

Simple integer-based pulmonary outcome prediction (POP) tools to estimate lung function at age 6 in patients aged 2–5 years (POP2–5) and lung function change over a 4-year period in patients aged 6–17 years (POP6–17) were developed using variables available at clinical visits. Both tools included WFA, clubbing, crackles, wheeze, number of exacerbations, and presence of Pa. POP2-5 added daily cough; POP6-17 added age, sex, and ppFEV₁.

Risk factors for lung function decline in young adulthood [30].

Rate of ppFEV₁ decline was greater in young adulthood (18.5 to 22.0 years) than in adolescence (14.0–17.5 years). Factors during adolescence that predicted substantial decline (≥5 ppFEV₁ per year) in young adulthood were: slower rate of FEV₁ decline, greater FEV₁ variability, faster BMI decline, male sex, chronic inhaled antibiotics, presence of Haemophilus influenzae, and absence of MARPA. Additional risk factors were lower than expected FEV₁ and BMI around age 18 (Figure 4B).

Year-to-year changes in FEV₁ [21].

The year-to-year change in ppFEV₁ for individuals (using the highest ppFEV₁ for each year) was compared to the change in the population median for the corresponding ages. Individual ppFEV₁ decreased 1–3 points per year, with maximal decreases in 14–15 year olds. Population changes agreed with individual changes up to age 15. After age 30, population change approximated zero while individual ppFEV₁ decreases were 1–2 points per year, indicating loss of FEV₁ is a persistent risk in 6–45 year old CF patients (Figure 5).

FIGURE 5.

Change in lung function by age. mean, median, and quartile changes in best recorded ppFEV1 from the prior year are shown as a function of age. Bars represent 95% CI. Adapted with permission from the Journal of Cystic Fibrosis. 95% CI, 95% confidence interval; ppFEV1, percent predicted forced expiratory volume in 1 s

Using decline in lung function to predict future decline [46].

For each year of age, the best ppFEV₁ and associated ppFVC and ppFEF_25–75, were used to calculate 2-year slopes. Slopes were most negative among those 13–17 years old, especially for FEF_25–75. Rate of lung function decline had a moderate correlation with later level of lung function, but contrary to expectations did not predict future rate of decline 3-8 years later either within or across spirometric variables.

Lung function variability [47].

Five different ppFEV₁ variability measures from a 2-year baseline period were predictive of the best ppFEV₁ in the subsequent 2 years both alone and after controlling for baseline demographic and clinical factors and the slope and level of ppFEV₁. The easy-to-calculate median deviation from the best ppFEV₁ in the baseline period had the highest contribution to explanatory power and exceeded the total contribution of all other factors excluding the rate of ppFEV₁ decline (Figure 6). Additional analyses revealed that number of exacerbations did not explain much of the variability [59,60].

FIGURE 6.

Prediction of best ppFEV1 in subsequent 2 years after2-year baseline. Contribution to percentage of variance explained(R2) in a regression model of ppFEV1 variability (as measured by median deviation from best ppFEV1; open bars), trendline slope (rate of ppFEV1 decline; gray bars), and all other factors (demographics, clinical factors, and trendline intercept; black bars). Adapted with permission from the Journal of Pediatrics. ppFEV1, percent predicted forced expiratory volume in 1 s

Comment:

Clinicians should not be reassured by high lung function, particularly in young children, because it is associated with more rapid decline in ppFEV₁. This is possibly due to failure to treat acute drops in lung function in patients with high ppFEV₁. Crackles predict future lung function decline in all ages and warrant attention in clinical care as well as research studies and are now captured in the CFFPR. Variability in ppFEV₁ is getting increased attention and may play an important role not only as a predictor but also as a potential surrogate variable for lung function decline.

DIABETES, PREGNANCY, AND LIVER DISEASE

Cystic fibrosis-related diabetes (CFRD) [10].

Patients with CFRD (defined as using insulin or an oral hypoglycemic agent) had more liver disease, more advanced lung disease, more exacerbations, and worse nutritional status. Previously reported CFRD risk factors of older age, female sex, and pancreatic insufficiency were confirmed.

Relationship between pregnancy, therapies and health outcomes [11].

Pregnant women (n=216) were compared with a group of never-pregnant women matched using propensity scores. Baseline ppFEV₁ was higher for the pregnant women but declines from baseline to follow-up were not different. Compared to the nonpregnant group, outpatient visits for pregnant patients were more frequent before and increased during pregnancy. Exacerbations and hospitalizations were similar between the groups at baseline but increased during pregnancy. Treatment for diabetes rose from 9.3% to 20.6% during pregnancy and then dropped to 14.4%. Although pregnant women experienced similar respiratory and health trends, they used a greater number of therapies and received more intense monitoring than nonpregnant women.

Pregnancy and motherhood [38].

A group of 119 women reporting a pregnancy and followed a median of 6.0 years from the pregnancy were propensity-score matched with never-pregnant women. No group differences were found in change from baseline for FEV₁ or BMI, in respiratory signs and symptoms, or in prescribed chronic therapies. Women who had been pregnant had lower health-related quality of life (HRQOL) scores, more exacerbations and more illness-related visits but showed no increase in chronic therapies.

Liver disease in Hispanics [43].

Abnormal liver findings were present in 20.8% of Hispanic patients compared with 16.0% of non-Hispanic white patients. This higher prevalence of liver involvement persisted after adjusting for demographics and meconium ileus and was most pronounced in the first year of life. Follow up of infants with elevated liver enzymes found that 10% progressed to more severe liver disease.

Comment:

Research to understand what occurs in patients before diagnosis with CFRD. With improvements in survival, CFRD is likely to become a more important problem. Women with CF can successfully navigate pregnancy and motherhood. Disease outcomes and associated risk factors for Hispanics and other minorities with CF need further investigation.

HEALTH-RELATED QUALITY OF LIFE (HRQOL)

Patient-reported outcomes are increasingly used in clinical trials to assess the natural history of chronic diseases and the efficacy of new treatments. Beginning in 2003, there were 14,258 Cystic Fibrosis Questionnaire-Revised (CFQ-R) forms completed by 7,763 patients and/or their parents from 160 sites.

Socioeconomic status (SES) and HRQOL [20].

Low SES was associated with lower CFQ-R scores on a majority of questionnaire domains. Even after controlling for lung function and SES, African American and Hispanic patients reported worse emotional and social functioning.

Change in CFQ-R [24].

Increased respiratory signs/symptoms were associated with worse CFQ-R Respiratory Symptom scores and declining weight was associated with worse scores on nutritional health domains. A Treatment Complexity Score (TCS) based on 37 chronic therapies was developed. For parents, increases in TCS were associated with worsened CFQ-R Treatment Burden scores.

Treatment complexity [39].

Treatment complexity as measured by the TCS were highest in adults and increased over the three years 2003-2005 for all ages. Sites were stratified by quartiles based on ppFEV₁, BMI, or CFQ-R Treatment Burden score. TCS did not differ by ppFEV₁ quartile; was higher in the highest BMI quartile; and was lower in the CFQ-R Treatment Burden quartile with lowest burden.

CFQ-R psychometrics [29].

The CFQ-R showed strong internal consistency reliability, few floor or ceiling effects, good discrimination between clinically defined groups, strong parent-child agreement on respiratory symptoms, and convergence between CFQ-R scales and health outcomes. Females reported better HRQOL than males on scales related to body image and weight and worse HRQOL on other scales.

Comment:

Health-related quality of life measures such as those from the CFQ-R are increasingly used in clinical care as well as in research.

DESIGN OF PROSPECTIVE STUDIES

When designing a clinical trial, study duration, outcome measures, and inclusion/exclusion criteria need to be determined. Observational studies can inform these decisions.

FEV₁ decline as endpoint [25].

Increased study duration and exclusion of lower risk patients both substantially reduced sample size requirements but increasing visit frequency had only modest impact. Studies of 1.5 years in duration appeared feasible, but shorter studies are unlikely to be adequately powered, irrespective of visit frequency.

Exacerbation rates as endpoint [28].

Rates of clinician-identified exacerbations (those treated with any antibiotic or only those treated with IV antibiotics) were estimated for durations from 3 to 12 months for subpopulations stratified by age, FEV₁, sex, WFA percentile, respiratory signs and symptoms, and history of exacerbations and bacterial culture. Prospective studies including patients with lower ppFEV₁, older age, female sex, recent history of exacerbations, and presence of Pa required lower sample sizes.

Pulmonary function outcomes [45].

Six potential outcome variables (best FVC, FEV₁, and FEF_25–75 in 2002 and rate of decline for each variable from 2000 to 2002) were used to rank sites and associations with practice patterns and follow-up pulmonary function assessed. Top quartile sites had more frequent monitoring, treatment of exacerbations, and use of chronic therapies and oral corticosteroids, but follow-up rates of FEV₁ decline were not predicted by site ranking. Annual FEV₁ was at least as good as any other pulmonary function measure. The current site ranking only moderately predicted future ranking.

Choice of FEV₁ reference equation [51].

The dornase alfa rate of decline study [26] was reanalyzed to compare the effect of using different reference equations: Knudson [61], Wang [62] & Hankinson [63], Stanojevic [64], and GLI [65]. The study conclusions about acute change and change in rate of decline were minimally affected. Based on this analysis, the recent GLI equations were recommended for future studies, but prior results based on older equations should be accepted as reliable.

Comment:

Observational studies such as ESCF can provide guidance for clinical trial design. While choice of ppFEV₁ reference equations matters for individual care decisions, intervention evaluations are not greatly affected.

COMPARATIVE EFFECTIVENESS

In contrast to randomized clinical trials, which can determine the efficacy of an intervention, observational studies such as ESCF are ideal for determining the real-world effectiveness of interventions. Although limited by various types of bias, especially indication bias, observational studies have the strengths of large sample sizes, broad inclusion criteria, and extended observation periods. These strengths allow the evaluation of long-term outcomes such as long-term lung function decline or survival that are not practical in clinical trials.

Dornase alfa [4,26].

The randomized controlled trial for dornase alfa demonstrated a 5.8 ppFEV₁ sustained improvement in ppFEV₁ [66]. In a 1997 analysis, dornase alfa patients in ESCF age ≥6 years with baseline ppFEV₁ ≥40 had lower pulmonary function, more bacterial infections, and more exacerbations at baseline than untreated comparators [4]. Treated patients had a 3.9 ppFEV₁ increase compared with a 1.6 ppFEV₁ decrease for untreated patients that was sustained for a year.

In a later study using data through 2005, a change-point model was used to relate initiation of dornase alfa to acute ppFEV₁ change and subsequent change in rate of ppFEV₁ decline, an endpoint not previously explored [26]. Change-point analysis of lung function looks before and after an event (such as the initiation of a new therapy) to see if there is an acute change (change in intercept) or a trend change (change in slope). This analysis introduced innovative methods, described in some detail in the online supplement [26]. Age-specific ppFEV₁ deciles accounted for the fact that patients with relatively high lung function have a shallower decline before and a steeper decline after an arbitrary point in time and the opposite for those with relatively low lung function. In addition, patients could contribute multiple four-year time spans to the comparator group (as was done for the inhaled corticosteroid analysis) but in addition dornase alfa patients could be included in the comparator group up to the time they received dornase alfa. Sensitivity analyses showed that this approach had greater power than using just one four-year time interval chosen at random and was additionally less biased than using the first or last possible interval.

Patients treated with dornase alfa had lower ppFEV₁ at treatment initiation and a more rapid decline prior to treatment relative to comparators. There was an acute improvement and the rate of ppFEV₁ decline was reduced by 32.1% among children (95% CI 15.1% to 46.8%) and by 28.8% among adults (−1.7% to 51.9; Figure 7A). For the comparators, there was no material change in intercept and the rate of decline worsened in children and did not change in adults.

FIGURE 7.

Change-point models of ppFEV1 before and after initiation of chronic treatments. Panel A. Change in ppFEV1 among children (upper lines) and adults (lower lines) receiving chronic dornase alfa treatment (DA; solid lines) versus an untreated comparator group (dashed lines). Mean ppFEV1 decline rates (slopes) are shown above lines; differences in intercept between lines are shown with arrows. Adapted with permission from Pediatric Pulmonology.26Panel B. Change in ppFEV1 among children receiving chronic ICS (solid lines) versus an untreated comparator group (dashed lines). Derived from previously published data.17Panel C. Change in ppFEV1 among patients ages 8 to38 years receiving chronic TIS (solid lines) versus an untreated comparator group (dashed lines). Mean ppFEV1 decline rates (slopes) are shown next to lines; differences in intercept between lines are shown with arrows. Derived from previously published data.41ICS, inhaled corticosteroids; ppFEV1, percent predicted forced expiratory volume in 1 s; TIS, tobramycin inhalation solution

Chronic inhaled corticosteroid (ICS) [17].

The change-point model for ICS allowed for a change in rate of ppFEV₁ decline at initiation of therapy but no acute change. For ICS patients, rate of ppFEV₁ decline was reduced by 71% (95% CI 40% to 98%), whereas the rate of ppFEV₁ decline in the comparator population increased 43% (Figure 7B). ICS use was associated with decreased HFA and increased insulin/oral hypoglycemic use.

Tobramycin inhalation solution (TIS) [41].

Using the same methods as the dornase alfa change-point investigation, acute ppFEV₁ change and subsequent change in rate of ppFEV₁ decline associated with initiation of chronic TIS were evaluated; comparators used TIS<10% of the time. TIS patients improved at initiation versus no change for comparators, whereas the rate of ppFEV₁ decline worsened in both groups by similar amounts (Figure 7C).

High-dose ibuprofen (HDI) [54].

Although treatment-associated reduction in rate of lung function loss should theoretically improve survival, no such relationship had been shown for a chronic CF therapy. In part, this is because the ages of most rapid lung function decline [14,21] -- early adolescence – occur more than a decade before the median age of death. HDI had previously been shown to reduce rate of ppFEV₁ decline without a corresponding acute improvement, but no survival benefit had been published. Children age 6-17 years receiving HDI for 2 years were matched 1:5 with comparator children not treated with HDI. HDI patients had a 37.5% reduction in rate of ppFEV₁ decline during 2 years of treatment and an 18% reduction in mortality (in ESCF and subsequently up to 16 years in the U.S. CFFPR) (Figure 8).

FIGURE 8.

Survival after initiation of at least 2 years of chronic HDI in children (6–17 years). The 2-year HDI exposure period is shown as a gray box. Survival among patients receiving at least 2years of chronic HDI (N= 775) is represented as a black line and survival among propensity-matched comparator patients (N= 3665 )is shown in gray. Adapted with permission from the Annals of the American Thoracic Society.54HDI, high-dose ibuprofen

Comment:

Despite disease complexities and polypharmacy, observational studies can be used to evaluate comparative effectiveness of chronic therapies. Although an acute increase in ppFEV₁ is desirable, it is more important to reduce the long-term rate of lung function decline which is not necessarily associated with an acute increase. The HDI results are consistent with the hypothesis that treatment-associated reduction of lung function decline in children with cystic fibrosis leads to improved survival. Although short term improvements in ppFEV₁ can be assessed in clinical trials, only with long-term observational studies is it possible to evaluate lung function decline and survival.

IMPLICATIONS FOR FUTURE DATA REGISTRIES AND PATIENT CARE PRACTICES

ESCF was the first encounter-based patient registry for CF and has the informed the design of similar registries. In addition to the scientific publications, ESCF also provided site-specific reports with regional and national data for comparison, allowing for quality improvement initiatives. Although practice patterns will change dramatically with the introduction of highly effective modulator therapies (HEMT), patient registries such as the CFFPR remain critical to evaluate long-term outcomes including survival in relation to practice patterns. Furthermore, unlike the development of new therapeutic interventions, existing practices are unlikely to be evaluated formally outside of observational studies. Requirements of pharmacovigilance are no substitute for long-term registries that provide a comprehensive look at patients and their care over their lifetime.

As patients become healthier, we may become complacent or reliant on untested assumptions or analyses that are no longer relevant. As the natural history of CF changes, we need to update our understanding of risk factors for disease progression and poor outcomes so that we can continue to anticipate disease progression and deliver care accordingly. As survival improves, we will likely need to evaluate lung function much later in life than has been feasible in the CF population to date. This would allow the evaluation of risk factors for the development of progressive lung disease analogous to COPD and non-CF bronchiectasis.

ESCF and CFFPR has been an important source of information on practice patterns and also serve to provide historical control populations for new therapies. Until therapies are developed for the approximately 10% of patients not currently appropriate for HEMT, we will need registries to supply similar information for those patients.

Care summaries provided in connection with point of service data collection may allow clinicians to take action based on information such as ppFEV1 variability or acute drop. An electronic version of the POP tool or similar risk assessment tool could be used to guide care by highlighting patients at high risk of lung function decline.

Although attempts to develop more sensitive measures of lung disease (MBW, HRCT, MRI) are laudable, but greater attention to simple, available measures (chest exam, spirometry, height, and weight) is likely to improve outcomes.

• Overall, 32,178 patients contributed 869,136 encounters.

• Survival information was updated from the U.S. CFFPR for two-thirds of patients.

• Sites with better lung function outcomes monitor and treat more aggressively [7,12]

• Over the course of ESCF, there was an improvement in health outcomes and increase in the use of therapies [15,16,23]

• Poorer health outcomes previously associated with low SES cannot be accounted for by decreased access to care or decreased diagnosis/treatment [19,27]

• Respiratory signs and symptoms and nutritional indexes in young children predict lung function at age 6 years [8]

• Persistent signs and symptoms of lung disease begin early, often by age 2 years [31]

• Pediatric wheeze is an important predictor of reduced ppFEV₁ at age 6 to 8 years [42]

• BMI and BMI percentile fail to identify children who are stunted or have potentially poor nutritional status as measured by WFA and/or HFA percentile [48]

• Signs and symptoms at an early age remain independent risk factors for mortality even after controlling for lung function [50]

• Differences in respiratory culture frequency and conditions contribute to differences in apparent prevalence [6]

• Likelihood of treatment was higher in MRSA patients [13] and MARPA patients [32], but neither incident MRSA nor incident MARPA was associated with a change in rate of ppFEV₁ decline [18,32]

• Clinical characteristics most associated with exacerbation treatment were identified [9]

• Clinicians fail to respond to acute ≥10% relative FEV₁ decline events, particularly in patients with higher lung function [40,55]

• Limiting analyses to IV exacerbations excludes a majority of exacerbations [35]

• Reduced exacerbation treatment would have been more pronounced if not for lower treatment thresholds [37]

• Sites with highest lung function were more likely to treat exacerbations identified by Rabin criteria [44]

• Antibiotic treatment of children presenting with at least one Rabin criterion was associated with a lower proportion with crackles, cough, or Pa at follow-up [34]

• Patients with inpatient exacerbation treatment had better likelihood of lung function recovery than those with outpatient treatment which in turn is better than no treatment [49,53]

• Lung function recovery from exacerbations as return to baseline is prone to artifact [56]

• High ppFEV₁ and persistent crackles are risk factors for lung function decline across all ages [14,22,33]

• Variability in ppFEV₁ is a major risk factor for lung function decline [30,47]

• Decline in lung function is greatest in adolescents and although population decline is minimal after age 30, individual decline continues [21]

• Past lung function decline does not predict future rate of decline ≥3 years later [46]

• CFRD is a common complication associated with more severe disease [10]

• Pregnancy and motherhood do not accelerate disease progression but are associated with more intense monitoring, more therapies, more illness-related visits and exacerbations, and a decrease in quality of life [11,38]

• Higher prevalence of liver disease in Hispanics is not explained by demographics and prevalence of meconium ileus [43]

• The CFQ-R health-related quality of life assessment tool and a newly-developed Treatment Complexity Score (TCS) were validated in multiple studies [20,24,39,29]

• ESCF data has been used to quantify the effects of duration, measurement frequency, and inclusion/exclusion criteria on the statistical power of studies [25,28]

• Annual FEV₁ is at least as good a measure for distinguishing among site practice patterns as other spirometric parameters including slopes [45]

• The GLI reference equations for ppFEV₁ are recommended, but results based on older equations should be accepted [51]

• Initiation of dornase alfa is associated with an acute ppFEV₁ increase and then a slower rate of decline [4,26]

• Initiation of ICS is associated with a slower rate of ppFEV₁ decline without an acute increase [17]

• Initiation of TIS is associated with an acute ppFEV₁ without a change in rate of decline [41]

• Initiation of HDI is associated with a slower rate of ppFEV₁ decline and improved survival [54]