Published ahead of print on August 13, 2003, doi:10.1164/rccm.200303-434OC
© 2003 American Thoracic Society Bronchopulmonary Disease in Children with Cystic Fibrosis after Early or Delayed DiagnosisDepartments of Pediatrics, Biostatistics/Medical Informatics, Radiology, and Nutritional Sciences, University of Wisconsin, Madison; and Department of Pediatrics, Medical College of Wisconsin, Milwaukee, Wisconsin Correspondence and requests for reprints should be addressed to Philip M. Farrell, M.D., Ph.D., University of Wisconsin Medical School, Room 1217 MSC, 1300 University Avenue, Madison, WI 53706-1532. E-mail: pmfarrel{at}facstaff.wisc.edu
Although early diagnosis of cystic fibrosis (CF) can lead to nutritional benefits, there has been uncertainty about pulmonary outcomes. Using a randomized controlled trial with unique unblinding/surveillance, we evaluated patients with CF who received similar treatment after being assigned to an early diagnosis (screened) group or to a standard diagnosis (control) group. When the youngest patient was 7 years of age, we compared outcomes using pulmonary function data and quantitative chest radiology. In the screened group (56 patients), diagnosis was made at a younger age of 12.4 weeks, compared with the diagnosis in control group (47 control patients) at the age of 95.8 weeks, but included a significantly greater proportion of patients with  F508 genotypes and pancreatic insufficiency. The first chest radiograph showed significantly fewer abnormalities in the screened group; but, over time, the two groups converged, and after 10 years of age the screened patients showed worse chest X-ray scores associated with earlier acquisition of Pseudomonas aeruginosa. No differences were detected in any measure of pulmonary dysfunction, which was generally mild in each group. Although CF neonatal screening provides a potential opportunity for better pulmonary outcomes, it appears that respiratory infections and pancreatic status are the dominant factors in pulmonary prognosis.
Key Words: cystic fibrosis • newborn • screening • pulmonary disease • Pseudomonas aeruginosa Although cystic fibrosis (CF) generally shows characteristic signs or symptoms during infancy (1–3), diagnostic delays are common so that the median and mean ages of recognition in the United States are 0.5 and 3 years (3, 4), respectively, with the former skewed by patients with meconium ileus or a positive family history and the latter by adult patients with pancreatic sufficiency. When diagnosis was made by traditional methods that rely on suspicion and a sweat test, nearly half the patients showed severe malnutrition (3) and many were suffering from chronic respiratory disease (1, 2, 4). Early diagnosis through neonatal screening became possible after 1979 when Crossley and coworkers (5) determined that measurement of immunoreactive trypsinogen on dried blood specimens from newborns could be used to identify infants with CF. A variety of studies conducted during the past two decades have demonstrated that nutritional benefits are achievable with early diagnosis (6–11). On the other hand, pulmonary outcomes after screening have been less certain and are a source of controversy (12). Observational studies (8–11, 13), however, suggest that potentially milder bronchopulmonary disease is present in screened patients compared with those diagnosed by traditional methods. In addition, research associated with the Colorado screening program has demonstrated conclusively that CF respiratory disease with recurrent infection/inflammation often develops early during infancy (1, 2), despite histologically normal lungs at birth (14). The Wisconsin Cystic Fibrosis Neonatal Screening Project was established in 1984 as a randomized, controlled trial to assess the benefits, risks, and costs of CF neonatal screening (15) and the epidemiology of childhood CF. Our data demonstrated nutritional benefits by Year 10 of the study (6), particularly in patients without meconium ileus (7), and randomization was therefore stopped, whereas longitudinal evaluations continued. Pulmonary outcomes, however, have proved more difficult to evaluate because CF lung disease has a variable onset in children, with multiple risk factors and problematic evaluation methods in young children. The diagnostic tools for quantitating pediatric lung disease were especially limiting when we began this randomized trial in 1985, although better methods such as high-resolution computed tomography have become available since then. In addition to these challenges, we experienced unexpected confounders when some postrandomization group imbalances emerged (7) and a subgroup of screened patients acquired Pseudomonas aeruginosa prematurely during 1985 to 1990 at an old clinic of one of Wisconsin's two CF Centers collaborating in this investigation (16). When all the children in this study reached 7 years of age and could perform adequate pulmonary function tests (PFTs), we analyzed the quantitative pulmonary outcome measures such as spirometry data and chest X-ray scores and now report important findings about the course of bronchopulmonary disease in CF as related to the age at diagnosis and other variables.
Design of a Randomized Trial of CF Neonatal Screening The design of the Wisconsin Cystic Fibrosis Neonatal Screening Project is described in detail elsewhere (15). In summary, we performed an assessment of the potential benefits of early diagnosis through neonatal screening in two concurrent groups of randomly identified children with CF who were managed prospectively with the same protocol. The two groups of patients with CF were selected through a 1985-to-1994 randomization process, proven statistically valid (7), in which newborns' blood specimens were assigned either to an early diagnosis (screened) group or to a standard diagnosis (control) group (15). To avoid selection bias, a unique unblinding/surveillance system (12) was used to identify all the control patients by approximately 4 years of age (7). Other methods, described in detail elsewhere (15), were also used to ensure an unbiased assessment of outcomes. A sweat chloride level of 60 mEq/L or greater at one of Wisconsin's two CF Centers was required to establish the diagnosis. This investigation was approved by the institutional review boards at the University of Wisconsin and at the Childrens' Hospital of Wisconsin. Patients included in this assessment were those in whom diagnosis was made through newborn screening and those not having meconium ileus, those randomized to the control group, with of signs or symptoms other than meconium ileus, and those with a positive family history.
Pulmonary Outcome Assessments
Statistical Analyses
Age at Diagnosis and Clinical Characteristics Table 1 lists the age at diagnosis and characteristics of patients identified and enrolled from the onset (April 1985) of randomized screening to the completion (April 1998) of the unblinding and surveillance efforts. The unblinding method enabled us to identify and include in this analysis 11 control subjects (23%) in whom diagnosis was yet to be made on the basis of positive family history, signs, or symptoms of CF (7). The mean age at diagnosis based on a positive sweat test was 12.4 weeks in the screened group and 95.8 weeks in the control group (p < 0.001). Included in the screened group were four patients with false-negative screening tests in whom diagnosis was made by a positive family history or by symptoms at 7, 8, 21, and 281 weeks and another screened patient who had an initial negative sweat test following a high immunoreactive trypsinogen level but in whom diagnosis was made at 21 months because of symptoms and a subsequent positive sweat test. Similar patient distributions were found when comparing center and sex; however, apparently by chance (7), there were significant differences in genotype and pancreatic status between groups, despite randomizing 650,341 newborns (6). Specifically, as identified in previous analyses (7), we found a greater proportion of F508-containing genotypes in the screened group and significantly more subjects with pancreatic sufficiency or probable pancreatic sufficiency (6) in the control group (Table 1).
Pulmonary Function Data PFT results are shown in Figure 1 for 7- to 16-year-old patients, as related to age using all Pediatric Alternate Spirometry System "acceptable" (19) data. These results were analyzed using either % predicted values or absolute values, and the same conclusions were reached. The % predicted FEV1/FVC results at 7 years of age for the screened and control groups were not statistically different (94 ± 1.8% vs. 95 ± 1.3%; p = 0.54). In addition, there were no differences over time between the groups in % predicted FEV1/FVC (p = 0.32) or in % predicted FEV1 (p = 0.18), nor were the absolute values different (e.g., FEV1, p = 0.58). It should be pointed out, however, that the opportunity to detect differences in pulmonary function is limited by the fact that 88 and 75% of the observations in 7-year-old screened and control patients, respectively, were within the normal limits (21) at 89% predicted or greater (p = 0.30). Figure 1 shows that the PFT data demonstrated relatively mild pulmonary dysfunction in both groups throughout the evaluation. Even the most sensitive measures such as mean forced expiratory flow during the middle half of the FVC (FEF25–75) and residual volume (RV)/total lung capacity (19, 24) showed very little deterioration with age. The comparisons of FEF25–75 (p = 0.37) and RV/total lung capacity (p = 0.25) also revealed no group differences. In addition, using all acceptable data obtained after 4 years of age, there were no statistically significant differences in any of the PFT measures analyzed.
Chest Imaging Results at Diagnosis Quantitative chest radiography was more useful than PFTs in assessing bronchopulmonary disease. Table 2 presents the results of the initial chest radiographs available in 49 screened patients compared with 40 control patients. At an average age of 14.3 weeks compared with 108 weeks (p < 0.001), the screened group showed better WCXR and BCXR mean scores. After adjusting these data for age (or for age, genotype, and pancreatic status), the differences remain significant (Table 2). Using discriminators of potentially irreversible lung disease described elsewhere (19) (i.e., WCXR  5 and BCXR  21), we found that approximately half the number of the control patients had reached this stage of radiographic lesions compared with about one-fourth the number of the screened patients (Table 2). The group differences were especially prominent in the subscores that evaluate structural airways abnormalities, reflecting apparent lower respiratory tract infection, i.e., peribronchial thickening and bronchiectasis (18, 19). Thus, significant differences were found in the summation subscores of these two components (p = 0.022), whereas measures of hyperinflation were not different (p = 0.85). In other words, after a delay in diagnosis averaging just over 2 years, half of the control group had developed structural changes by chest imaging that reflect permanent bronchopulmonary abnormalities, and this is twice the proportion in the screened group.
Longitudinal Assessment of Chest Radiographic Abnormalities Other differences emerged in two groups as they aged and were assessed with quantitative chest radiology. Examining only the data recorded after 5 years of age to avoid selection bias and a GEE Model (I) that included patient group, CF Center, sex, and age, our analysis revealed worse scores over time in the screened group (p = 0.017 for WCXR, and p = 0.041 for BCXR); however, if genotype, pancreatic status, and indicators of P. aeruginosa infection were also included (Model II), there were no significant differences (p = 0.10 for WCXR, and p = 0.20 for BXCR). Using all available quantitative chest radiology data, we found that GEE analysis with Model I revealed apparently worse scores in the screened group (p = 0.027 for WCXR, and p = 0.089 for BCXR), but again, if genotype, pancreatic status, and P. aeruginosa-culture results are added, the significant difference disappears (p = 0.081 for WXCR, and p = 0.21 for BXCR). Thus, although there were apparent differences in chest X-ray scores during the longitudinal period analyzed, it is evident that confounders clearly influence the outcomes.
Association of Bronchopulmonary Disease Progression with Respiratory Infections
Differential Impact of P. aeruginosa Infection Further evaluations of the two groups using t test are shown in Table 3 and revealed that at 12 and 14 years of age there were significant differences between the two groups in WCXR and BCXR scores. In our age-bracketed subgroup analyses, however, the differences in clinical characteristics became less significant with fewer patients. Consequently, factors such as genotype and pancreatic functional status that potentially influenced WCXR scores in the group comparisons before the age of 10 years were less influential in older patients, and the significant difference in WCXR scores after the age of 10 years appears to be mostly attributable to P. aeruginosa infection (the third factor in GEE Model II). However, the differences in WCXR scores at age 12 years were related in part to genotype status, especially F508 homozygotes (p = 0.006), although this was not the case in the smaller subgroup at 14 years of age (p = 0.16). Analysis of covariance in the 36 patients at age 12 years revealed that P. aeruginosa infection explained more variability (i.e., 15%) than genotype (i.e., 6%) did on WCXR scores, although both are significant.
Figure 4
sheds more light on why the two groups' WCXR scores are diverging after 10 years of age. When WCXR scores were examined as a function of whether or not the patients had acquired P. aeruginosa at a younger age, we found that respiratory infections with this organism over time seem to largely explain the differences between the two groups. More specifically, it is the screened patients with P. aeruginosa infection who show increasing WCXR scores as they age. In the screened group, P. aeruginosa–positive status was clearly associated with rising WCXR scores (p = 0.008). In addition, we found that the screened group with P. aeruginosa had significantly higher scores than the control group with P. aeruginosa (p = 0.001). Further analyses revealed that screened patients had a much longer duration of P. aeruginosa infection. Specifically, patients in the screened group were culture positive for P. aeruginosa earlier than control patients (median time to culture positive status of 3.01 years, compared with 6.04 years; p = 0.007, Figure 5A)
. Detailed analyses on subgroups of the screened and control groups revealed that it was actually the screened patients in Center B who acquired P. aeruginosa much earlier than all other subgroups (2.04 vs. 5.27, 6.00, 6.50, years in median time) and that Center B had many more patients associated with an old, small clinic (17) than all other subgroups. Specifically, 19 of 28 patients in the screened B subgroup (contrasted with 0 of 28 for screened A, 7 of 21 for control B, and 0 of 28 for control A) received care in an old, small clinic setting that integrated young and old patients irrespective of their pulmonary status, as described previously (17). In addition, the age of P. aeruginosa acquisition is different between the
Other Pulmonary Outcome Assessments We also used GEE analysis to assess longitudinal data on patients who were aged 5 years or older with the binary discriminators of WCXR and BCXR scores because this measure had proved in past analyses to be our most sensitive indicator of potentially irreversible lung disease (19). In comparing 54 screened patients with 44 control patients there was an apparent difference between groups (p = 0.024 for WCXR, and p = 0.055 for BCXR), but again, the difference diminished once the intrinsic risk factors (genotype and pancreatic status) and P. aeruginosa infection were considered (p = 0.19 for WCXR, and p = 0.18 for BCXR). In addition, Shwachman–Kulczycki clinical scores (16) were analyzed, including 1,253 total scores obtained from diagnosis up to 17 years of age. Using all the data, there were no differences in Shwachman–Kulczycki total scores between the two groups (p = 0.75 for GEE Model I, and p = 0.84 for Model II). Although using the data recorded when the patient was aged 5 years or older revealed that the screened group had significantly worse clinical scores than the control groups (p = 0.027), the significance disappeared again once the risk factors and P. aeruginosa infection were adjusted for in the model (p = 0.23).
Results from this randomized, controlled trial demonstrated previously that there are significant nutritional benefits associated with early diagnosis of CF through neonatal screening (6, 7). However, we recognized during the design phase, and subsequently as other CF screening studies were reported (1, 2, 8–11, 13), that greater difficulty would be encountered in assessing pulmonary outcomes. Comparison of screened and control patients has been complicated by center differences and dissimilar group characteristics such as genotype and pancreatic status, despite a large randomization process that was proven to be statistically valid (7). In an effort to avoid disproportions, all newborns were randomized into the screened or control group based strictly on the terminal digit (odd or even) of the laboratory number assigned to the first neonatal screening specimen and not on the basis of reason for diagnosis (15). For example, a patient in whom diagnosis of CF was made in early infancy because of positive family history or because of the presence of meconium ileus had an equal chance of being in the screened as in the control group. Conversely, a patient in whom diagnosis was made later in childhood due to symptoms could have been assigned to the screened group if the child had a false-negative screening test. Although most clinical and demographic characteristics were similar in the two groups (7), it should be emphasized that our control group has intrinsically less risk (19) of bronchopulmonary disease, especially with 28% of the patients having intact pancreatic function compared with 9% of screened patients. In addition, the patients' ages and duration of assessment varied markedly when this analysis was performed due to accrual over 13 years; in fact, only 57.1% of the screened group and 51.1% of the control group have been fully assessed thus far during the critical period of 10 to 14 years. Another unexpected challenge we encountered is that the severity of lung disease has been less than anticipated at the onset of the study (25). Thus, PFT data revealed surprisingly mild degrees of pulmonary dysfunction throughout the evaluation period, and even the CXR scores are in the mild to moderate range of severity (18, 20). This is in accordance with the limited need both groups of patients showed for hospitalization (0.85 hospital days/patient/year), with no group differences either. Nevertheless, quantitative chest radiology, which has proved more sensitive than PFT data in our previous studies (19), enabled us to proceed with meaningful comparisons of the two randomly generated groups and to examine risk factors such as P. aeruginosa acquisition. In summary, our radiographic data revealed two significant findings (1) screening provides a potential opportunity for a better pulmonary outcome in patients in whom diagnosis was made early because the initial chest radiograph was less severely affected and (2) the course of radiographically quantitated lung disease evolves predominately in relation to P. aeruginosa infection. The first point is especially significant because the lungs of patients with CF are histologically normal at birth (14) and therapies developed since this project began (26, 27) offer great potential to delay the onset of irreversible, progressive lung disease. Furthermore, there are newer and better modalities to monitor lung disease in young children with CF (28–30). In retrospect, methods such as high-resolution chest computed tomography imaging (29) and infant PFT assessment (30) would have added value to this study had they been available in 1985. As identified previously (17, 28, 31, 32) and confirmed with these analyses, screened patients in whom diagnosis was made from 1985 to 1989 at one CF Center setting with a small, old clinic acquired P. aeruginosa at younger ages and therefore have had a longer duration of associated respiratory infection. We have also demonstrated that acquisition of P. aeruginosa accelerates radiographically quantitated lung disease (32) when there is no delay in diagnosis with an effect after age 10 years in this birth cohort. Thus, the results shown in Table 3 and Figure 3 indicate that P. aeruginosa infection had a stronger influence on measures of bronchopulmonary infection than did the age when diagnosis of CF was made in the case of patients followed longitudinally in this randomized trial. Our results also demonstrate the interaction of various risk factors that influence pulmonary prognosis. Clearly, genotype and pancreatic functional status are important intrinsic factors that predispose children with CF to chronic lung disease when they develop respiratory infections (19). Undoubtedly, prognosis in CF is determined by the interplay of multiple genetic, nutritional, and environmental determinants. In a randomized trial such as this, however, these factors can become confounders when unavoidable group imbalances occur. Other reports on pulmonary outcomes associated with CF neonatal screening have been based exclusively on observational studies (1, 2, 8–11), but there are supportive epidemiologic data also (33). All published experience suggests that early diagnosis was beneficial, but the differences between screened patients and a variety of control groups are difficult to interpret. In Australia for instance, the screened patients compared with historic controls showed a higher % FEV1 and % FVC at 5 and 10 years of age; however, the mean values were normal and chest radiograph scores were similar between the two groups (5). Observational studies in northeastern Italy (13) also revealed that better pulmonary outcomes were evident with early diagnosis compared with historic controls. In the Netherlands (8, 9), screened patients showed less of a decrease in % FEV1 compared with a concurrent birth cohort that did not have access to CF screening. Finally, a recently completed 10-year observational study in Brittany (11) compared concurrent outcomes in the only nonscreening region (Loire-Atlantique) with data obtained on screened patients and found significantly worse BCXR scores in the former group but no differences in P. aeruginosa colonization or PFTs. Other assessments of P. aeruginosa acquisition compared with age at diagnosis have not revealed differences such as we report in this study, but these reports do not include complete information on extrinsic or intrinsic risk factors (34). It has become increasingly clear that there are many potential risk factors for P. aeruginosa infection in children with CF. Our previous findings (17, 28) combined with other studies (35–37) incriminate long-term oral (non–P. aeruginosa) antibiotic administration, aerosol use, malnutrition, person-to-person exposures, and certain intrinsic characteristics such as pancreatic insufficiency. Although some of these factors are fundamentally iatrogenic, it should be recognized that at the time of diagnosis at a CF Center in the United States, nearly one-third of new patients registered with the CF Foundation already have acquired P. aeruginosa (38) and acquisition/infection can occur in early infancy (1, 2). In addition, our results show that in some clinic settings half the number of patients have P. aeruginosa by 2 years of age (Figure 5) and half the number of those identified by traditional methods already have irreversible lung disease at diagnosis (Table 2). The deleterious effects of P. aeruginosa have been known for many years (39), but the quantitative impact of this pathogen on children with respect to morbidity and mortality has become more evident during the past decade (32, 35, 40–42). For instance, Emerson and coworkers (40) recently reported that P. aeruginosa-positive patients with CF have 2.6 times higher risk of death and significantly lower % FEV1 values. Reduced survival to age 16 years was also reported by Wilmott and coworkers (43), whereas Hudson and coworkers (41) found higher mortality if P. aeruginosa was detected in cultures before 2 years of age. A recent epidemiologic assessment with P. aeruginosa molecular typing in Melbourne (44) identified cross-infection in ambulatory clinics and/or hospital wards as the key factor responsible for five unexpected deaths of children with CF who were less than 5 years old. This alarming experience with a virulent P. aeruginosa strain argues strongly for segregated clinics, as used in one of our ambulatory care settings (17, 31) and in Denmark (45). On the other hand, this issue remains controversial as arguments have been raised against P. aeruginosa–driven segregated care (46). Among many potential advantages, CF neonatal screening provides a special opportunity to reorganize care (15) and emphasize proactive, preventive strategies such as eliminating malnutrition (7) and potentially eradicating P. aeruginosa when it first appears in young patients (26, 27, 47, 48). Fortunately, improved methods of P. aeruginosa surveillance (28) and more effective therapies are now available (26, 27, 47–49). Therefore, regions that are implementing or planning (50, 51) CF neonatal screening should give priority to enhancing care delivery methods because our results confirm that early diagnosis per se does not ensure a better outcome but only provides that important opportunity.
The authors thank the following investigators who have participated in the Wisconsin Cystic Fibrosis Neonatal Screening Project: University of Wisconsin Medical School, Madison—N. Fost, E. Mischler, M. Palta, A. Tluczek, M. Block, L. A. Davis, L. Feenan, D. Pfeil, K. Moucha, L. J. Wei, B. S. Wilfond, A. von Egmond, M. Weatherly, J. Sharp, L. Makholm, L. Loveland, and R. Brown; Medical College of Wisconsin, Milwaukee—W. T. Bruns, H. Colby, W. Gershan, C. McCarthy, L. Rusakow, M. E. Freeman, and K. Riley; and Wisconsin State Laboratory of Hygiene, Madison—G. Hoffman, D. J. Hassemer, and R. H. Laessig.
Supported by National Institutes of Health grants DK 34108 and M01 RR03186 and by Cystic Fibrosis Foundation grant A001-5-01. Conflict of Interest Statement: P.M.F. has no declared conflict of interest; Z.L. has no declared conflict of interest; M.R.K. has no declared conflict of interest; A.L. has no declared conflict of interest; C.G.G. has no declared conflict of interest; J.C. has no declared conflict of interest; H-C.L. has no declared conflict of interest; M.J.R. has no declared conflict of interest; M.L.S. has no declared conflict of interest. Received in original form March 26, 2003; accepted in final form August 6, 2003
This article has been cited by other articles:
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
TOP
ABSTRACT
METHODS
levels of 0.005, 0.01, and 0.035 during the years 2000, 2001, and 2002, respectively. For the last analysis, there were 746 chest films scored and 539 PFTs. Generalized estimating equations (GEE) were used for repeated measures of radiographic scores and PFTs, adjusting for various covariates with the identity link for continuous outcomes and the logit link for dichotomous outcomes (
2 tests of homogeneity were used for discrete variables, and t tests were used for continuous variables at individual time points (
