| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
1 Department of Radiology, 2 Departments of Orthopaedic Surgery and Biostatistics, and 3 Division of Pulmonary Medicine, Children's Hospital, Harvard Medical School, Boston, Massachusetts
Correspondence and requests for reprints should be addressed to Robert H. Cleveland, M.D., Department of Radiology, Children's Hospital, 300 Longwood Avenue, Boston, MA 02115. E-mail: Robert.Cleveland{at}childrens.harvard.edu
ABSTRACT
We compare a chest radiographic scoring system developed by our group to spirometry in quantifying the longitudinal progression of lung disease among cystic fibrosis (CF) patients, and we evaluate the use of this radiographic scoring system in identifying the treatment effect of an inhaled antibiotic. Results suggest that longitudinally acquired chest radiographs, when scored using our scoring system, are at least as sensitive as lung function in detecting the progression of lung disease in CF patients.
Key Words: chest radiograph • cystic fibrosis
In this article, we discuss the development and use of a system based on scoring of longitudinally acquired chest X-rays (CXR) to evaluate progression of lung disease in cystic fibrosis (CF) patients. The first part will be divided into two subsections, (1) a presentation of the scoring system (1, 2); (2) the use of the system to assess treatment effect (3).
THE AGE-BASED SEVERITY SCORE (ABS)
Several years ago a system was developed based on scoring longitudinally acquired CXR in a large cohort of patients with CF. Scoring was performed using the Brasfield system (4). The system is referred to as the Age-Based Severity score (ABS). The methodology and statistical analyses employed in development of this system have been thoroughly presented elsewhere (1, 2). Comparisons of rates of decline of ABS score were expressed in Z scores based on a mixed-model ANOVA analysis. Images were acquired as needed for clinical purposes, therefore the number of images varied from patient to patient and yearly. Those wishing greater detail concerning methodology and statistical analysis are referred to the prior publications (1, 2).
The ABS was derived from individual scoring of 3,038 CXR from 230 patients. This number of CXR was chosen to develop 95% confidence intervals around one Brasfield point across the spectrum of ages. Male/female distribution was essentially equal (m = 106, f = 124); the age range for which radiographs were available spanned three days to over 50 years. The mean number of CXR per patient was 13 (range ± 12) (1, 2). The Brasfield system (4) was the preferred scoring system because of its high inter- and intra-rater reliability and its high correlation with pulmonary function tests (PFT) (1, 2). For the ABS (Figure 1), the solid line represents the mean; the dotted lines are two standards errors of the mean. Two readers were employed. The senior reader scored 2,120 images from 194 patients, the other scored 918 CXR from 34 patients. Eighty patients with 833 radiographs were randomly selected for blinded second reading by both observers. Interobserver reliability was high at a value of 0.81 (p < 0.001). The intraobserver reliabilities were also quite high with values of 0.92 for reader 1 and 0.85 for reader 2 (p < 0.001 for both).
|
|
In a subsequent study, the ABS system was used to evaluate the effectiveness of tobramycin solution for inhalation (TSI) (3). The methodology and statistical analyses employed in this study have been thoroughly presented elsewhere (3). Comparisons of rates of decline were expressed in Z scores based on a mixed-model ANOVA analysis. Those wishing greater detail concerning methodology and statistical analysis than presented below are referred to the prior publication (3). There were 38 patients in the database who had received treatment with TSI; 24 had received DNase simultaneously and were excluded from the study. The remaining 14 patients had 282 CXR. Male/female ratio was 9/5; age range of youngest to oldest patient at time of image acquisition was 2 months to 22 years. There was a high range of variability in the number of images acquired per patient (mean = 20; range, 2–57). The mean duration of TSI therapy was 16.5 months (range, 1–48 months). Scores were regressed, and three curves for rate of decline were generated: (1) Brasfield scores for the total ABS group previously published (1, 2) (surrogate control group), (2) the TSI group of patients including time on TSI therapy, and (3) the TSI-treated group limited to those scores acquired before the commencement of TSI treatment (Figure 3). Comparison of the TSI-treated group and the surrogate control group revealed an ABS annualized rate of decline of 0.150 Brasfield points per year for the TSI-treated group and 0.175 Brasfield points per year for the surrogate controls. This represents a 14 percent improvement for patients on TSI (p < 0.001). In addition, the rate of decline for the TSI-treated patients before therapy was compared with that of the ABS control group. The TSI group before treatment had a rate of decline of 0.169 Brasfield points per year as opposed to the 0.175 Brasfield points per year for the ABS control group, representing a difference of 3%, which was not significant (p = 0.859).
|
DISCUSSION
An ABS-like database could potentially be a useful surrogate measure for large observational studies at a cost in radiation exposure and dollars significantly lower than other imaging modalities. These arguments are further emphasized by the fact that in many instances the radiographic images have already been acquired for clinical purposes and thus there is no additional X-ray exposure or financial costs. The images are retrospectively available, often back to infancy. The multiplicity of images available in creating a chest radiograph database is impractical for any other imaging modality. Infant PFTs are not widely available, and widespread use of CT in infants is impractical; therefore, serial radiographs often provide the only information for children under 1 year of age. Interest, therefore, has been expressed in the value of the ABS in very young children. The original ABS study (1, 2) showed that infants often were first seen because of an acute disease process (either viral airways disease or pneumonia) that was reflected in a lower Brasfield score. After the acute process resolved, the scores improved. For many of the infants, the initial process affecting their scores was bronchial wall thickening. Assessing subtle changes in bronchial wall thickening by CXR remains difficult. These factors combine to limit the ability to evaluate progression under age 1 year. However, there was a detectable overall improvement in aggregate scores between ages 1 and 2 years (presumably reflecting a transiently lower Brasfield score in infants who present with acute disease) and a relative plateauing of scores between 2 and 4 years with the beginning of an initial decline after 4 years of age. This suggests that assessing progression of CF lung disease under 1 year of age remains difficult, but that between ages 1 and 5 years, an ABS-like database can provide information that may not otherwise be available.
The use of a pre-existing ABS-like database suffers from being developed from retrospectively acquired data. Also, since there may be secular changes in trends of CF lung disease progression, any such database which is "frozen in time" may be affected. However, for the published ABS (1, 2), there were no significant effects when differences in rates of decline for the database from 1965–1980 was compared with data acquired from 1981–1995 (p = 0.27). Since the ABS is derived from an unselected group of patients with CF with variable disease severity, trends may be biased by individual patient disease severity. However, when the published ABS (1, 2) was compared with that of the TSI-treated patients before they were begun on TSI, there was no significant difference in rate of decline (p = 0.859). To our knowledge, no similar study evaluating CF lung progression using CXR has been published. Studies using serial PFTs as the outcome parameter in the evaluation of effectiveness of TSI (5, 6) have shown similar results.
Although the creation of an ABS-like database may be time consuming, it is relatively easy to do so with a high degree of inter- and intra-reader reliability. For the original ABS development, the senior reader, a pediatric radiologist with 17 years of experience, trained the other reader, a radiology resident, for approximately 30 minutes in the use of the Brasfield system. No further training or joint efforts were needed to produce the high inter-and intra-reader correlations reported (1, 2). Although CT provides more precise anatomic detail for any one study, the need for multiple repeated studies over time to develop longitudinal data makes CT impractical for this purpose because of radiation dose, cost, and accessibility. The sensitivity of the CXR-based tool has been shown to equal PFTs when specific subsets of the CF population have been evaluated over periods of time within which their CXR and PFT have been compared against each other (3).
Furthermore, our preliminary results from a study investigating CFTR class effect suggests that for groups containing relatively few CXRs and PFT data points, the CXR data may be more sensitive in detecting differences than either FEV1 or FVC (unpublished data).
CONCLUSIONS
These studies suggest that aggregate scores derived from serial chest radiographs of large cohorts of patients correlate with PFT data. In the study of effectiveness of TSI, the use of an ABS-like database as the outcome parameter has produced results which parallel studies using PFT as the outcome measure. In large cohorts, the ABS (or an ABS-like) radiographic database and PFT database demonstrate similar rates of decline. However, in small cohorts, the imaging data may be more sensitive in showing statistically significant differences than either FEV1 or FVC. This presumably relates to the fact that the X-ray database is looking at the evolution of the anatomic disease before it is manifested as a physiologic abnormality. This has also been observed when CT results were compared with PFT (7).
FOOTNOTES
* Current address: Division of Pediatric Pulmonology, University of Miami, Batchelor Children's Institute, Miami, Florida. 
Current address: Paediatric Respiratory Physician, Children's Hospital, University College Dublin and Mater Misericordiae Hospital, Dublin, Ireland.
Conflict of Interest Statement: None of the authors has a financial relationship with a commercial entity that has an interest in the subject of this manuscript.
(Received in original form November 30, 2006; accepted in final form March 27, 2007)
REFERENCES
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
5 yr 6 mo and < 6 yr 6 mo). Age 1 represents ages younger than 1 year 6 months; age older than 30 includes all ages older than 30 years 6 months. Note that although scores slowly decline overall, relative plateaus in decline occur from 2 to 4 years, from 8 to 15 years, and after 22 years. Reprinted by permission from Reference 1.
