Jk. Sont et al., REPEATABILITY OF MEASURES OF INFLAMMATORY CELL NUMBER IN BRONCHIAL BIOPSIES IN ATOPIC ASTHMA, The European respiratory journal, 10(11), 1997, pp. 2602-2608
Airway pathology is increasingly considered to be a major outcome in a
sthma research. The aim of this study was to examine the intra-observe
r, within-section and between-biopsy repeatability, together with the
implications for statistical power of a computerized quantitative anal
ysis of inflammatory cell numbers in the lamina propria in bronchial b
iopsy specimens from atopic asthmatic subjects. Thirty six atopic adul
ts (aged 19-40 yrs) with mild to moderate asthma (baseline forced expi
ratory volume in one second (FEV1) greater than or equal to 50% of pre
dicted value, methacholine (PC20) range 0.02+/-18.2 mg.mL(-1)) at vari
ous levels of treatment (25 subjects on inhaled steroids) entered the
study. Biopsies were taken from the (sub)segmental carinae of the righ
t lower and middle lobe and from the main carina. Specimens were snap-
frozen and immunohistochemical staining was performed on cryostat sect
ions with monoclonal antibodies against: (secreted) eosinophil cationi
c protein (EG1, EG2), mast cell tryptase (AA1), CD45, CD22, CD4, CDS,
CD25, and CD45RO. Using a computerized system, the number of positivel
y stained cells in the lamina propria was counted. When considering al
l cell types together, satisfactory intraclass correlation coefficient
s (ICC) values were obtained for intra-observer, within-section and be
tween-biopsy repeatability, being 0.90, 0.80 and 0.81, respectively. T
he analysis of repeatability for individual cell types revealed ICC va
lues ranging 0.47-0.82 for intra-observer, 0.44-0.76 for within-sectio
n and 0.37-0.67 for between-biopsy repeatability. The results imply th
at a sample-size between eight and 25 subjects is needed to detect at
least one doubling difference in cell number per 0.1 mm(2) for a parti
cular inflammatory cell type in a study, using a within-group design w
ith alpha=0.05 and power of 0.80. A sample-size of 13-48 subjects per
group is required to detect the same difference between the groups in
a parallel design.