 |
|
 |
|
 |
 |
|
 |
 |
 |
 |
 |
 |
 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
David J. Henderson-Smart, MB, PhD
Centre for Perinatal
Health Services Research
University of Sydney
Sydney, New South Wales 2006, Australia
Judy M. Simpson, PhD
School of Public Health
University of Sydney
Sydney, New South Wales 2006, Australia
Nicholas J. Evans, DM, MRCPH
Royal Prince Alfred Hospital
University of Sydney
Sydney, New South Wales 2050, Australia
In Reply.—
We thank Manzoni and colleagues for their interest in our report1 and their comments.
Since the first description of retinopathy of prematurity (ROP) in the 1940s, a notable feature of the condition has been the variation in incidence among hospitals. Reasons for this variation are likely to include differences in case ascertainment, case mix, sampling variability (expected to be greater for smaller units), and aspects of both obstetric and neonatal clinical practice. The aim of our study was to identify antenatal and intrapartum factors significantly associated with increased risk of severe ROP. In another study, currently unpublished, we analyzed the differences in the incidence of severe ROP among units of the Australia and New Zealand Neonatal Network, adjusting for case mix as well as for sampling variability,2 to identify units with significantly better outcomes. These data will facilitate investigation of potentially better practices associated with a reduced risk of severe ROP.3
Manzoni and colleagues raise the interesting question of how much weight should be given to data from population-based studies versus local-institution audit when discussing risks of preterm birth with parents. In many cases, the database from individual units will simply be too small to allow unbiased analysis, a factor that has been a key motivation for the growth of neonatal networks.4,5
Manzoni and colleagues reported their data by birth weight groups. As we noted in our report, significantly growth-restricted infants will be overrepresented in such groupings, and different associations may be apparent when analysis is undertaken by gestational age bands.6,7 Looking at the example of vaginal versus operative delivery cited by Manzoni et al, caesarian section (CS) with no labor showed a protective effect for severe ROP on unadjusted analysis in our data set, but it did not retain significance on multivariate analysis.1 In fact, when the effect of method of birth was adjusted for gestation alone, the effect disappeared, giving an odds ratio for CS with no labor of 1.13 (95% confidence interval: 0.76, 1.67).
An additional problem with the analysis presented by Manzoni and colleagues is that they seem to be concluding an effect of mode of birth in 1 subgroup and not the other based on an inappropriate analysis. As many authors have noted, the appropriate test for performing subgroup analysis is a test of statistical interaction; it is not valid to draw this conclusion based on the fact that the P value is significant in 1 subgroup but not the other.8 However, in this example, even if a unit in the Australia and New Zealand Neonatal Network did have sufficient local institutional data to show a valid protective association with CS, although it would be reasonable to convey that information, it would also be appropriate to emphasize that overall, in Australia and New Zealand, there was no such association.
We agree with Manzoni and colleagues that it is important for units to collect their own data and continually review their practices. Our study is aimed at informing and enhancing such a process for all regional neonatal units in both Australia and New Zealand. We would encourage units elsewhere to contribute data to an appropriate network.
REFERENCES
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
 |
||
|
||
Home | My Pediatrics | Journal Information | Current Issue | Past Issues | Subscriptions & Services | Contact Us Click here for faster international access © 2005 American Academy of Pediatrics. All rights reserved. |
||
|
||
