Variation Among Immunoreactive Trypsinogen Concentrations, Michigan Newborn Screening, 10/2007-4/2008 Steven J. Korzeniewski, MA, MSc, Maternal & Child Health Epidemiology Section Manager Grigorescu, V., Young, W., Hawkins, H., Cavanaugh, K., Nasr, S.Z., Langbo, C.
12
Embed
Steven J. Korzeniewski, MA, MSc, Maternal & Child Health Epidemiology Section Manager
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Steven J. Korzeniewski, MA, MSc,Maternal & Child Health Epidemiology
Section Manager
Grigorescu, V., Young, W., Hawkins, H., Cavanaugh, K., Nasr, S.Z., Langbo, C.
Outline
•Background•Research question •Methods •Results•Discussion•Public Health Implications
Background • CF Screening in MI commenced Oct. 2007
• IRT is used to identify infants at increased risk of CF for DNA testing.
• Mutation analysis, using a panel of 40 CF mutations among > 96th percentile.
• In the absence of a mutation sweat testing is recommended only among infants having IRT concentrations > 99.8th percentile.
Research Question•Anecdotal evidence suggested a high rate
of false positives among NICU infants
•This study explores variations in IRT concentrations in hopes of developing a strategy to reduce false positives.
•R1: Do IRT concentrations vary among the general population by sex, race, birth weight, gestational age, and fetal growth ratio?
Methods• Data: Newborn screening IRT concentrations and infant
demographic data collected from Oct 2007-April 2008 were used for this study.
• Analysis: Crude and adjusted generalized linear models (GLM) of the association between demographic variables and IRT concentrations▫ Least squares means and p-values are reported
▫ LS-means are within-group adjusted means, they estimate the marginal means for a balanced population (as opposed to the unbalanced design). Also called estimated population marginal means by Searle, Speed, and Milliken (1980).
▫ We also calculated means and percentiles (96th, 99.8th) by race and gestational age strata
ResultsGeneralized Linear Models of the Crude & Adjusted Associations between Initial Screening IRT Values & Subject
Demographics, Michigan Newborn Screening, Oct. 2007- April 2008
DemographicsCrude Adjusted*
IRT (LS Mean) P-value IRT (LS Mean) P-value
Race (n=59150)
American Indian 26.4
<.0001
27.7
<.0001
Arab Descent 26.4 27.7
Asian/Pacific Islander 22.9 24.1
Black 33.2 34.3
Multi-Racial 27.1 28.4
White 25.9 27.1
Gestational Age (n=67643)
LT 28 wks 27.1
<.0001
28.3
<.000128-37 wks 29.0 28.5
>=37 wks 27.1 27.9
Birth Weight (n=67643)
<1800 grams 29.3
<.0001
29.6
<.00011800g-2500g 28.7 27.9
>=2500g 27.0 27.1
Sex (n=64803)Female 26.9
<.000127.9
<.0001Male 27.7 28.5
FGR (n=48039)
< 25th % 29.1
<.0001
30.9
<.0001>= 25th% & < 75th% 27.5 29.6
>=75th% 26.6 28.9
Race*GA 0.0005
*Adjusted for other demographic covariates included in the table (Race, GA, BW, Sex) and age at specimen collection. (N=58,789)
ResultsIRT Means & Percentiles (96th & 99.8th) by Race & Gestational Age, Michigan Newborn Screening, Oct/2007-April/2008, N=59,150
*Total N exceeds the sum of racial categories because it includes those records missing race information.
Updated Results•At one year (Oct 2007-Oct 2008)
▫Effect modification of race by gestational age absolved
▫Racial variation remained significant in both crude & adjusted models
Conclusion•Failure to account for racial variation results
in:▫Over sampling of black infants
those at lower risk of CF▫Under sampling of white infants
those at the greater risk of CF
•False positive and false negative rates could be inflated ▫However, no false negatives have been
detected thus far
Public Health Implications• Calculation of IRT % cutoffs stratified by race would:
▫ Reduce the FPR & Improve PPV
▫ Require further research to discern appropriate cutoffs, particularly for racial minorities or those with missing data
▫ Require significant change in laboratory operating procedures Sorting of cards Verification of Race information Development of strategy to calculate cutoffs over time
Acknowledgements
•Co-Investigators: Grigorescu, V., Young, W., Hawkins, H., Cavanaugh, K., Nasr, S.Z., Langbo, C.