STATISTICAL-METHODS FOR THE BLOOD BERYLLIUM LYMPHOCYTE-PROLIFERATION TEST

Citation
El. Frome et al., STATISTICAL-METHODS FOR THE BLOOD BERYLLIUM LYMPHOCYTE-PROLIFERATION TEST, Environmental health perspectives, 104, 1996, pp. 957-968
Citations number
24
Categorie Soggetti
Public, Environmental & Occupation Heath","Environmental Sciences
ISSN journal
00916765
Volume
104
Year of publication
1996
Supplement
5
Pages
957 - 968
Database
ISI
SICI code
0091-6765(1996)104:<957:SFTBBL>2.0.ZU;2-Q
Abstract
The blood beryllium lymphocyte proliferation test (BeLPT) is a modific ation of the standard lymphocyte proliferation test that is used to id entify persons who may have chronic beryllium disease. A major problem in the interpretation of BeLPT test results is outlying data values a mong the replicate well counts (approximate to 7%). A log-linear regre ssion model is used to describe the expected well counts for each set of Be exposure conditions, and the variance of the well counts is prop ortional to the square of the expected count. Two outlier-resistant re gression methods are used to estimate stimulation indices (SIs) and th e coefficient of variation. The first approach uses least absolute val ues (LAV) on the log of the well counts as a method for estimation; th e second approach uses a resistant regression version of maximum quasi -likelihood estimation. A major advantage of these resistant methods i s that they make it unnecessary to identify and delete outliers. These two new methods for the statistical analysis of the BeLPT data and th e current outlier rejection method are applied to 173 BeLPT assays. We strongly recommend the LAV method for routine analysis of the BeLPT. Outliers are important when trying to identify individuals with beryll ium hypersensitivity, since these individuals typically have large pos itive SI values. A new method for identifying large Sis using combined data from the nonexposed group and the beryllium workers is proposed. The log(Sl)s are described with a Gaussian distribution with location and scale parameters estimated using resistant methods. This approach is applied to the test data and results are compared with those obtai ned from the current method.