       Document 3066
 DOCN  M94A3066
 TI    A community-based study of the reliability of absolute CD4 lymphocyte
       counts in HIV infected subjects.
 DT    9412
 AU    Jimmerson C; Bockmon K; Harrison J; Houston Clinical Research Network,
       TX 77006.
 SO    Int Conf AIDS. 1994 Aug 7-12;10(1):161 (abstract no. PB0070). Unique
       Identifier : AIDSLINE ICA10/94369509
 AB    OBJECTIVES: CD4 cell counts are commonly used as surrogate markers. High
       variability within these markers could limit their usefulness as a tool
       for making treatment decisions. The variability of CD4 counts and
       percentages performed by commercial laboratories was investigated.
       METHODS: Three nationwide, contract laboratories were shipped blinded
       samples from HIV-infected subjects (N = 52; CD4 > 100 and < 900). Two
       samples per patient per laboratory were drawn and shipped concurrently.
       Fixed effects, repeated measures ANOVA, and generalizability theory were
       used to quantify sources of variation associated with results. RESULTS:
       Absolute CD4 Lymphocyte Counts--Fourteen percent of the variance across
       laboratories was associated with various sources of error rather than
       differences among patients. The standard error of measurement for CD4
       counts was approximately 70 cells/mm3. ANOVA indicated a significant (p
       < .001) laboratory effect. The difference between samples was
       significant (p < .02) within one of the laboratories and was often
       clinically significant within all three laboratories. The overall
       generalizability coefficient was .88 for a single sample from a single
       laboratory. CD4 lymphocyte counts were not generalizable across
       laboratories and thus not reliable for clinical decision making. CD4
       Percentages--Five percent of the variance was associated with error. The
       standard error was approximately 2%. A significant laboratory effect was
       also observed (p < .001), and a difference (p < .01) between samples was
       also found for the one laboratory noted above. The generalizability
       coefficient was .95. CD4 percentages were more generalizable than
       absolute cell counts. Variances in CD4 percentages were rarely
       clinically significant. CONCLUSIONS: Future research involving a larger
       number of laboratories and more stringent quality control mandates are
       indicated. CD4 percentages may be more reliable than absolute cell
       counts.
 DE    Biological Markers  Human  HIV
       Infections/CLASSIFICATION/DIAGNOSIS/*IMMUNOLOGY  *Leukocyte Count
       Predictive Value of Tests  Reproducibility of Results  T4
       Lymphocytes/*IMMUNOLOGY  MEETING ABSTRACT

       SOURCE: National Library of Medicine.  NOTICE: This material may be
       protected by Copyright Law (Title 17, U.S.Code).

