Background and objective: Multivariate analyses on clusters of metabolic an
d hemostatic risk indicators implicitly assume good test-retest reliability
of these variables, substantial covariance among the various indicators, s
tability of covariance structure over time, and comparable covariance struc
ture in different subpopulations. The aim of the present study is to invest
igate these assumptions.
Methods: Repeated samples were taken of fasting insulin, triglycerides (TG)
, high-density cholesterol (HDL-C), low-density cholesterol (LDL-C), fibrin
ogen, tissue-type plasminogen activator (t-PA) antigen, t-PA activity, plas
minogen activator inhibitor-1 (PAI-1) antigen to address their intra-week r
eliability and covariance structure. In the same work-week blood was drawn
three times from 125 sedentary males (age 45.2 +/- 5.3 years) and twice fro
m 132 female nurses (age 33.7 +/- 8.0 years).
Results: About half (44.8%) of these women were oral contraceptives (OC) us
ers. Only minor intra-week changes in absolute levels were found. Intra-wee
k test-retest correlations varied between 0.52 (t-PA activity) and 0.94 (HD
L-C) with an average value of 0.81. In men, non-OC using women, and OC usin
g women, the covariance matrices of the eight risk indicators were equal at
day 1 and day 3, testifying the good stability of covariance structure ove
r time. Differences in covariance structure of all three groups were observ
ed, which remained after correction for BMI and age. In men and non-OC-usin
g women, significant correlation was found on all days between insulin and
the other risk indicators with exception of fibrinogen and LDL-C. In OC use
rs, insulin was correlated with TG, LDL-C, and fibrinogen.
Conclusion: The metabolic and hemostatic risk indicators showed good test-r
etest reliability, and their covariance is stable over time. Multivariate a
nalyses of this cluster should be performed separately for men, non-OC-usin
g women, and OC-using women. (C) 2001 Harcourt Publishers Ltd.