the mean values of variables used in the comparison of groups. As with the regres sion models used, the analysis of covariance provides a statistical rather than an experimental method to control and adjust for the effects of one or more uncon trolled variables.583 The influence of the covariate is removed using a linear regres sion method, residual sums of squares are then used to provide variance estimates that are then employed in tests of significance.583 Adjusted means of the dependent variable are obtained, and the adjusted mean shows which part of the variation in the m eans betw een groups rem ains when the v ariatio n due to the co v ariate is removed.583 Necessary in the analysis is the calculation of the within group variable variance and the between group variable variance.
In this analysis, the covariates used were obtained simultaneously as the dependent variable, using the same blood sample. In addition, the variables assessed are intrin sic attributes of the subjects studied. Furthermore, a number of assumptions are made in this analysis. In view of the lack of overt familial forms of hyperlipoprotei- naemia or other known genetic influences on the dependent variable, it will be assumed that the differences between the groups are not genetically determined for the factors analysed. A direct causal effect is assumed between the dependent vari able and the covariates. The statistical assumption for the presence of homogeneous regression coefficients is also made in this method (MANOVA in the SPSSX).
4.5.3 Results
4.5.3.1 Platelet count. There was no measurable relationship between the platelet count and plasma TXB2 levels, plasma BTG levels, Hb, He or WCC in the young normal (control) group. A very small but consistent positive relation between the platelet count and the serum TXB2 level was present in the groups with CHD, but not in the young normal group (Table 4.5).
4.5.3.2 Serum Thromboxane B2. The amount of TXB2 formed after incubation for 1 hour at 37°C had a significant negative relationship with the PCR in the older CHD group and in the young control group (Table 4.5). Although the correlation with the PCR is only small, there is clearly a reproducible association in these groups but not in the young CHD group. Besides the platelet count, the serum TXB2 level correlated with Hb, He and WCC in the two CHD groups but not in the control group (Table 4.5). The serum TXB2 level correlated negatively with the BTG level in the normal group (r=-0.38, p<0.005), although the serum TXB2 level was not related to the plasma TXB2 or plasma BTG levels in patients with CHD.
Using a stepwise regression model, with serum TXB2 as the dependent variable and age, Hb, He, WCC and PC as the independent variables, only the PC and WCC re mained significantly and independently associated with serum TXB2 levels in the young case group, while the WCC was the only variable associated with serum TXB2 level in the older CHD group (Table 4.6). There were no significant associa tions in the control group for the regression analysis using the same variables.
An analysis of covariance was performed using MANOVA (SPSSX), with the de pendent variable being the serum TXB2 level, the covariates being Hb, He, PC and WCC, and the case and control groups being analysed as factors, that is, as categor ical variables defined in two groups. In fact, these two groups were analysed togeth er by the analysis of covariance method. The cases and controls were considered as categorical values and defined as two factors in the analysis of covariance. The adjusted means for serum TXB2 values, that is the means adjusted for the effects of the covariates Hb, He, WCC and PC, remained very different for the case group (adjusted logarithmic mean being 4.56 ng/ml) compared to the control group (adjust ed logarithmic mean being 1.33 ng/ml). The summary of the analysis is presented in Table 4.7. The means are expressed as the logarithm since the serum TXB2 value was originally logarithmically transformed to obtain a normal distribution to allow Gaussian statistical methods to be used.
4.5.3.3 Platelet Count Ratio. The PCR is a measure of ex vivo platelet reactivi ty. A consistent and significant relationship between the PCR and other blood cell indices was only observed in the older CHD group (Table 4.8), but not in the young CHD group, except for a negative relationship with the WCC (r=-0.45 ,p<0.001). The PCR correlated negatively with serum TXB2 levels in the control group and in the older CHD group. In order to evaluate the importance of other cells in determin ing the association between the serum TXB2 level and the PCR in the older CHD group, the correlation was further examined using partial correlations. This method adjusts for the effects of independent variables on any of the measured zero order partial correlations. The association between the serum TXB2 levle and the PCR diminished after adjusting for Hb, He, WCC and PC, although only to a small degree (Table 4.9).
4.5.3.4 In vitro platelet aggregation. A positive association was demonstrated between the reactivity of platelets in PRP to specific platelet agonists and serum TXB2 values in both the control group and the case group. In particular, in the control group, a strong consistent association existed between platelet aggregation and the amount of TXB2 formed (Table 4.10). This measurement was not performed
in the older group with CHD.
4.5.3.5 Influence of cardiac medications on the associations. In the older CHD