Background The measurement of risk factors at the beginning of the study is the base of up to date various statistic analyzes concerning risk of incidence occurrence in prospective studies. At follow-up period the risk factors change as well as their predictive value. The factors’ predictive value is connected by natural time-dependent variability. Then the dilution ratio was evaluated. This study aimed to investigate the regression dilution ratio of blood pressure SBP and DBP on the risk of cardiovascular death (CD) during 14-years follow-up period.
Material and methods The CD risk was evaluated in the population of 2646 Warsaw inhabitants (male and female) aged 35-64. 99 individuals were first time examined in 1984, then followed-up in 1988 and 1993. At every examination blood pressure-SBP and DBP, age and other CVD risk factors were assessed. For the estimation of risk of CD the Cox proportional hazards model was applied.
Results During 14-years period the 249 CDs were confirmed in whole examined group. SBP and DBP at baseline were significant risk factors of CD. The relative risks were 1.027 in men and 1.028 in women for SBP and 1.04 in men and 1.04 in women for DBP. For the next points of examinations the influence of blood pressure on the risk of CD decreased. According to the quartiles of SBP and DBP dissection at baseline, the increase of mean values these factors were observed in the l quartile and the decrease of these means were observed in IV quartile. The regression dilution ratios were ca 0.70 for SBP and 0.67 for DBP between 1984 and 1988 and 0.56 for SBP and 0.54 for DBP between 1984 and 1993. After correction of SBP and DBP according to the dilution ratios the relative risks were almost the same for 1984, 1988 and 1993 screens.
Conclusions 1. The regression dilution ratio measures the decrease of predictive power of risk factor in follow-up study.
2. If the dilution ratio in not included in the analysis the predictive value of risk factors in underestimated.
3. Especially regression dilution ratio is recommended in meta-analysis studies.