Article
Modelling coronary calcium progression – the learning sample: Five year follow-up data of the Heinz Nixdorf Recall Study
Search Medline for
Authors
-
Nils Lehmann
- Institut für Medizinische Informatik, Biometrie und Epidemiologie, Essen
-
Susanne Moebus
- Institut für Medizinische Informatik, Biometrie und Epidemiologie, Essen
-
Stefan Möhlenkamp
- Westdeutsches Herzzentrum, Essen
-
Amir Mahabadi
- Westdeutsches Herzzentrum, Essen
-
Axel Schmermund
- Kardioangiologisches Zentrum Bethanien, Frankfurt / Main
-
Andreas Stang
- Institutfür klinische Epidemiologie, Halle / Saale
-
Thomas Budde
- Klinik für internistische Medizin und Kardiologie, Alfried Krupp Krankenhaus, Essen
-
Klaus Mann
- Abteilung für Endokrinologie, labormedizinische Forschung, Essen
-
Raimund Erbel
- Westdeutsches Herzzentrum, Essen
-
Karl-Heinz Jöckel
- Institut für Medizinische Informatik, Biometrie und Epidemiologie, Essen
| Published: | September 2, 2009 |
|---|
Outline
Text
Background: It is now acknowledged that coronary artery calcium (CAC) is both positively associated with cardiovascular risk factors and predictive of coronary events. This suggests studying CAC progression and its relation to risk factors. Present data of the Heinz Nixdorf Recall Study allow to investigate this topic in a large (n=4487) population-based cohort.
Methods: Two electron beam scans were performed roughly 5 years apart. The learning sample comprised 1600 participants (age at baseline 45-75, 54.4% women) without coronary heart disease. Clinically meaningful classes of CAC (0-9;10-99;100-399;400-999;>=1000) are used to determine rates of progression. We further studied the association of progression with cardiovascular risk factors applying linear regression to log(Q)=5*(log(CAC(t1)+1)-log(CAC(t0)+1))/(t1-t0). Here t0 stands for baseline and t1 for follow-up examination date, and (t1-t0) is given in years. Q is the 5-year factor of progression in CAC+1. As an example for progression modelling, we investigate the effect of baseline smoking behavior.
Results: CAC progressed by one or more classes in 26.7%. Regression (mostly from 2nd to 1st class) was present in 2.6%. Median (Q1;Q3) 5-year progression was by a factor Q of 1.41 (1.00;2.59). The distribution of log(Q) was fairly symmetric and unimodal without long tails. In crude, age- and sex-adjusted, and in Framingham-risk-factor-adjusted analyses, 5-year progression was increased in baseline smokers [former smokers] vs. never smokers by
16.3% (95% CI 1.0%;33.9%; p=0.04) [5.7% (-6.2%,19.1%; p=0.36],
27.3% (10.2%;47.2%; p=0.001) [5.7% (-6.6%,19.6%; p=0.38],
28.1% (10.7%;48.1%; p=0.0009) [5.7% (-6.6%,19.7%; p=0.38], respectively. The models explained only a small fraction of the total log(Q)-variance, namely 0.3%, 3% and 4%.
Discussion: Progression to higher CAC classes in the study period was frequent, while regression was rare and may be mostly attributable to image noise. The proposed linear modelling of log(Q) yields reasonable associations between progression and risk factors. Further aspects of model building will be discussed.
