https://orcid.org/0000-0001-6107-7421
TO THE EDITOR:
We thank Cameron Norman McIntosh for the critical review of our work and appreciate his suggestion that generalized additive logit models (GALMs) may provide further benefit to our data. Our analysis focused on citrullinated histone H3–deoxyribonucleic acid (H3Cit-DNA) complexes as a marker for thromboinflammation in ST-elevation myocardial infarction (STEMI). We found that we can reliably identify patients at high risk of major adverse cardiac events (MACE). Additionally, receiver operating characteristic analyses confirmed that the addition of C-reactive protein (CRP) leads to an improvement in prognostic significance.1 Now, we conducted further analyses on our data using GALMs and we received similar results to our initial analysis (Figure 1). Additionally, we could identify an inflection point for H3Cit-DNA (165.07 μg/L) above which the risk of MACE increases clearly (Figure 1A). Compared with the original logistic multivariate regression analysis, the combinations of H3Cit-DNA and CRP, as well as H3Cit-DNA and CRP-positive glomerular filtration rate (Figure 1B), also proved to be the most suitable to predict MACE in patients with STEMI. The area under the curve of the receiver operating characteristic analysis improved only slightly in each case in comparison with our initially used statistical model. The addition of further parameters did not improve the predictive power of MACE. In this specific case, GALMs just slightly improved predictive power of our model. Nevertheless, they strengthen our previous findings and further demonstrate the important role of thromboinflammation in the predictive power of MACE after STEMI. Modern and significantly more complex models have been used for years, particularly in medicine.2-4 Those models, some of which are artificial neural networks, do not always show the desired advantage in terms of the predictive power of a particular event.5 Nevertheless, GALMs have many advantages, even if their application, as in this case, is not always superior to the “classic” logistic multivariate regression analysis.
GALMs for the prediction of MACE within 1 year after STEMI. (A) Associations between H3Cit-DNA complexes, glomerular filtration rate (GFR), and C-reactive protein (CRP) and occurrence of MACE are presented. The smooth blue line shows the spline function, a piecewise cubic polynomial function. Red dashed line represents 95% confidence interval. Red dot shows inflection point at 165.07 μg/L H3Cit-DNA complexes. (B) ROC analyses: AUC of the combination of H3Cit-DNA level to CRP and GFR is shown. AUC, area under the curve; ROC, receiver operating characteristic.
GALMs for the prediction of MACE within 1 year after STEMI. (A) Associations between H3Cit-DNA complexes, glomerular filtration rate (GFR), and C-reactive protein (CRP) and occurrence of MACE are presented. The smooth blue line shows the spline function, a piecewise cubic polynomial function. Red dashed line represents 95% confidence interval. Red dot shows inflection point at 165.07 μg/L H3Cit-DNA complexes. (B) ROC analyses: AUC of the combination of H3Cit-DNA level to CRP and GFR is shown. AUC, area under the curve; ROC, receiver operating characteristic.
Contribution: K.A. analyzed the data; M.B. wrote the manuscript; and A.P. supervised the data analysis and revised the manuscript.
Conflict-of-interest disclosure: The authors declare no competing financial interests.
Correspondence: Amin Polzin, Heinrich Heine University Düsseldorf, Moorenstr 5, 40225 Düsseldorf, Germany; email: amin.polzin@med.uni-duesseldorf.de.
