Response to the Letter by Kumarswamy et al
We are appreciative of the interest that Kumarswamy and coauthors show in our description of circulating microRNAs (miRNAs) as novel biomarkers in heart failure (HF).1 The argument they put forward on the one hand reinforces our findings and on the other reiterates our final remark that “larger studies are needed to confirm the diagnostic capacity of identified miRNAs.”2
The first point of Kumarswamy et al1 is that the c statistic we used is “relatively insensitive.”
This argument actually strengthens the case for the miRNAs we identified: relative insensitivity is not a problem for the biomarkers we already identified and pleas for the robustness of the biomarkers we did find. This may simply mean that we have discarded other miRNA based biomarkers that may still be relevant but did not meet our statistical stringency.
Kumarswamy et al1 also question the use of the c statistic as they propose that appropriateness (of the c statistic) in identifying prognostic biomarkers of potential clinical utility was questioned recently.3
However, while according to the paper3 cited by Kumarswamy et al, the c statistic is less suited for the assessment of prognostic risk models this measure is appropriate in the evaluation of diagnostic accuracy,3 which is exactly what we have used c statistics for in our report.2
The third argument is that we did not provide a “net reclassification improvement for certain or combination of miRNAs on top of the traditional markers.”1
The now established traditional marker in HF is (NT-)proBNP. As described in our report,2 the diagnosis of “heart failure” was not only based on clinical criteria, but also on NT-proBNP levels. Therefore, it is fundamentally impossible to test reclassification within our study against a marker that we used to classify the patients.
Kumarswamy et al1 emphasize the age difference between the healthy controls and the HF patients. However, as stated in our paper2 “we explicitly chose to compare HF cases not only to controls, but also to dyspneic patients who were free of clinical HF.” It seems to have escaped the attention of Kumarswamy et al that in this clinically most important comparison, the dyspneic patients free of clinical HF were of similar same age as the HF patients (mean age 65.5 versus 68.2, P=0.527, see Table 1 in the article2). In addition to this, our data do not suggest a correlation between age and levels of miR423-5p (Spearman correlation=0.073, P=0.614 in the dataset including HF patients and dyspneic patients free of clinical HF).
The final remark that larger studies are needed1 is almost identical to the final remark in our Discussion,2 where we already stated that our study is limited by the number of patients and that larger studies are needed.
In conclusion, Kumarswamy et al1 reiterate our call for larger studies. They further stipulate that we used a relatively insensitive method to select circulating miRNAs as putative biomarkers, which, however, actually reinforces that those we did identify are most likely to be robust. Finally, they suggest that age differences may play a role. However, the 2 clinically most relevant groups we compared (dyspneic patients with or without HF) were of comparable age. Therefore, we share the excitement of this first identification of this novel class of biomarkers in HF and look forward to larger clinical studies to appreciate the clinical value of the miRNAs we identified.
Sources of Funding
Kumarswamy R, Anker SD, Thum T. MicroRNAs as circulating biomarkers for heart failure: questions about MiR-423-5p. Circ Res. 2010; 106: e8.
Tijsen AJ, Creemers EE, Moerland PD, de Windt LJ, van der Wal AC, Kok WE, Pinto YM. MiR423-5pAs a Circulating Biomarker for Heart Failure. Circ Res. 2010; 106: 1035–1039.
Cook NR. Use and misuse of the receiver operating characteristic curve in risk prediction. Circulation. 2007; 115: 928–935.