The authors reply

Abstract

Our study (1), recently published in Critical Care Medicine, is based on fundamental fluid dynamics and rigorous mathematical analyses (2). We derived the intrinsic frequency (IF) concept (3) from experiments and simulations and subsequently demonstrated the correlation between IF and left ventricle (LV) systolic function (3). Estimation of LV ejection fraction (LVEF) based on the IF of the arterial waveform was a logical next step. We disagree with Michard et al (4) that echocardiography remains the gold standard technique to accurately assess LVEF. Though echocardiography is ubiquitous, numerous studies have shown the superiority of MRI over echocardiography for assessing LVEF and LV volumes (see references in Table 2 of Pahlevan et al [1]). The criticism by Michard et al (4) of using MRI as the reference for our derived LVEF therefore seems at odds with established knowledge ([5, 6] and references within), based on which we specifically chose MRI as the most accurate method to study the relationship of IF to LVEF ("…MRI as the reference standard for assessment of global LV function" [6]). We reported r values to be consistent with previously reported comparison studies of LVEF (references in Table 2 of [1]). In this respect, we are puzzled by conversion of r value to r2 in the letter by Michard et al (4). They argued that our r2 is low but failed to show a basis for comparison. We emphasize that based on data presented in our Table 2 (1) (which summarizes other relevant clinical studies), the r value of our results is comparable with r values reported for 2D-echocardiogram. Multiple studies ([5] and references in Table 2 of [1]) have reported limits of agreement (LOA) of 2D-echocardiography compared with MRI and radionuclide ventriculography (RNV) to be around ±20%. Therefore, the LOA of ±19.1% for our iPhone method is a very reasonable value considering its cost and ease-of-use in comparison with 2D-echo. The same drawback that Michard et al (4) mentioned about our LOA is also applicable to 2D-echo. For example, we refer to LOA of 2D-echo when it was compared to RNV (LOA = –27%, +12%) and MRI (LOA = –24%, +20%) in the study by Bellenger et al (5) in heart failure (HF) patients. Furthermore, our correlation (r = 0.74) and LOA (±19.1%) are well within the range of 2D-echo correlation and LOA as reported by multiple investigators (Table 2 in [1]). We acknowledged that further studies with more patients with lower LVEF (specially ejection fraction < 20) are required to assess sensitivity and specificity of IF accurately. In order to help readers judge the screening value of our iPhone method, we provided the boxplot (Fig. 1) based on the data presented in Figure 2 of Pahlevan et al (1). Further studies are required to confirm and validate our proposed technology before introducing it to the market. However, we believe that an r value of 0.74 and LOA equals to ±19.1% is a reasonable start for a new technology; importantly, the higher r value for HF patients suggests the potential for IF to be used to adjust medicines and prevent readmissions. We look forward to optimizing the IF method for clinical use based on ongoing research.

Additional details