To the Editor We read with great interest the article by Lanspa et al. (1), in which they demonstrate that both vena cava collapsibility index (VCCI) and stroke volume variation (SVV) may be used to “predict hemodynamic response” in “non–mechanically ventilated” patients. The authors are to be congratulated for doing novel research on spontaneously ventilating patients with septic shock. Unfortunately, there are a number of serious issues we would like to address relative to the study design and the authors’ conclusions that SVV can be used to reliably assess volume responsiveness in these patients when this is not supported by the data. Although the authors acknowledge a very small sample size (N = 14) and that their results lack external validation, their methodology is still weak. With a sample size this small, all statistical inferences should be nonparametric, and it appears that only some were. The main analysis was calculating three c-statistics (area under the receiver operating characteristic curve) for three independent variables: VCCI, SVV, and aortic blood velocity variation with the binary outcome of an increased cardiac index 15% or greater. The authors conclude that VCCI and SVV were predictive of outcome because their c-statistics were high (0.83 and 0.92, respectively). This is misleading because the 95% confidence intervals [CIs] were wide because of the sample size (0.58–1.00 and 0.73–1.00, respectively). The aortic blood velocity variation c-statistic was 0.67 (95% CI, 0.32–1.00). Had the authors compared the c-statistics among these three predictors, they would all be statistically similar as the confidence intervals all have substantial overlap. Thus, one may not conclude that any of the three predictors had better predictive (discriminatory) ability than the other two, and all three predictors’ c-statistics may only be in the 0.7 to 0.75 range, indicating modest discrimination as all three CIs overlapped those values. The conclusions from their positive and negative predictive value analysis in Table 3 are also similarly dubious because of the sample size. Furthermore, only one predictor (VCCI) had a borderline significant difference in medians by response (P = 0.04), with the others being nonsignificant, as seen in Table 1. A limitation not acknowledged is that the study results are all unadjusted for confounding. If we assume a regular r-r interval, for any specific preload condition, dynamic indices, such as SVV, will vary with a patient’s tidal volume (Vt) (2) and respiratory rate (RR) and any variability in RR. This means Vt and RR are confounders and need to be adjusted for via regression to determine the true relationship between SVV and hemodynamic response. Indeed, Kim and Pinsky (3) have shown that SVV varies with RR and that one must hold Vt constant if changes in SVV are to assess preload responsiveness. Furthermore, the ability of either PPV or SVV to predict preload responsiveness varies with Vt (4). To this point, De Backer et al. (5) have shown that a minimum of 8 mL/kg is required for positive-pressure ventilation to reliably predict volume responsiveness during mechanical ventilation. Although it is true that “dynamic fluctuations in thoracic pressure occur” regardless of ventilation mode, it is highly unlikely that these spontaneously breathing patients (ostensibly with near-normal tidal breaths yet all being different) had values near 8 to 10 mL/kg. Third, the authors decry the passive-leg-raise maneuver, although it reliably predicts volume responsiveness with a pooled area under the curve of 0.95 (6) and is independent of either ventilatory mode or cardiac rhythm. The authors contend that limitations of the passive leg raise include patient movement, the requirement of a device, and that it is affected by raised intra-abdominal pressure. Indeed, although SVV does not require patient movement, it most certainly requires a device and is affected by intra-abdominal pressure, as are all dynamic indices. (7) Lanspa et al. attempted to answer an important research question, and it is unfortunate that they could not obtain an adequately powered sample. Pilot studies such as this do not contribute in a meaningful way to our understanding of the hemodynamic response to fluid therapy and may even be harmful if their findings are not interpreted cautiously. This study is remindful of Ioannidis’ (8) first corollary, which states that “the smaller the studies conducted in a scientific field, the less likely the research findings are to be true.” No financial support was offered or obtained for the writing of this editorial letter. In addition, no NIH, Wellcome Trust, or Howard Hughes funding was used or obtained. J.A.B. and W.T.M. are on the speaker’s bureau for Edwards LifeSciences. B.H.N. has no potential conflicts of interest. Joshua A. Bloomstone Department Anesthesiology and Perioperative Medicine, Banner Thunderbird Medical Center Glendale, Arizona Brian H. Nathanson OptiStatim, LLC, Longmeadow William T. McGee Baystate Critical Care Medicine Springfield, Massachusetts
Search from 250 M+ research papersSearch from 250 M+ research papers
Open Access
Aug 4, 2023
Marcus Karl-Heinz Auth + 9
