Selection Bias in the Hypotension Prediction Index: Comment.

We read with interest the article by Enevoldsen and Vistisen about the predictive model of intraoperative hypotension.1 Regarding the criticism the authors made about what they named “gray-zone outcome,” we would like to provide some comments.The Hypotension Prediction Index is a new parameter provided by the Hemosphere platform from Edwards Lifesciences (USA). The platform shows also other parameters: systolic, diastolic, and mean arterial pressure; the markers of fluid responsiveness (stroke volume variation and pulse pressure variation), and their ratio, dynamic elastance (pulse pressure variation/stroke volume variation ); and the increase of pulse pressure over time (dP/dt, i.e., the increase of blood pressure in one second, the slope). The latter parameter (dP/dt) is well correlated with the left ventricle elastance and reflects cardiac contractility,.2–7The alarm level of the index has been set by the manufacturer at 85/100, but we believe the anesthetist should check the cardiovascular system when it is higher than 50/100 because we consider it is a marker for the level of cardiovascular compensation in response to a specific hemodynamic impairment. The higher the index, the lower the compensatory reserve, and the shorter the time before hypotension occurs. When the index trend begins to rise, the other parameters can help to understand the potential cause of the ongoing hemodynamic alteration: stroke volume variation and pulse pressure variation regarding fluid dependency, dynamic elastance regarding the pressure response to vasoconstrictor drug or fluid administration, and dP/dt regarding contractility failure, if the patient is not preload-dependent.5From our anecdotal experience with its use in major surgery, we have noticed that when the index trend increases close to 50/100, the first hemodynamic parameter that starts to decrease is the diastolic arterial pressure. In our ongoing analysis (preliminary results of 990 measurements in 10 patients submitted to major surgery), we plotted the index and the diastolic arterial pressure when the mean arterial pressure (MAP) is 65 to 75 mmHg. As shown in figure 1, when diastolic arterial pressure is lower than 60 mmHg, the index is generally already higher than 50/100 (P < 0.0001).Figure 2, A and B, shows representative data from two patients with periods of the monitoring time during which the diastolic arterial pressure is lower than 60 mmHg and the index is greater than 85/100, even with a normal trend of the MAP. Figure 2A shows a long-lasting period of diastolic hypotension that was untreated since the MAP and other parameters were in the normal range. The same situation but with a different outcome is shown in figure 2B, where a bolus of vasoconstrictor (gray triangle) restored the impairment before the predictive index could change and MAP could drop. After correction of what was likely a vasodilation event, the hemodynamics profile normalized.In addition, our experience derives from intraoperative employment of this type of monitoring during major surgery, and it is known that diastolic arterial pressure is generally also a marker of the vasodilation caused by general anesthesia.We consider that neglecting reductions in diastolic pressure may be harmful for the patient’s outcome, given the mostly exclusive dependency on diastolic flow of heart, kidney, liver, and brain. On the other hand, we are aware that the risk of overtreatment is real.Enevoldsen and Vistisen reported that a bias in the chosen range of blood pressure could overestimate the performance of the predictive model.1 The algorithm of the index considered the lowest level of normal MAP 65 mmHg and a “no-risk zone” of a hypotensive event if the MAP is greater than 75 mmHg. This assumption could imply that a patient with a MAP less than 75 mmHg could experience a hypotensive event anyway. Nevertheless, Hatib et al. recommended strict attention when the MAP is between 65 and 75 mmHg.6 In our opinion, such a “gray zone” could be checked by the diastolic arterial pressure trend, and if this persists at a low level, correcting diastolic arterial pressure may be appropriate to prevent a subsequent drop in MAP.To our knowledge, diastolic arterial pressure as a trigger to the Hypotension Prediction Index to increase has not been investigated and is a topic worthy of specific analysis.Finally, is this just a matter of computation of MAP, since the formula 2/3 of mean arterial pressure is due to the diastolic arterial pressure? We are not able to answer this currently as we do not know the 23 features of the proprietary model of the predictive index.Anyway, mind the diastolic arterial pressure! We could be proactive even earlier than the Hypotension Prediction Index.Dr. Giustiniano is an Edwards Lifesciences (Irwin, California) Proctor for Italy/Europe. This program supports operators using the Hemosphere platform for hemodynamic monitoring in the operating room or intensive care unit (the Proctor receives a fee for each event). This activity does not involve research support. The other author declares no competing interests. The authors declare that this paper has been written and submitted without any fund and/or any grant and/or any extramural support received. Moreover, the authors did not receive any salary from Edwards Lifesciences or from any other institutional body.