Using generalized additive models to decompose time series and waveforms, and dissect heart-lung interaction physiology

Common physiological time series and waveforms are composed of repeating cardiac and respiratory cycles. Often, the cardiac effect is the primary interest, but for, e.g., fluid responsiveness prediction, the respiratory effect on arterial blood pressure also convey important information. In either case, it is relevant to disentangle the two effects. Generalized additive models (GAMs) allow estimating the effect of predictors as nonlinear, smooth functions. These smooth functions can represent the cardiac and respiratory cycles' effects on a physiological signal. We demonstrate how GAMs allow a decomposition of physiological signals from mechanically ventilated subjects into separate effects of the cardiac and respiratory cycles. Two examples are presented. The first is a model of the respiratory variation in pulse pressure. The second demonstrates how a central venous pressure waveform can be decomposed into a cardiac effect, a respiratory effect and the interaction between the two cycles. Generalized additive models provide an intuitive and flexible approach to modelling the repeating, smooth, patterns common in medical monitoring data.

Existing fluid responsiveness studies using the mini-fluid challenge may be misleading: Methodological considerations and simulations

BACKGROUND

The mini-fluid challenge (MFC) is a clinical concept of predicting fluid responsiveness by rapidly infusing a small amount of intravenous fluids, typically 100 ml, and systematically assessing its haemodynamic effect. The MFC method is meant to predict if a patient will respond to a subsequent, larger fluid challenge, typically another 400 ml, with a significant increase in stroke volume.

METHODS

We critically evaluated the general methodology of MFC studies, with statistical considerations, secondary analysis of an existing study and simulations.

RESULTS

Secondary analysis of an existing study showed that the MFC could predict the total fluid response (MFC + 400 ml) with an area under the receiver operator characteristic curve (AUROC) of 0.92, but that the prediction was worse than random for the response to the remaining 400 ml (AUROC = 0.33). In a null simulation with no response to both the MFC and the subsequent fluid challenge, the commonly used analysis could predict fluid responsiveness with an AUROC of 0.73.

CONCLUSION

Many existing MFC studies are likely overestimating the classification accuracy of the MFC. This should be considered before adopting the MFC into clinical practice. A better study design includes a second, independent measurement of stroke volume after the MFC. This measurement serves as reference for the response to the subsequent fluid challenge.

Changes in arterial blood pressure characteristics following an extrasystolic beat or a fast 50 ml fluid challenge do not predict fluid responsiveness during cardiac surgery

Prediction of fluid responsiveness is essential in perioperative goal directed therapy, but dynamic tests of fluid responsiveness are not applicable during open-chest surgery. We hypothesised that two methods could predict fluid responsiveness during cardiac surgery based on their ability to alter preload and thereby induce changes in arterial blood pressure characteristics: (1) the change caused by extrasystolic beats and (2) the change caused by a fast infusion of 50 ml crystalloid (micro-fluid challenge). Arterial blood pressure and electrocardiogram waveforms were collected during surgical preparation of the left internal mammary artery in patients undergoing coronary artery bypass surgery. Patients received a fluid challenge (5 ml/kg ideal body weight). The first 50 ml were infused in 10 s and comprised the micro-fluid challenge. Predictor variables were defined as post-ectopic beat changes (compared with sinus beats preceding ectopy) in arterial blood pressure characteristics, such as pulse pressure and systolic pressure, or micro-fluid challenge induced changes in the same blood pressure characteristics. Patients were considered fluid responsive if stroke volume index increased by 15% or more after the full fluid challenge. Diagnostic accuracy was calculated by the area under the receiver operating characteristics curve (AUC). Fifty-six patients were included for statistical analysis. Thirty-one had extrasystoles. The maximal AUC was found for the extrasystolic change in pulse pressure and was 0.70 (CI [0.35 to 1.00]). The micro-fluid challenge method generally produced lower AUC point estimates. Extrasystoles did not predict fluid responsiveness with convincing accuracy in patients undergoing cardiac surgery and changes in arterial waveform indices following a micro-fluid challenge could not predict fluid responsiveness. Given a low number of fluid responders and inherently reduced statistical power, our data does not support firm conclusions about the utility of the extrasystolic method. CLINICAL TRIAL REGISTRATION: Unique identifier: NCT02903316. https://clinicaltrials.gov/ct2/show/NCT02903316?cond=NCT02903316&rank=1 .

Current use of inotropes in circulatory shock

Background

Treatment decisions on critically ill patients with circulatory shock lack consensus. In an international survey, we aimed to evaluate the indications, current practice, and therapeutic goals of inotrope therapy in the treatment of patients with circulatory shock.

Methods

From November 2016 to April 2017, an anonymous web-based survey on the use of cardiovascular drugs was accessible to members of the European Society of Intensive Care Medicine (ESICM). A total of 14 questions focused on the profile of respondents, the triggering factors, first-line choice, dosing, timing, targets, additional treatment strategy, and suggested effect of inotropes. In addition, a group of 42 international ESICM experts was asked to formulate recommendations for the use of inotropes based on 11 questions.

Results

A total of 839 physicians from 82 countries responded. Dobutamine was the first-line inotrope in critically ill patients with acute heart failure for 84% of respondents. Two-thirds of respondents (66%) stated to use inotropes when there were persistent clinical signs of hypoperfusion or persistent hyperlactatemia despite a supposed adequate use of fluids and vasopressors, with (44%) or without (22%) the context of low left ventricular ejection fraction. Nearly half (44%) of respondents stated an adequate cardiac output as target for inotropic treatment. The experts agreed on 11 strong recommendations, all of which were based on excellent (> 90%) or good (81–90%) agreement. Recommendations include the indications for inotropes (septic and cardiogenic shock), the choice of drugs (dobutamine, not dopamine), the triggers (low cardiac output and clinical signs of hypoperfusion) and targets (adequate cardiac output) and stopping criteria (adverse effects and clinical improvement).

Conclusion

Inotrope use in critically ill patients is quite heterogeneous as self-reported by individual caregivers. Eleven strong recommendations on the indications, choice, triggers and targets for the use of inotropes are given by international experts. Future studies should focus on consistent indications for inotrope use and implementation into a guideline for circulatory shock that encompasses individualized targets and outcomes.

The response of a standardized fluid challenge during cardiac surgery on cerebral oxygen saturation measured with near-infrared spectroscopy

Near infrared spectroscopy (NIRS) has been used to evaluate regional cerebral tissue oxygen saturation (ScO₂) during the last decades. Perioperative management algorithms advocate to maintain ScO₂, by maintaining or increasing cardiac output (CO), e.g. with fluid infusion. We hypothesized that ScO₂ would increase in responders to a standardized fluid challenge (FC) and that the relative changes in CO and ScO₂ would correlate. This study is a retrospective substudy of the FLuid Responsiveness Prediction Using Extra Systoles (FLEX) trial. In the FLEX trial, patients were administered two standardized FCs (5 mL/kg ideal body weight each) during cardiac surgery. NIRS monitoring was used during the intraoperative period and CO was monitored continuously. Patients were considered responders if stroke volume increased more than 10% following FC. Datasets from 29 non-responders and 27 responders to FC were available for analysis. Relative changes of ScO₂ did not change significantly in non-responders (mean difference - 0.3% ± 2.3%, p = 0.534) or in fluid responders (mean difference 1.6% ± 4.6%, p = 0.088). Relative changes in CO and ScO₂ correlated significantly, p = 0.027. Increasing CO by fluid did not change cerebral oxygenation. Despite this, relative changes in CO correlated to relative changes in ScO₂. However, the clinical impact of the present observations is unclear, and the results must be interpreted with caution.Trial registration:http://ClinicalTrial.gov identifier for main study (FLuid Responsiveness Prediction Using Extra Systoles-FLEX): NCT03002129.

Automated echocardiography for measuring and tracking cardiac output after cardiac surgery: a validation study

Echocardiographic measurement of cardiac output with automated software analyses of spectral curves in the left ventricular outflow tract has been introduced. This study aimed to assess the precision and accuracy of cardiac output measurements as well as the ability to track cardiac output changes over time comparing the automated echocardiographic method with the continuous pulmonary artery thermodilution cardiac output technique and the manual echocardiographic method in cardiac surgery patients. Cardiac output was measured simultaneously with all three methods in 50 patients on the morning after cardiac surgery. A second comparison was performed 90-180 min later. Precisions for each method were measured. Bias and limits of agreement (LoA) between methods were assessed and concordance- and polar plots were used for evaluating trending of cardiac output. When comparing the automated echocardiographic method with the thermodilution technique, the mean bias was 0.72 L/min with LoA - 1.89; 3.33 L/min corresponding to a percentage error of 46%. The concordance rate was 47%. The mean bias between the automated- and the manual echocardiographic methods was - 0.06 L/min (95% LoA - 2.33; 2.21 L/min, percentage error 42%). The concordance rate was 79%. The automated echocardiographic method did not meet the criteria for interchangeability with the thermodilution technique or the manual echocardiographic method. Trending ability was poor when compared to the continuous thermodilution technique, but moderate when compared to the manual echocardiographic method.Trial registry number: NCT03372863. Retrospectively registered December 14th 2017.

What the anaesthesiologist needs to know about heart-lung interactions

The impact of positive pressure ventilation extends the effect on lungs and gas exchange because the altered intra-thoracic pressure conditions influence determinants of cardiovascular function. These mechanisms are called heart-lung interactions, which conceptually can be divided into two components (1) The effect of positive airway pressure on the cardiovascular system, which may be more or less pronounced under various pathologic cardiac conditions, and (2) The effect of cyclic airway pressure swing on the cardiovascular system, which can be useful in the interpretation of the individual patient's current haemodynamic state. It is imperative for the anaesthesiologist to understand the fundamental mechanisms of heart-lung interactions, as they are a foundation for the understanding of optimal, personalised cardiovascular treatment of patients undergoing surgery in general anaesthesia. The aim of this review is thus to describe what the anaesthesiologist needs to know about heart-lung interactions.

Journal of clinical monitoring and computing end of year summary 2018: hemodynamic monitoring and management

Hemodynamic management is a mainstay of patient care in the operating room and intensive care unit (ICU). In order to optimize patient treatment, researchers investigate monitoring technologies, cardiovascular (patho-) physiology, and hemodynamic treatment strategies. The Journal of Clinical Monitoring and Computing (JCMC) is a well-established and recognized platform for publishing research in this field. In this review, we highlight recent advancements and summarize selected papers published in the JCMC in 2018 related to hemodynamic monitoring and management.

Current use of vasopressors in septic shock

Background

Vasopressors are commonly applied to restore and maintain blood pressure in patients with sepsis. We aimed to evaluate the current practice and therapeutic goals regarding vasopressor use in septic shock as a basis for future studies and to provide some recommendations on their use.

Methods

From November 2016 to April 2017, an anonymous web-based survey on the use of vasoactive drugs was accessible to members of the European Society of Intensive Care Medicine (ESICM). A total of 17 questions focused on the profile of respondents, triggering factors, first choice agent, dosing, timing, targets, additional treatments, and effects of vasopressors. We investigated whether the answers complied with current guidelines. In addition, a group of 34 international ESICM experts was asked to formulate recommendations for the use of vasopressors based on 6 questions with sub-questions (total 14).

Results

A total of 839 physicians from 82 countries (65% main specialty/activity intensive care) responded. The main trigger for vasopressor use was an insufficient mean arterial pressure (MAP) response to initial fluid resuscitation (83%). The first-line vasopressor was norepinephrine (97%), targeting predominantly a MAP > 60–65 mmHg (70%), with higher targets in patients with chronic arterial hypertension (79%). The experts agreed on 10 recommendations, 9 of which were based on unanimous or strong (≥ 80%) agreement. They recommended not to delay vasopressor treatment until fluid resuscitation is completed but rather to start with norepinephrine early to achieve a target MAP of ≥ 65 mmHg.

Conclusion

Reported vasopressor use in septic shock is compliant with contemporary guidelines. Future studies should focus on individualized treatment targets including earlier use of vasopressors.

Using extra systoles and the micro-fluid challenge to predict fluid responsiveness during cardiac surgery

Fluid responsiveness prediction is difficult during cardiac surgery. The micro-fluid challenge (micro-FC; rapid central infusion of 50 ml) and the extrasystolic method utilising post-extrasystolic preload increases may predict fluid responsiveness. Two study windows during coronary artery bypass graft surgery were defined, 1: After anaesthesia induction until surgical incision, 2: Left internal mammarian artery surgical preparation period. Each window consisted of 10-15 min observation for extrasystoles before a micro-FC was performed, after which a traditional fluid challenge (FC) was performed (5 ml/kg). Extrasystolic and micro-FC induced changes in hemodynamic variables were derived as predictors of fluid responsiveness defined as stroke volume increasing > 10% following FC. 61 patients were studied. Post-ectopic changes in pulse pressure (PP) predicted fluid responsiveness with receiver operating characteristic area (AUC) of 0.69 [CI 0.40;0.97] in the first study window and 0.64 [0.44;0.86] in the second window. Other post-ectopic predictors such as pre-ejection period (PEP) and systolic blood pressure (SBP) had similar or lower AUCs. Heart rate was 52.9 (SD ±8.4) min^- 1 and 53.6 (± 8.8) min^- 1 in the two study windows. Micro-FC induced changes in PEP had AUC of 0.74 [0.57;0.90] in the first window and 0.60 [0.40;0.76] in the second window. Correcting micro-FC induced changes in PEP for the micro-FC induced changes in heart rate had AUCs of 0.84 [0.70;0.97] in the first window and 0.63 [0.47;0.79] in the second window. The investigated methods revealed insufficient validity during cardiac surgery. RR interval corrected changes during a micro-FC should be investigated further. Trial registration Clinicaltrials.gov: NCT03002129.

Gastrointestinal transit time and heart rate variability in patients with mild acquired brain injury

Background

Constipation is suspected to occur frequently after acquired brain injury (ABI). In patients with ABI, heart rate variability (HRV) is reduced suggesting autonomic dysfunction. Autonomic dysfunction may be associated with prolonged gastrointestinal transit time (GITT). The primary aim of this study was to investigate if GITT is prolonged in patients with ABI. Secondarily, HRV and its correlation with GITT was investigated.

Methods

We included 25 patients with ABI (18 men, median age: 61.3 years, range [30.7-74.5]). GITT was assessed using radio-opaque markers and HRV was calculated from 24-hour electrocardiograms. Medical records were reviewed for important covariates, including primary diagnosis, time since injury, functional independence measure, and use of medication. The GITT assessed in patients was compared to a control group of 25 healthy subjects (18 men, median age: 61.5 years, range [34.0-70.9]).

Results

In ABI patients, the mean GITT was significantly longer than in healthy controls (2.68 days, 95% CI [2.16-3.19] versus (1.92 days, 95% CI [1.62-2.22], p = 0.011)). No correlation was found between HRV and GITT.

Conclusion

Patients with mild to moderate ABI have prolonged GITT unrelated to the HRV.

Should dynamic parameters for prediction of fluid responsiveness be indexed to the tidal volume?

BACKGROUND

The respiratory variation in the pre-ejection period (Delta PEP) has been used to predict fluid responsiveness in mechanically ventilated patients. Recently, we modified this parameter (PEPV) and showed that it was a reliable predictor for post-cardiac surgery, mainly paced, patients when moderately low tidal volumes were used. One of the modifications involved tidal volume indexation, which had not been proposed before for dynamic parameters. The aim of the present animal study was to investigate whether indexation to tidal volume should be part of a new definition of dynamic parameters such as the case for our newly proposed PEPV.

METHODS

Eight prone, anesthetized piglets (23-27 kg) were subjected to a sequence of 25% hypovolemia, normovolemia, and 25% and 50% hypervolemia. At each volemic level, tidal volumes were varied in three steps: 6, 9, and 12 ml/kg. PEP variations (ms) and pulse-pressure variation (PPV) were measured during the three tidal volume steps at each volemic level.

RESULTS

PEP variations increased significantly with increasing tidal volume at all volemic levels but 50% hypervolemia and were proportionally related to the tidal volume at normovolemia. PPV increased significantly with increasing tidal volume at all volemic levels and was roughly proportional to the tidal volume at all volemic levels but hypovolemia.

CONCLUSION

Our study indicates that dynamic parameters are improved by indexing to tidal volume.

Automated pre-ejection period variation predicts fluid responsiveness in low tidal volume ventilated pigs

INTRODUCTION

The respiratory variation in the pre-ejection period (Delta PEP) has been used to predict fluid responsiveness in mechanically ventilated patients. Recently, we automated this parameter and indexed it to tidal volume (PEPV) and showed that it was a reliable predictor for post-cardiac surgery, mainly paced, patients ventilated with low tidal volumes. The aims of the present animal study were to investigate PEPV's ability to predict fluid responsiveness under different fluid loading conditions and natural heart rates during low tidal volume ventilation (6 ml/kg) and to compare the performance of PEPV with other markers of fluid responsiveness.

METHODS

Eight prone, anesthetized piglets (23-27 kg) ventilated with tidal volumes of 6 ml/kg were subjected to a sequence of 25% hypovolemia, normovolemia, and 25% and 50% hypervolemia. PEPV, Delta PEP, pulse pressure variation (PPV), central venous pressure (CVP), and pulmonary artery occlusion pressure (PAOP) were measured before each volume expansion.

RESULTS

Sensitivity was 89% and specificity was 93% for PEPV, 78% and 93% for Delta PEP, 89% and 100% for PPV, 78% and 93% for CVP, and 89% and 87% for PAOP.

CONCLUSION

PEPV predicts fluid responsiveness in low tidal volume ventilated piglets.

Automated pre-ejection period variation indexed to tidal volume predicts fluid responsiveness after cardiac surgery

BACKGROUND

Reliable continuous monitoring of fluid responsiveness is an unsolved issue in patients ventilated with low tidal volume. We hypothesised that variations in the pre-ejection period (PEP) defined as the time interval between electrocardiogram (ECG) R-wave and onset of systolic upstroke in arterial blood pressure could reliably predict fluid responsiveness in patients ventilated with moderately low tidal volume. Furthermore, we hypothesised that indexing dynamic parameters to tidal volume would improve their prediction. The aim was to refine and automate a previously suggested algorithm for PEP variation (DeltaPEP) and to test this new parameter indexed to tidal volume (PEPV), as a marker of fluid responsiveness along with central venous pressure (CVP), pulse pressure variation (PPV) and DeltaPEP. Additionally, the aim was to evaluate the concept of indexing dynamic parameters to tidal volume.

METHODS

Arterial pressure, CVP, ECG and cardiac index (CI) were acquired from 23 mechanically ventilated post-cardiac surgery patients scheduled for volume expansion. PEPV, PPV and DeltaPEP were extracted.

RESULTS

Using responder/non-responder classification (response=change in CI>+15%), sensitivity and specificity were 100% and 83%, respectively, for PEPV, 94% and 83% for DeltaPEP, and 94% and 83% for PPV. CVP offered no relevant information. Tidal volume indexing improved sensitivity for DeltaPEP to 100%.

CONCLUSION

In this study in post-cardiac surgery patients, a refined parameter, PEPV, predicted fluid responsiveness better than PPV and DeltaPEP. Our results suggest that dynamic parameters using variations in PEP should be indexed to tidal volume.

Using an expiratory resistor, arterial pulse pressure variations predict fluid responsiveness during spontaneous breathing: an experimental porcine study

Introduction

Fluid responsiveness prediction is difficult in spontaneously breathing patients. Because the swings in intrathoracic pressure are minor during spontaneous breathing, dynamic parameters like pulse pressure variation (PPV) and systolic pressure variation (SPV) are usually small. We hypothesized that during spontaneous breathing, inspiratory and/or expiratory resistors could induce high arterial pressure variations at hypovolemia and low variations at normovolemia and hypervolemia. Furthermore, we hypothesized that SPV and PPV could predict fluid responsiveness under these conditions.

Methods

Eight prone, anesthetized and spontaneously breathing pigs (20 to 25 kg) were subjected to a sequence of 30% hypovolemia, normovolemia, and 20% and 40% hypervolemia. At each volemic level, the pigs breathed in a randomized order either through an inspiratory and/or an expiratory threshold resistor (7.5 cmH₂O) or only through the tracheal tube without any resistor. Hemodynamic and respiratory variables were measured during the breathing modes. Fluid responsiveness was defined as a 15% increase in stroke volume (ΔSV) following fluid loading.

Results

Stroke volume was significantly lower at hypovolemia compared with normovolemia, but no differences were found between normovolemia and 20% or 40% hypervolemia. Compared with breathing through no resistor, SPV was magnified by all resistors at hypovolemia whereas there were no changes at normovolemia and hypervolemia. PPV was magnified by the inspiratory resistor and the combined inspiratory and expiratory resistor. Regression analysis of SPV or PPV versus ΔSV showed the highest R² (0.83 for SPV and 0.52 for PPV) when the expiratory resistor was applied. The corresponding sensitivity and specificity for prediction of fluid responsiveness were 100% and 100%, respectively, for SPV and 100% and 81%, respectively, for PPV.

Conclusions

Inspiratory and/or expiratory threshold resistors magnified SPV and PPV in spontaneously breathing pigs during hypovolemia. Using the expiratory resistor SPV and PPV predicted fluid responsiveness with good sensitivity and specificity.