Post-Induction Low Cardiac Output Syndrome in Coronary Artery Bypass Graft Surgery

BACKGROUND AND RATIONALE Peri-operative hemodynamic management is built almost entirely around arterial pressure. International consensus recommends that mean arterial pressure be kept at or above 60-65 mmHg in patients at risk, notes that organ injury depends on both the depth and the duration of hypotension, and identifies the area beneath a pressure threshold as the metric most closely associated with injury. That framework arose from a large observational literature linking intra-operative hypotension to acute kidney injury, myocardial injury and death. Arterial pressure, however, is not perfusion. Perfusion is determined by flow, and pressure and flow can separate, particularly where vascular tone is changing rapidly under the influence of anesthetic agents, surgical stimulation and vasoactive drugs. A recent prospective cohort in isolated coronary artery bypass grafting reported a mean of 86.4 minutes of low cardiac index across the whole operation, of which 69% occurred while the patient was normotensive, with a correlation between cardiac index and mean arterial pressure of only 0.33. Low cardiac output after cardiac surgery has itself been defined in at least 262 different ways, under which reported adult incidence ranges from 1.5% to 91%. The interval that follows induction of anesthesia has a particular standing among these periods. It precedes surgical stimulation, it is short, and hemodynamic conditions within it are determined by anesthetic management alone. Post-induction hypotension has been characterized in detail and is conventionally measured over the first 20 minutes after induction. What happens to flow during this interval has not been described. This study sets out to characterize it. STUDY DESIGN AND SETTING This is a prospective, observational, two-centre cohort study conducted at two tertiary cardiovascular surgery centre. The study is observational in the strict sense: no aspect of anesthetic technique, monitoring, fluid or vasoactive management, or surgical conduct is altered for the purposes of the study, and the investigators take no part in clinical decisions. The study is reported in accordance with the STROBE statement. Consecutive adults scheduled for isolated coronary artery bypass grafting with cardiopulmonary bypass are screened and enrolled. Patients are enrolled before induction of anesthesia so that baseline measurement can begin while they are awake. HAEMODYNAMIC MONITORING AND DATA ACQUISITION An arterial catheter is sited before induction while the patient is awake, in keeping with routine practice at both centre for this operation. Cardiac output is derived from the arterial waveform by pulse contour analysis via pressure recording analytical method and indexed to body surface area. Cardiac index is recorded continuously at 30-second intervals from before induction until the end of the operation. Recorded at the same resolution are mean arterial pressure, systolic and diastolic pressure, heart rate, stroke volume and stroke volume index. Body surface area is computed once for each patient from height and weight. The time of induction is recorded prospectively on the case report form as the moment at which the intravenous induction agent is given, and is cross-checked against the monitor clock. Where the two disagree, the discrepancy is documented and resolved before analysis by a prespecified rule. THE ANALYSIS WINDOW The analysis window is the 30 minutes that begin at induction of anesthesia. This length was chosen for two reasons. It extends beyond the 20 minutes over which post-induction hypotension is conventionally measured, so that the study can determine whether the conventional window is adequate for flow. It also ends before surgical stimulation is established in most patients, so that the exposure measured remains attributable to anesthetic management rather than to surgery. The period of cardiopulmonary bypass is excluded from all analyses. Flow during bypass is generated by the pump rather than by the heart, so cardiac index derived from the arterial waveform has no physiological meaning in that interval. The bypass period is identified from the perfusion record and confirmed against the arterial waveform. PRIMARY OUTCOME The primary outcome has two components. Both are defined over the 30-minute window and both are referenced to a cardiac index threshold of 2.0 L/min/m2. The first component is time below threshold: the total duration for which cardiac index lies below 2.0 L/min/m2, expressed in minutes. The second component is cumulative low-output burden: the integral over time of the difference between the threshold and the measured cardiac index, taken only where that difference is positive. Formally, for threshold T and cardiac index CI(t), burden is the integral of the positive part of T minus CI(t) over the window. The two components answer different questions and neither is sufficient alone. Time below threshold states how long the exposure lasted but is indifferent to its depth: one minute at 1.9 L/min/m2 and one minute at 1.0 L/min/m2 count equally. Cumulative burden states how deep the exposure was but is indifferent to its shape: a brief profound fall and a prolonged shallow one may yield the same value. Reported together they define the exposure. The unit of burden reduces under dimensional analysis. Cardiac index is expressed in L/min/m2 and time in minutes, so their product is L/m2, a volume of undelivered flow indexed to body surface area. Multiplied by the patient's body surface area it becomes an absolute flow deficit expressed in litres, which is the quantity of blood flow that the circulation failed to deliver during the window. Both components are computed between consecutive measurement pairs by the trapezoidal rule with linear interpolation, and threshold crossings are located exactly rather than rounded to the sampling grid. Intervals between consecutive measurements longer than the sampling period are not counted as observed time. SECONDARY OUTCOMES Time below threshold and cumulative burden are additionally computed at cardiac index thresholds of 2.2 and 2.5 L/min/m2 over the same window, so that the findings do not depend on a single threshold. The incidence of PiLCOs is defined as the proportion of patients with at least one measurement below 2.0 L/min/m2 within the window. Two measures of depth are reported. Time-weighted average depth is cumulative burden divided by the length of the window and describes the mean shortfall spread across the whole interval. Depth during exposure is the mean shortfall computed only over the time spent below threshold and describes how far below the threshold the circulation actually ran while the event was occurring. The temporal distribution of exposure is described by the share of cumulative burden accruing within each of the three consecutive ten-minute segments of the window, and by the time from induction to the first measurement below threshold. FLOW AND PRESSURE A prespecified secondary analysis addresses whether pressure-based monitoring is able to detect the event. The proportion of sub-threshold exposure occurring while the patient is normotensive is quantified, a normotensive episode being defined as a cardiac index below 2.0 L/min/m2 recorded at a mean arterial pressure of 65 mmHg or above. The correlation between cardiac index and mean arterial pressure within the window is examined over the same period, and the agreement between the two in classifying a patient as exposed or unexposed is described. EXPLORATORY ANALYSES The association between the presence and the magnitude of PiLCOs and the postoperative course is examined for acute kidney injury, serum lactate, duration of mechanical ventilation, intensive care unit and hospital length of stay, and mortality. Patient-level and procedural characteristics, among them age, body-mass index, left ventricular ejection fraction, comorbidity burden, induction agent dose and cardiopulmonary bypass duration, are examined as correlates of exposure. All outcome analyses are exploratory and hypothesis-generating. The number of covariates admitted to any model is constrained by the rule of at least ten events per variable, and no adjustment is made for multiple comparisons. Associations found here are intended to inform the design of subsequent studies rather than to establish causation. STATISTICAL ANALYSIS Continuous variables are summarized as mean and standard deviation where the distribution is symmetrical and as median and interquartile range where it is skewed. Distributional shape is assessed by the Shapiro-Wilk test together with bias-corrected coefficients of skewness and excess kurtosis, and uni-modality is examined using the bi-modality coefficient derived from those two moments. Confidence intervals for proportions are constructed by the Wilson score method and for means and medians by percentile bootstrap. Groups are compared using the Mann-Whitney U test for continuous variables and Fisher's exact test for categorical variables. Effect size is reported as the Hodges-Lehmann estimate of location shift with its distribution-free confidence interval and as Cliff's delta for overlap. Associations between continuous variables are examined by Spearman rank correlation. A two-sided p value below 0.05 is taken as the threshold for statistical significance. No value-based artifact filter is applied to the cardiac index record, and no measurement is excluded other than those flagged as invalid by the monitor itself. Removing low readings as artifact would cause systematic under-ascertainment of the very event under study. Recording coverage within the window is quantified for each patient across the three ten-minute segments. A patient whose weighted coverage falls below a prespecified minimum is not included in the primary analysis, because the exposure cannot be estimated reliably from a sparse record. SAMPLE SIZE Sample size is based on the precision of estimation rather than on the power of a hypothesis test, because the study estimates the frequency and the magnitude of an event that has not previously been characterized rather than testing a prespecified superiority hypothesis. An internal pilot of 30 patients, comprising the first 15 consecutive patients enrolled at each centre, was used to estimate variability. In that pilot the mean time below threshold was 6.59 minutes with a standard deviation of 8.14 minutes, and the incidence of PiLCOs was 70.0%. Two precision targets were set. For mean time below threshold the half-width of the 95% confidence interval is to be no greater than 1.7 minutes, which requires 88 patients. For the incidence of PiLCOs the half-width is to be no greater than 9.5 percentage points, which requires 90 patients. The larger of the two figures was carried forward and rounded to 90, allocated as 45 patients per centre. Allowing 15% for withdrawal, protocol deviation and unusable recordings gives a planned enrollment of 103 patients. ETHICS The study is approved by the institutional review board at each participating centre and written informed consent is obtained from every patient before enrollment. Because the study imposes no intervention and alters no aspect of care, the risk to participants is confined to the handling of clinical data, which is stored in de-identified form.