Hemodynamic Changes after Protamine Administration: Association with Mortality after Coronary Artery Bypass Surgery

With institutional review board approval (Duke University Medical Center, Durham, North Carolina), we retrospectively studied 6,921 consecutive patients undergoing primary elective CABG surgery at an institution between 1993 and 2000. Patients were excluded if the surgery was emergent, if the surgery involved additional procedures or circulatory arrest, or if protamine administration was not documented. Patient characteristics and other variables were obtained from the prospectively collected Duke Cardiovascular Database,14a quality assurance database from contemporaneous medical records, custom data sheets, and records of laboratory results. Quality assurance involves random chart review for data confirmation and assessment of data completeness.

Risk factors for in-hospital mortality after CABG surgery, identified by Hannan et al.  15in the New York State population, were tested as a marker of preoperative mortality risk. These include age, sex, severity of presenting condition, preoperative intraaortic balloon counterpulsation, left main coronary artery disease, history of unstable angina, congestive cardiac failure, left ventricular ejection fraction, and comorbidities such as chronic obstructive pulmonary disease, diabetes, morbid obesity, and dialysis dependency. Cardiopulmonary bypass (CPB) duration was included for analysis as a marker of procedure complexity and risk because it is directly correlated with factors such as number of coronary grafts performed and is significantly associated with in-hospital mortality.16In addition, postoperative need for intraaortic balloon counterpulsation was recorded.

Protamine and hemodynamic data were obtained by querying an automated anesthesia record–keeping system (Arkive; Arkive Inc., San Diego, CA). Drug administration is manually entered when it occurs17; hemodynamic data are automatically downloaded by this system at 1-min intervals.

To develop variables representing the range of hemodynamic perturbation, minute-by-minute accumulated “area under the line” for systolic arterial blood pressures less than 100 mmHg during the first 30 min after protamine administration was used as an index of systemic hypotension (T^{sBP < 100}); for all patients with a pulmonary artery catheter, an equivalent “over the line” metric was determined for pulmonary artery systolic pressures greater than 30 mmHg (T^{sPAP > 30}). For example, a systolic arterial blood pressure of 80 mmHg for 30 min would equal a T^{sBP < 100}of 600 mmHg · min; a systolic pulmonary artery blood pressure of 40 mmHg for 30 min would equal a T^{sPAP > 30}of 300 mmHg · min. These variables do not correspond with formal definitions of systemic hypotension and pulmonary hypertension but are indices that include almost all patients, providing sensitive continuous variables to assess the relation of hemodynamic perturbations and mortality.

In addition to linear analyses, we tested binary definitions for “severe” systemic hypotension (≥ 30 mmHg below baseline) and pulmonary hypertension (> 40 mmHg, including an increase of at least 10 mmHg above baseline). Baseline for each variable was the median value derived from the 5 min before protamine administration.

Anesthesia included methadone or diazepam premedication, midazolam, fentanyl, thiopental, and isoflurane. Pancuronium was used for muscle relaxation. Nonpulsatile, hypothermic (28°–34°C) CPB was conducted using a membrane oxygenator (Cobe CML; Cobe Inc., Arvada, CO) after a crystalloid and mannitol prime and using an arterial line filter (Cobe). Porcine heparin was administered (300 U/kg) and supplemented during CPB to maintain an activated clotting time (Hemochron 801; International Technidyne Corp., Edison, NJ) of greater than 480 s. Protamine sulfate (Eli Lilly, Indianapolis, IN) was administered until the activated clotting time returned to baseline or a ratio of 1 mg:100 U heparin was reached. Although we do not have access to the specific rates of protamine administration for our study patients, a typical strategy would be to administer a 20-mg test dose that, if tolerated, is then followed by removal of the aortic cannula and the subsequent administration of 3–3.5 mg/kg as a slow, intermittent bolus over 10 min through a peripheral vein. This is a low dose (typically approximately 0.7–0.8 mg protamine/100 U heparin given), and further protamine in 25- to 50-mg increments is given as needed based on an activated clotting time test performed 5 min after the initial dose is completed. Occasionally, further optional protamine doses are given in the intensive care unit during the postoperative period, again based on activated clotting time testing.

After separation from CPB and in the intensive care unit, inotropic and vasoactive agents were used as necessary to maintain a cardiac index of greater than 2.0 l · min⁻¹· m⁻²and a mean arterial pressure of greater than 60 mmHg. A postoperative care pathway was followed with the aim of discharge from the intensive care unit on postoperative day 1 and from the hospital on postoperative day 5.

The primary outcome measure was in-hospital mortality. Descriptive statistics are presented as mean and SD for continuous data and percentage for categorical data. The Hannan score was described by the median and 25th–75th percentile range. Thirty-minute postprotamine linear hemodynamic measures and the odds ratio and their 95% confidence intervals (CIs) were calculated. The relations between 30-min postprotamine linear hemodynamic measures and mortality was assessed using multiple logistic regression, adjusting for preoperative and intraoperative risk factors. Covariates were the components of the Hannan score. In addition, duration of CPB and number of grafts were included as markers of procedure complexity. Initially, all terms were included as predictors in a multivariable model, and terms not significant at P < 0.05 were eliminated from the model one at a time until only significant terms remained. A c-index, or area under the receiver operating characteristic curve, was calculated to describe the predictive ability of the model. The Hosmer–Lemeshow goodness-of-fit statistic was calculated to assess how well the models fit the data. Tolerance and variance inflation diagnostics were performed on a model containing all significant covariates to determine whether the presence of multicollinearity among predictors. Predicted probabilities of death for each patient were generated from these models for graphing purposes, using mean values to control for preoperative and intraoperative risk factors.

The primary analyses were repeated for 15-min postprotamine intervals to assess the relation of proximity of hemodynamic responses to outcome. Equivalent analyses were repeated using the binary definitions of hemodynamic change as predictors in place of the continuous variables.

Because degree-duration integrals do not distinguish between large-brief and modest-sustained hemodynamic changes, we repeated the primary analysis, excluding patients who met our binary criteria as a marker for large “severe” hemodynamic changes.

After a logistic regression model was developed for the sample, further testing of our data included investigation of nonlinear terms and combinations of linear terms or splines. For example, if outliers were overly influential in determining the slope of the line, we might expect to see a deviation from linearity, or an “elbow,” where the line was flat and then increased rapidly. By testing restricted cubic splines, typically used to describe smooth curves rather than elbows, we wanted to assess whether higher-order terms would contribute explanatory power to our model.

SAS version 8.02 (SAS Institute, Inc., Cary, NC) was used for all analyses; a P  value of less than 0.05 was considered significant. All tests are two tailed.

Of the 7,306 patients who met our inclusion criteria with clinical outcomes data, we were able to match 6,846 (93.7%) to our database containing intraoperative blood pressure data. Of the 6,846 patients included in the study, a subgroup of 5,649 (82.5%) also had pulmonary artery pressure monitoring. Seventy-five patients were excluded because no protamine dose was recorded; only one case of protamine allergy was documented in the anesthesia records. Patient characteristics and other variables are listed in table 1. Predicted in-hospital mortality (median Hannan score) was 2%; actual in-hospital mortality was 2%. The mean time between first and last dose of protamine in the operating room was 22 min (SD 18), and the mean protamine dose was 257 ± 95 mg.

Figure 1illustrates the range of systolic arterial and pulmonary artery systolic blood pressures measured over the 30-min period after protamine administration. Hemodynamic perturbations occurred within 10 min of protamine administration for most patients (e.g. , 20% decrease in systemic blood pressure, 52% of patients; 20% increase in pulmonary artery pressure, 92% of patients).

Strong associations of both systemic hypotension and pulmonary artery hypertension related to protamine administration with mortality are present in our data (table 2). After accounting for comorbidities and markers of procedure complexity, the Hosmer–Lemeshow goodness-of-fit statistic, calculated for the systemic hypotension model (8 df , chi-square 2.56, P = 0.96) and the pulmonary hypertension model (8 df , chi-square 5.58, P = 0.69), did not support rejection of the null hypothesis that the models provide a good fit to the data. The c-index (area under the receiver operating characteristic curve) was 0.78 for both the systemic hypotension model and the pulmonary hypertension model. Tolerance values of the predictors range from 0.81 to 0.96 (corresponding to variance inflation values ranging from 1.03 to 1.24). Because the tolerance values are all greater than 0.4, we conclude that the model is appropriate statistically and no significant collinearity exists among our covariates. Other covariates associated with mortality risk are similar for both primary hemodynamic analyses and resemble those previously reported,15including advanced age, female sex, history of congestive heart failure, need for postoperative intraaortic balloon counterpulsation, extended duration of CPB, reduced left ventricular ejection fraction, and greater number of grafts. Figures 2A and Billustrate the probability of death with increasing T^{sBP < 100}and T^{sPAP > 30}, respectively, while controlling for preoperative and intraoperative risk factors set at mean values. For example, T^{sBP < 100}values of 300, 600, and 900 mmHg · min are associated with mortality rates of 3.0, 5.0, and 8.7%, respectively (fig. 2A); the mean T^{sBP < 100}is 100 ± 148 mmHg · min. Based on this SD, an increment of 150 mmHg · min was used for odds ratio calculations of the association of T^{sBP < 100}(and T^{sPAP > 30}for consistency). Similarly, T^{sPAP > 30}values of 300, 600, and 900 mmHg · min are associated with in-hospital mortality rates of 2.6, 4.4, and 7.4%, respectively (fig. 2B); the mean T^{sPAP > 30}is 56 ± 110 mmHg · min.

Proximity of hemodynamic response to protamine administration strengthens the relation between systemic hypotension and pulmonary hypertension and mortality. Analyses using 15- and 30-min observation periods after protamine administration demonstrate a greater adjusted odds ratio of mortality per 150 mmHg for the 15-min assessment for both systemic hypotension (T^{sBP < 100}: 1.40 [95% CI, 1.11–1.73]vs.  1.28 [95% CI, 1.14–1.43]) and pulmonary hypertension (T^{sPAP > 30}: 1.50 [95% CI, 1.07–2.04]vs.  1.27 [95% CI, 1.06–1.48]; table 2). To determine whether the associations of systemic blood pressure and pulmonary pressure are independent of each other, we constructed a model for the 30-min data in which both blood pressure indices were included as predictors, along with the previously described covariates. Both the systemic and pulmonary pressure variables retained their significance in the model (T^{sBP < 100}: odds ratio, 1.28 [95% CI, 1.13–1.42]; P < 0.0001; T^{sPAP > 30}: odds ratio, 1.26 [95% CI, 1.05–1.47]; P = 0.007), indicating that they have independent effects on mortality.

Table 3illustrates binary definitions of systemic hypotension (≥ 30 mmHg below baseline) and pulmonary hypertension (> 40 mmHg, including an increase of at least 10 mmHg above baseline) with mortality, adjusted for preoperative and intraoperative risk factors; odds ratios were calculated after controlling for preoperative and intraoperative risk factors in a logistic regression model. Overall, “severe” hemodynamic responses related to protamine were common; 23.6% of patients met the definition of either systemic hypotension or pulmonary hypertension, and mortality in this group was 3.4%, compared with 1.8% for the remainder. All typical hemodynamic responses were associated with increased mortality risk (table 3). To assess smaller sustained hemodynamic changes, we performed a secondary analysis excluding patients meeting our criteria for severe hemodynamic responses outlined above; a strong relation persisted with mortality for both systemic hypotension (P = 0.008) and pulmonary hypertension (P = 0.03).

To further assess the association of hemodynamic responses with mortality, we tested the possibility that our data were not well described by linear models by investigating nonlinear terms and combinations of linear terms or splines. We found no point in our data at which a significant deviation from linearity could be detected. In addition, a test of restricted cubic splines, which fit smooth curves, determined that higher-order terms did not add explanatory power to our model. These findings indicate that a logistic regression model (which is linear in the logit) is an appropriate description of the relations. This linearity supports an association between hemodynamic responses and mortality throughout the observed range of values.

In our study, we found an independent relation between hemodynamic changes after protamine administration and in-hospital mortality after primary CABG surgery. Even after accounting for comorbidities and procedure complexity, systemic hypotension and pulmonary hypertension during the 30 min after protamine administration are both independently associated with in-hospital mortality. The relations are continuous over the range of hemodynamic response and persist even when severe reactions are removed from analysis. Tests for linearity confirm associations even in the lowest observed range of values for both pressure effects. Although our findings suggest that common “minor” protamine reactions may not be benign and may even contribute to adverse outcome, because no equivalent drug devoid of adverse effects is available for comparison in a randomized trial, it is only possible to speculate that a direct causal link exists between protamine and mortality risk.

In a retrospective chart review of 2,159 cardiac surgical patients, Kimmel et al.  4reported an increased mortality rate in the 2.6% of patients with major complications, including hemodynamic instability, temporally related to protamine. We have extended this concept using continuous variables that document the full range of hemodynamic changes. Our findings indicate that indices of hemodynamic change reflecting the degree and duration of systemic hypotension and pulmonary hypertension during the minutes after protamine administration can be used to predict mortality risk.

One of the limitations of a retrospective investigation is the risk of incomplete or biased data. We used hemodynamic data from an automated anesthesia record–keeping system that should not be prone to error. Outcomes for our study were taken from a prospectively gathered quality-assurance database entered within several days of hospital discharge that should be more accurate and less biased than a traditional chart review. A timing bias with regard to the charting of protamine dosing is possible; however, it is standard at our institution to document drugs through the touch-screen format located on the anesthesia machine as they are given. The relation of hemodynamic responses to protamine dosing in our study is supported by the strengthening of the relation with increasing proximity to recorded time of drug administration.

The relation of pulmonary artery hypertension after protamine administration with adverse outcome in our study may reflect the responses of patients with poor myocardial function and increased baseline pulmonary artery pressure. However, our secondary analysis using a more stringent definition to assess acute pulmonary hypertensive response still identifies an association of hemodynamic response with mortality, even after controlling for markers of preoperative and intraoperative risk, including left ventricular ejection fraction.

A criticism of our study is that we do not account for clinical interventions to treat hemodynamic perturbations. However, the majority of these would obscure the hemodynamic perturbations and therefore reduce the likelihood of identifying statistical association between these responses and adverse outcome.

Other factors may be related to hemodynamic instability after CPB involving surgical complexity or underlying premorbid conditions that increase the risk of adverse outcome. We designed our analysis to include markers of preoperative mortality risk and intraoperative complexity described in the literature15and tested to confirm association with mortality in our population; we still found hemodynamic instability after protamine administration to be independently associated with mortality.

Cause and effect cannot be conclusively demonstrated in a retrospective study, because uncontrolled confounding may persist despite adjustments. However, it is important to explore potential pathophysiological mechanisms. A direct link between catastrophic hemodynamic collapse after protamine and mortality is easy to comprehend, but mechanistic links between the common, modest hemodynamic changes and excess mortality in our study have not been investigated. Complement activation and other related components of the inflammatory response are linked to outcomes after cardiac surgery, providing a plausible interpretation for our findings. Heparin–protamine complexes are known to activate the classic complement pathway9and trigger release of inflammatory mediators.2,11Complement fragments C3a and C5a (anaphylotoxins) degranulate basophils and mast cells, releasing histamine, leukotrienes, and thromboxanes. C5a activates macrophages, monocytes, and neutrophils, resulting in proteolytic enzyme and interleukin-1 release, free radical generation, margination, chemotaxis, and increased vascular permeability.18Histamine contributes to cause the peripheral vasodilation and systemic hypotension,19consistently observed after protamine reversal of heparin,8whereas C5a and thromboxane, released after protamine, can precipitate pulmonary hypertension and bronchospasm,11thereby linking our observations to pathophysiologic changes associated with protamine. Complement activation after CABG surgery is related to outcome, including postoperative arrhythmias20and pulmonary dysfunction.21Similarly, inhibition of the complement cascade reduces leukocyte and platelet activation22and improves outcome after CABG surgery.23Although proinflammatory mediators are activated during heparin reversal by protamine, a direct link between these factors and adverse outcome, including mortality risk, is unconfirmed and speculative only.

Because the full range of blood pressure responses related to protamine are associated with mortality risk after cardiac surgery, hemodynamic “protamine reactions” may be a useful metric to evaluate future therapies aimed at avoiding associated adverse outcome.24,25In addition, investigations are necessary to assess whether equivalent associations between protamine and outcome exist in other procedures and settings where protamine is used.

In summary, hemodynamic responses in the 30 min after protamine administration for heparin neutralization during CABG surgery are associated with in-hospital mortality after controlling for preoperative and intraoperative risk factors, even in the lowest observed range of values for both systemic hypotension and pulmonary hypertension. This is consistent with and extends previous findings that major adverse events, temporally related to protamine administration, are associated with increased mortality risk.4Future randomized trials are necessary to differentiate association from causality in the relation between hemodynamic “protamine reactions” and adverse outcome. These trials may also identify therapeutic strategies that reduce such hemodynamic perturbations and improve patient outcome after cardiac surgery.

The authors thank Cheryl J. Stetson, A.A.S. Paralegal Technology, Staff Assistant, Department of Anesthesiology, Duke University Medical Center, Durham, North Carolina, for editorial assistance.