Red Blood Cells and Mortality After Coronary Artery Bypass Graft Surgery: An Analysis of 672 Operative Deaths

Gaetano Paone, MD, MHSA, Morley A. Herbert, PhD, Patricia F. Theurer, BSN, Gail F. Bell, MSN, Jaelene K. Williams, MSN, Francis L. Shannon, MD, Donald S. Likosky, PhD, and Richard L. Prager, MD Division of Cardiac Surgery, Henry Ford Hospital, Detroit, Michigan; Sections of Health Services Research and Quality and Adult Cardiac Surgery, Department of Cardiac Surgery, University of Michigan, Ann Arbor, Michigan; Michigan Society of Thoracic and Cardiovascular Surgeons Quality Collaborative, Ann Arbor, Michigan; Southwest Data Consultants, Dallas, Texas; and Division of Cardiovascular and Thoracic Surgery, William Beaumont Hospital, Royal Oak, Michigan

Background. Prior studies have implicated transfusion as a risk factor for mortality in coronary artery bypass graft surgery (CABG). To further our understanding of the true association between transfusion and outcome, we specifically analyzed the subgroup of patients who died after undergoing CABG. Methods. A total of 34,362 patients underwent isolated CABG between January 2008 and September 2013 and were entered into a statewide collaborative database; 672 patients (2.0%) died and form the basis for this study. Univariate analysis compared preoperative and intraoperative variables, as well as postoperative outcomes, between those with and without transfusion in both unadjusted cohorts and those matched by predicted risk of mortality (PROM). Mortality was further evaluated with phase of care analysis. Results. Of the 672 deaths, 566 patients (84.2%) received a transfusion of red blood cells. The PROM was 7.5% for the transfused patients versus 4.3% for those not transfused (p < 0.001). Transfused patients

were older, more often female, had more emergency, on-pump, and redo procedures, and had a lower preoperative and on-bypass nadir hematocrit. Most other demographics were similar between the groups. Postoperatively, transfused patients were ventilated longer, had more renal and multisystem organ failure, and were more likely to die of infectious and pulmonary causes after longer intensive care unit and overall lengths of stay. Conclusions. Significant differences in PROM and the postoperative course leading to death between those with and without transfusion suggest the role of transfusion may be secondary to other patient-related factors. Recognizing that the relationship between transfusion and outcome after CABG remains incompletely understood, these findings are suggestive of a complex interaction of many variables.

P

In this study we attempt to further our understanding of the manner in which transfusion may be related to morbidity and mortality by specifically analyzing the subgroup of patients who died after undergoing CABG.

rior studies have implicated red blood cell (RBC) transfusion as a risk factor for increased morbidity as well as short- and long-term mortality after coronary artery bypass graft surgery (CABG) [1–4]. However, as these reports consistently demonstrate significant demographic differences between transfused and nontransfused cohorts, the true association between transfusion and outcome remains unclear. Despite the use of increasingly sophisticated statistical techniques to “match” these differing patient populations, it remains largely uncertain whether transfusion is in fact the cause of these worse outcomes or rather serves as a surrogate marker for a patient population at higher risk.

Accepted for publication Dec 16, 2014. Presented at the Sixty-first Annual Meeting of the Southern Thoracic Surgical Association, Tucson, AZ, Nov 5-8, 2014. Address correspondence to Dr Paone, Division of Cardiac Surgery, Henry Ford Hospital, 2799 W Grand Blvd, Detroit, MI 48202; e-mail: [email protected].

Ó 2015 by The Society of Thoracic Surgeons Published by Elsevier

(Ann Thorac Surg 2015;99:1583–90) Ó 2015 by The Society of Thoracic Surgeons

Material and Methods This study was approved by the Institutional Review Board of the University of Michigan Health System (IRB HUM00053934, Notice of Determination of “Not Regulated” Status).

Patient Population The Michigan Society of Thoracic and Cardiovascular Surgeons Quality Collaborative (MSTCVS-QC) is a multidisciplinary group consisting of all 33 hospitals that perform adult cardiac surgery in the State of Michigan [5]. All programs use the Society of Thoracic 0003-4975/$36.00 http://dx.doi.org/10.1016/j.athoracsur.2014.12.064

ADULT CARDIAC

Red Blood Cells and Mortality After Coronary Artery Bypass Graft Surgery: An Analysis of 672 Operative Deaths

ADULT CARDIAC

1584

PAONE ET AL RBC TRANSFUSION AND MORTALITY

Surgeons (STS) data collection form and submit data on a quarterly basis to both the STS database and the MSTCVS-QC data warehouse. Data collected include perioperative, operative, and outcomes data on all patients undergoing cardiac surgery at the participating hospitals. Data managers meet quarterly and audits are conducted biannually to ensure data integrity. Seventysix data elements, including those used in the STS morbidity and mortality risk models, are audited from a random sample of 20 procedures performed at each center. For consistency, all audit visits are conducted by a core group of trained quality collaborative nurses. Between January 2008 and September 2013, a total of 34,362 patients underwent isolated CABG in the State of Michigan. These included patients undergoing both onpump and off-pump procedures, and data for all were entered into the statewide collaborative database. Salvage procedures were excluded. Operative mortality was defined according to STS database guidelines and includes all deaths occurring inhospital at any duration and those deaths occurring after discharge from the hospital, but within 30 days of the procedure. A total of 672 patients (2.0%) died and form the basis for this report. Additional analysis of patient mortality was undertaken using the phase of care mortality analysis (POCMA) methodology initially introduced into the MSTCVS database in 2006. The POCMA methodology is predicated on the identification of a seminal event that initiates the subsequent clinical sequence ultimately leading to patient mortality [6]. This “death trigger,” as identified by the operating surgeon, is then categorized by its timing into the preoperative, intraoperative, postoperative intensive care unit (ICU), postoperative floor, or discharge phase of care. Further analysis attempts to subcategorize the initiating event within each phase of care. The intent of the POCMA process is to identify the time and location of the sentinel event or root cause ultimately leading to death.

Statistical Analysis Summary data are presented as proportions or means and standard deviations. Proportions are compared using c2 statistics, and continuous variables are compared with Student’s t tests. A subgroup of the mortalities was also matched using a 1:1 greedy matching algorithm. Initial univariate analysis of risk factors between transfusion and no transfusion groups provides a starting point. Factors are then added into the matching algorithm, and the matching is run. The preoperative risk factors are then compared between groups, with those having significant differences being added into the algorithm. The matching is run until there are no important differences between risk factors in the two groups. All analyses were performed using SAS 9.4 (SAS Institute, Cary, NC). The tests were considered significant at a probability value of less than 0.05.

Ann Thorac Surg 2015;99:1583–90

Results Study Population Of the 672 deaths, 566 patients (84.2%) received a transfusion of RBCs. Of these, 66.1% received an average of 3 units of RBCs in the operating room and 87.6% an average of 5.8 units during the postoperative period. Included in the group of 106 patients not transfused with RBCs were 7 patients who did receive platelets or fresh frozen plasma. Predicted risk of mortality (PROM) was 7.5% for the transfused patients versus 4.3% for those 106 patients not receiving a transfusion (p < 0.001). Transfused patients were older, larger, and more often female, and had a higher incidence of peripheral arterial disease. They also underwent more redo and emergency procedures and had more preoperative balloon pumps, a lower preoperative hematocrit, and a higher baseline creatinine level. Most other preoperative demographics were similar between the groups (Table 1). Those transfused had a significantly lower on-bypass nadir hematocrit than the not transfused group. Offpump CABG surgery, performed in less than 10% of patients overall, was associated with fewer transfusions (Table 2). Perfusion but not aortic cross-clamp times were longer in the transfused group. More transfused patients required intraoperative placement of balloon pumps (9.7% versus 1.9%; p ¼ 0.008). Postoperatively, transfused patients were more likely to undergo reoperation, most commonly for bleeding. They were ventilated longer and more often reintubated, had more renal and multisystem organ failure, and had longer ICU and total hospital lengths of stay (Table 3). Transfused patients more often died of infectious and pulmonary causes, whereas those not transfused were more likely to have died of neurologic or vascular causes as well as for reasons categorized in the database as unknown (Table 4).

Predicted Risk of Mortality Matched Population A total of 160 patients—80 from each cohort—were matched on the basis of preoperative predicted risk (transfused PROMadj 2.26 versus not transfused PROMadj 2.25; p ¼ 0.98). With few exceptions the matching process equalized many of the differences in preoperative characteristics. The relative frequency and distribution of postoperative outcomes and cause of death remained qualitatively similar to those of the unmatched populations (Tables 1–4).

Phase of Care Mortality Analysis As shown in Table 5, there were significant differences between the groups in the distribution of both phase of care assignment and what was considered the likely initiating event leading to mortality (p < 0.001). A higher proportion of deaths in the transfused patients were assigned to the preoperative phase of care. Differences within the operative phase were mostly related to either technical mishaps or surgeon judgment. Beyond the operating room, the majority of deaths in both groups

PAONE ET AL RBC TRANSFUSION AND MORTALITY

1585 ADULT CARDIAC

Ann Thorac Surg 2015;99:1583–90

Table 1. Preoperative Demographics of Dead Patients Who Received Blood Versus No Blood All Patients Variable Number of patients STS predicted risk of mortality Patient age (y) Sex (% male) Height (cm) Weight (kg) Whites Last preoperative hematocrit (%) Last preoperative creatinine (mg/dL) Diabetes (yes) Hypertension Chronic lung disease Moderate Severe Cerebrovascular disease Renal failure—dialysis Immunosuppression Peripheral arterial disease Myocardial infarction Heart failure within 2 weeks NYHA class III IV Operative status Elective status Urgent status Emergent status Left main disease >50% No. of diseased vessels Single-vessel Two-vessel Triple-vessel Preoperative IABP Cardiogenic shock Ejection fraction Preoperative b-blocker Preoperative aspirin ADP inhibitors Preoperative lipid-lowering agent Previous CABG

Transfused

No Blood

566 7.54 70.82 57.95% 167.75 83.91 82.69% 35.51 1.57 45.41% 90.46%

106 4.33 67.04 74.53% 172.38 98.96 90.91% 39.15 1.27 55.66% 89.62%

8.66% 14.31% 27.74% 9.72% 9.36% 34.45% 70.32% 33.75%

8.49% 10.38% 19.81% 3.77% 6.60% 21.70% 66.98% 26.42%

12.94% 17.38%

15.09% 7.55%

19.79% 65.19% 15.02% 46.64%

30.19% 64.15% 5.66% 37.74%

3.00% 14.31% 82.51% 20.49% 10.60% 46.26 82.86% 87.63% 16.28% 72.44% 7.95%

0.00% 20.75% 79.25% 10.38% 6.60% 47.54 88.68% 84.91% 8.49% 83.96% 1.89%

PROM Matched p Value <0.001 0.001 <0.001 <0.001 <0.001 0.04 <0.001 0.02 0.05 0.79 0.71

0.09 0.05 0.36 0.01 0.49 0.14 0.16

Transfused

No Blood

80 2.26 67.46 80.00% 172.33 90.67 95.00% 37.53 1.12 37.50% 83.75%

80 2.25 67.14 80.00% 173.56 98.53 95.00% 39.19 1.2 51.25% 90.00%

10.00% 11.25% 22.50% 1.25% 5.00% 25.00% 61.25% 27.50%

10.00% 7.50% 15.00% 3.75% 8.75% 21.25% 63.75% 22.50%

15.19% 8.86%

11.25% 7.50%

26.25% 73.75% — 38.75%

26.25% 73.75% — 37.50%

1.25% 11.25% 87.50% 8.75% 0.00% 48.88 91.25% 87.50% 16.25% 82.50% —

0.00% 12.50% 87.50% 7.50% 2.50% 47.85 91.25% 88.75% 7.50% 85.00% —

0.006

0.09 0.12

0.015 0.21 0.42 0.24 0.32 0.09 0.03 0.03

Comment The results of our study reveal significant clinical differences between those who die after CABG in terms of

0.98 0.84 1.00 0.43 0.04 1.00 0.03 0.58 0.08 0.24 0.48

0.22 0.31 0.35 0.57 0.74 0.47 0.80

1.00

ADP ¼ adenosine diphosphate; CABG ¼ coronary artery bypass graft surgery; IABP ¼ intraaortic balloon pump; Association; PROM ¼ predicted risk of mortality; STS ¼ The Society of Thoracic Surgeons.

were judged to be catastrophic in nature. Almost one third of all deaths in those not transfused occurred after discharge, compared with only 12.6% of those in the transfused group (p < 0.001).

p Value

0.87 0.59

0.77 0.16 0.64 0.87 0.60 0.09 0.67 —

NYHA ¼ New York Heart

those who were transfused versus those not transfused. Preoperative differences, in particular those regarding patient age, sex, body size, and baseline hematocrit, are consistent with factors predictive of transfusion in general after CABG, and are therefore likely responsible for the need for transfusion in at least some of these patients, independent of the subsequent clinical course. Transfused patients more frequently experienced sepsis and renal and multisystem organ failure. They were on the

ADULT CARDIAC

1586

PAONE ET AL RBC TRANSFUSION AND MORTALITY

Ann Thorac Surg 2015;99:1583–90

Table 2. Intraoperative Variables of Dead Patients Who Received Blood Versus No Blood All Patients Variable Number of patients Off-pump procedure No. of grafts (mean) Perfusion time (min) Cross-clamp time (min) CPB lowest hematocrit (%) RBC Mean RBC units FFP Cryoprecipitate Platelets IABP Delayed sternal closure CPB ¼ cardiopulmonary bypass; RBC ¼ red blood cell.

PROM Matched

Transfused

No Blood

p Value

Transfused

No Blood

p Value

566 9.01% 3.06 119.65 75.61 22.18 66.08% 2.98 26.33% 6.18% 34.45% 9.72% 9.20%

106 16.04% 3.04 106.4 74.88 27.35 0.00% — 0.94% 0.00% 1.89% 1.89% 2.86%

0.03 0.86 0.01 0.85 <0.001 <0.001 — <0.001 0.009 <0.001 0.008 0.21

80 5.00% 3.48 116.47 80.11 23.13 45.00% 2.39 20.00% 2.50% 26.25% 15.00% 11.11%

80 13.75% 3.23 103.78 75.72 26.87 0.00% — 0.00% 0.00% 1.25% 1.25% 3.85%

0.06 0.10 0.11 0.38 0.007 <0.001 — <0.001 0.16 <0.001 0.001 0.35

FFP ¼ fresh frozen plasma;

IABP ¼ intraaortic balloon pump;

ventilator and in the ICU longer, underwent more reoperations especially for bleeding, and more often died of infectious and pulmonary causes after longer hospital stays. Conversely, the not transfused group spent significantly less time on the ventilator and in the ICU and more often died of neurologic and vascular events.

PROM ¼ predicted risk of mortality;

Differences in PROM identify the transfused group as a higher risk cohort (7.5 versus 4.3). This is also consistent with the findings of our POCMA analysis whereby almost 30% of deaths in the transfused cohort were allocated to the preoperative phase and determined by the surgeon to have been the result of either judgment in patient

Table 3. Comparison of Dead Patients Who Received Blood Versus No Blood Postoperative Outcomes All Patients Variable Number of patients Any reoperation Reoperation for bleeding Postoperative IABP Deep sternal wound infection Prolonged ventilation Permanent stroke Postoperative renal failure Postoperative dialysis Postoperative atrial fibrillation Multisystem organ failure Postoperative creatinine (mg/dL) PRBC Mean RBC units Platelets FFP Cryoprecipitate Total ventilator hours Reintubated Total ICU hours Length of stay

PROM Matched

Transfused

No Blood

p Value

Transfused

No Blood

p Value

566 16.43% 11.15% 7.24% 2.30% 68.85% 9.56% 29.56% 16.96% 33.27% 17.80% 2.64 87.63% 5.84 36.93% 34.10% 10.60% 161.62 44.07% 234.84 11.51

106 0.00% 0.00% 1.89% 0.94% 31.13% 14.15% 11.32% 5.66% 29.25% 4.76% 1.84 0.00% — 3.77% 1.89% 0.00% 46.63 26.42% 97.85 7.53

<0.001 <0.001 0.04 0.37 <0.001 0.15 <0.001 0.003 0.412 0.002 <0.001 <0.001 — <0.001 <0.001 <0.001 <0.001 <0.001 <0.001 <0.001

80 20.00% 15.00% 7.50% 2.50% 62.50% 7.50% 41.25% 23.75% 37.50% 16.44% 2.45 86.25% 7.13 33.75% 40.00% 15.00% 219.36 41.25% 321.82 14.05

80 0.00% 0.00% 1.25% 1.25% 32.50% 16.25% 10.00% 7.50% 31.25% 4.76% 1.75 0.00% — 2.50% 2.50% 0.00% 49.16 23.75% 98.2 8.26

<0.001 <0.001 0.05 0.56 <0.001 0.09 <0.001 0.005 0.41 0.03 0.001 <0.001 — <0.001 <0.001 <0.001 0.002 0.02 0.005 0.03

FFP ¼ fresh frozen plasma; IABP ¼ intraaortic balloon pump; PROM ¼ predicted risk of mortality; RBC ¼ red blood cell.

ICU ¼ intensive care unit;

PRBC ¼ packed red blood ells;

PAONE ET AL RBC TRANSFUSION AND MORTALITY

1587 ADULT CARDIAC

Ann Thorac Surg 2015;99:1583–90

Table 4. Cause of Death of Dead Patients Who Received Blood Versus No Blood All Patients Variable Number of patients Discharge status—dead Cause of death Cardiac Infection Pulmonary Neurologic Vascular Other Unknown

Transfused

No Blood

566 80.39%

106 59.43%

47.00% 8.83% 15.02% 8.66% 1.94% 8.83% 7.95%

40.57% 2.83% 9.43% 14.15% 5.66% 9.43% 16.98%

PROM Matched p Value <0.001 <0.001

Transfused

No Blood

80 80.00%

80 53.75%

40.00% 17.50% 16.25% 8.75% 0.00% 7.50% 7.50%

35.00% 3.75% 11.25% 17.50% 6.25% 8.75% 16.25%

p Value <0.001 <0.001

PROM ¼ predicted risk of mortality.

selection, inadequate patient evaluation, or excessive risk profiles. All of these category designations characterize a patient who the operating surgeon in retrospect believed should have been considered a prohibitive risk for CABG surgery. Transfusion is unlikely to have contributed to death in these patients. Differences in PROM and baseline hematocrit levels between the groups would also then characterize the not transfused cohort as initially at lower risk for both transfusion and death. They nonetheless have died, more frequently after discharge of

catastrophic events, without having been previously transfused. This is also supported by results from our POCMA database as almost one third of deaths in this not transfused group were assigned to the discharge phase. Most prior studies have evaluated the relationship between transfusion and outcome by comparing results of those transfused with those not transfused. All routinely document significant differences in patient characteristics that categorize the transfused cohort as higher risk. These differences are then addressed statistically by

Table 5. Phase of Care Mortality Analysis Variable

Transfused

Number of patients Phase of care mortality category Preoperative phase Judgment Cardiac risk factor profile Noncardiac risk factor profile Patient evaluation Intraoperative phase Intraoperative phase surgeon Catastrophic event Anesthesia Postoperative ICU phase Catastrophic event Surveillance/recognition of decompensation Respiratory care Other Postoperative floor phase Catastrophic event Surveillance/recognition of decompensation Sepsis prevention/treatment Discharge phase Catastrophic event Adequate instruction and support network Other

566

ICU ¼ intensive care unit;

RBC ¼ red blood cells.

166 63 44 30 19 91 49 30 8 166 93 26 15 18 71 44 12 9 72 57 1 7

No RBC

p Value

106 (29.3%) (11.1%) (7.8%) (5.3%) (3.4%) (15.9%) (8.7%) (5.3%) (1.4%) (29.3%) (16.4%) (4.6%) (2.7%) (3.2%) (12.4%) (7.8%) (2.1%) (1.6%) (12.6%) (10.1%) (0.2%) (1.2%)

18 7 8 1 1 9 3 5 1 26 16 4 2 2 19 12 4 0 34 31 1 2

(16.8%) (7.5%) (7.5%) (0.9%) (0.9%) (8.4%) (2.8%) (4.7%) (0.9%) (24.3%) (15.0%) (3.7%) (1.9%) (1.9%) (17.8%) (11.2%) (3.7%) (0%) (31.8%) (29.2%) (0.9%) (1.9%)

0.01 0.16 0.92 0.09 0.30 0.04 0.06 0.99 1.00 0.32 0.73 0.91 0.90 0.68 0.14 0.23 0.50 0.40 <0.001 <0.001 0.720 0.941

ADULT CARDIAC

1588

PAONE ET AL RBC TRANSFUSION AND MORTALITY

a variety of matching techniques. Whether analyzing the specific outcome of mortality, other individual morbidities, or estimates of cost, the results almost always show that transfused patients have worse outcomes and greater cost than those not transfused, even with amounts as small as 1 and 2 units [7–11]. This has led to the dictum that transfusion is bad and a conclusion that avoidance of transfusion will result in less morbidity and mortality after CABG. This certainly seems rational, but one must always acknowledge that the relationship cannot be definitively proven to be one of cause and effect, but rather must be considered within the context of a variety of potential known and unknown confounding effects [12]. With the exception of the rare patient with a massive acute hemorrhage, essentially all transfusions during CABG are administered to address some predefined level of anemia. Anemia is common in ICU patients, with up to 90% of patients diagnosed as such by day 3 of the ICU stay. The cause is most frequently multifactorial and may include nutritional deficiencies with low iron stores, hemolysis owing to drug reactions, and blood loss related to coagulation abnormalities, gastrointestinal bleeding, and excessive phlebotomy [13, 14]. There is also evidence for impaired erythrocyte production as a consequence of abnormal iron metabolism and a diminished response of the bone marrow to endogenous erythropoietin, creating what has been termed the acute anemia of chronic disease [15, 16]. It is easy then to understand how the prolonged and complicated clinical conditions more common in the transfused group would lead to more profound levels of anemia and greater use of transfusion in a clinical scenario that ultimately led to death. However, why the transfusion itself should then become explicitly identified as either a significant contributor to or the immediate cause of death is considerably less apparent. A number of investigators have shown even worse outcomes when transfusion is added to anemia [17–19]. Engoren and colleagues [20] evaluated late mortality in 922 patients who underwent isolated CABG during a 3.5-year period. They found that patients with preoperative anemia who received transfusion had a threefold hazard of death compared with those without anemia and transfusion, and twice that of patients who were anemic but not transfused. Of interest, they limited their analysis to late mortality because “it is unlikely that in the absence of malignant transfusion reaction, the transfusion would be the proximate cause of acute perioperative death; rather, it would more likely be just a marker for a more complex and complicated perioperative course.” The findings of our study seem to agree with their reasoning; our results show that patients who died having been transfused generally had more prolonged and complicated postoperative courses. The controversy within this notion that “.these patients died in spite of blood transfusion, not because of it” is the essence of the present debate regarding the interaction of transfusion with outcome. Yun and associates [21] found that exposure to 1 to 2 units of RBCs did not significantly increase the risk

Ann Thorac Surg 2015;99:1583–90

of either early or late death in octogenarians undergoing various cardiac procedures. They postulated that perhaps transfusion was beneficial in the face of a greater incidence of anemia in the elderly, or alternatively some other age-related causes of morbidity confounded the findings. Salisbury and coworkers [22] analyzed 34,937 patients from 57 centers hospitalized for treatment of acute myocardial infarction. Of these, 1,778 (5.1%) were transfused with at least 1 unit of RBCs. There were significant clinical differences between the transfused and not transfused groups, and transfusion was associated with a significantly higher mortality (unadjusted odds ratio, 2.05; 95% confidence interval, 1.76 to 2.40). Unlike most other studies, their complex matching process included not only admission but also posttreatment patient variables, after which transfusion was now associated with a lower in-hospital mortality (adjusted odds ratio, 0.73; 95% confidence interval, 0.58 to 0.92). In fact, we did match the two groups in our study on the basis of preoperative PROM, thereby controlling for preoperative predictors of transfusion and death. Despite lower risk in both matched cohorts, the distribution of morbid outcomes and cause of death remained similar to those in the unmatched population. We believe this supports the idea that many transfusions are administered not as a precursor to but as a result of the clinical circumstances. This report is subject to the limitations of other retrospective observational database analyses. Our database does not include information regarding the timing or indication for transfusion. We are unable to account for a variety of other potential confounders, including those of practice patterns at the physician and institutional level [23]. There is the possibility that the determination of the cause of death may well be subjective and vary among physicians and centers. The POCMA data presented are subject to similar concerns regarding reporting inconsistency. In addition, approximately one fourth of the deaths in those not transfused are reported as other or unknown. In this study we sought to leverage the large size of our multicenter collaborative database to determine whether a comparison of the transfusion status of those who died after undergoing CABG could provide insight into the transfusion–outcome relationship not previously available. Our findings describe a number of differences in the clinical course and outcomes between the two patient groups, which may help explain why many patients who die after undergoing CABG are transfused while others are not. We believe these results challenge what has become the conventional wisdom that avoiding transfusion will improve results after CABG. Rather, it seems more reasonable to suggest that avoiding the clinical circumstances that lead to transfusion may improve survival. Finally, our attempt to better understand the relationship between transfusion and outcome should not be interpreted as an endorsement of transfusion or a recommendation to increase blood utilization in cardiac surgery. We continue to believe that RBC transfusion is often unnecessary, remains at least potentially “injurious” [12], and should therefore be administered cautiously.

Although a majority of patients who died after undergoing CABG received blood transfusion, significant differences in PROM and the postoperative course leading to death between those with and without transfusion suggest the role of transfusion may be secondary to other patient-related factors. Recognizing that the association between transfusion and outcome after CABG remains incompletely understood, these findings suggest a complex interaction of many other patient-, physician-, and institutional process–related variables not easily captured or accounted for in most clinical databases.

PAONE ET AL RBC TRANSFUSION AND MORTALITY

10.

11.

12. 13.

The MSTCVS Quality Collaborative recognizes the support of Blue Cross Blue Shield of Michigan and Blue Care Network.

14.

References

15.

1. Engoren MC, Habib RH, Zacharias A, Schwann TA, Riordan CJ, Durham SJ. Effect of blood transfusion on longterm survival after cardiac operation. Ann Thorac Surg 2002;74:1180–6. 2. Koch CG, Li L, Duncan AI, et al. Morbidity and mortality risk associated with red blood cell and blood–component transfusion in isolated coronary artery bypass grafting. Crit Care Med 2006;34:1608–16. 3. Moskowitz DM, McCullough JN, Shander A, et al. The impact of blood conservation on outcomes in cardiac surgery: is it safe and effective? Ann Thorac Surg 2010;90:451–9. 4. Paone G, Brewer R, Theurer PF, Bell GF, Cogan CM, Prager RL. Preoperative predicted risk does not fully explain the association between red blood cell transfusion and mortality in coronary bypass surgery. J Thorac Cardiovasc Surg 2012;143:178–85. 5. Prager RL, Armenti F, Bassett J, et al. Cardiac surgeons and the quality movement: the Michigan experience. Semin Thorac Cardiovasc Surg 2009;21:20–7. 6. Shannon FL, Fazzalari FL, Theurer PF, Bell GF, Sutcliffe KM, Prager RL, Michigan Society of Thoracic and Cardiovascular Surgeons. A method to evaluate cardiac surgery mortality: phase of care mortality analysis. Ann Thorac Surg 2012;93: 36–43. 7. Paone G, Likosky DS, Brewer R, et al. Transfusion of 1 and 2 units of red blood cells is associated with increased morbidity and mortality. Ann Thorac Surg 2014;97:87–94. 8. Karkouti K, Wijeysundera DN, Yau TM, et al. Acute kidney injury after cardiac surgery: focus on modifiable risk factors. Circulation 2009;119:495–502. 9. Murphy GJ, Reeves BC, Rogers CA, Rizvi SIA, Culliford L, Angelini GD. Increased mortality, postoperative morbidity,

16. 17.

18.

19.

20.

21. 22.

23.

1589

and cost after red blood cell transfusion in patients having cardiac surgery. Circulation 2007;116:2544–52. LaPar DJ, Crosby IK, Ailawadi G, et al. Blood product conservation is associated with improved outcomes and reduced costs after cardiac surgery. J Thorac Cardiovasc Surg 2013;145:796–804. Surgenor SD, Kramer RS, Olmstead EM, et al. The association of perioperative red blood cell transfusions and decreased long-term survival after cardiac surgery. Anesth Analg 2009;108:1741–6. Shander A, Goodnough LT. Can blood transfusion be not only ineffective, but also injurious? Ann Thorac Surg 2014;97: 11–4. McEvoy MT, Shander A. Anemia, bleeding, and blood transfusion in the intensive care unit: causes, risks, costs, and new strategies. Am J Crit Care 2013;22:eS1–14. Rodriquez RM, Corwin HL, Gettinger A, Corwin MJ, Gubler D, Pearl RG. Nutritional deficiencies and blunted erythropoietin response as causes of the anemia of critical illness. J Crit Care 2001;16:36–41. van Iperen CE, Gaillard CAJM, Kraaijenhagen RJ, Braam BG, Marx JJM, van de Wiel A. Response of erythropoiesis and iron metabolism to recombinant human erythropoietin in intensive care unit patients. Crit Care Med 2000;28:2773–8. Corwin HL, Krantz SB. Anemia of the critically ill: “acute” anemia of chronic disease. Crit Care Med 2000;28:3098–9. Habib RH, Zacharias A, Schwann TA, et al. Role of hemodilutional anemia and transfusion during cardiopulmonary bypass in renal injury after coronary revascularization: implications on operative outcome. Crit Care Med 2005;33: 1749–56. Swaminathan M, Phillips-Bute BG, Conlon PJ, Smith PK, Newman MF, Stafford-Smith M. The association of lowest hematocrit during cardiopulmonary bypass with acute renal injury after coronary artery bypass surgery. Ann Thorac Surg 2003;76:784–92. Surgenor SD, De Foe GR, Fillinger MP, et al. Intraoperative red blood cell transfusion during coronary artery bypass graft surgery increases the risk of postoperative low-output heart failure. Circulation 2006;114(Suppl 1):I43–8. Engoren M, Schwann TA, Habib RH, Neill SN, Vance JL, Likosky DS. The independent effects of anemia and transfusion on mortality after coronary artery bypass. Ann Thorac Surg 2014;97:514–20. Yun JT, Helm RE, Kramer RS, et al. Limited blood transfusion does not impact survival in octogenarians undergoing cardiac operations. Ann Thorac Surg 2012;94:2038–45. Salisbury AC, Reid KJ, Marso SP, et al. Blood transfusion during acute myocardial infarction: association with mortality and variability across hospitals. J Am Coll Cardiol 2014;64: 811–9. Jin R, Zelinka ES, McDonald J, Byrnes T, Grunkemeier GL, Brevig J. Effect of hospital culture on blood transfusion in cardiac procedures. Ann Thorac Surg 2013;95:1269–75.

DISCUSSION DR ALAN M. SPEIR (Falls Church, VA): It is my privilege to discuss this well-presented, thoughtful, and provocative paper, and I particularly would like to thank Dr Paone for providing me the manuscript in such a timely fashion. We are all familiar with the leading clinical accomplishments of the Michigan Society of Thoracic and Cardiovascular Surgeons Quality Collaborative. This presentation is certainly one of them. Additionally, for those not familiar with their Phase of Care Mortality Analysis, or POCMA, I would commend to you their groundbreaking mortality assessment that identifies a seminal event or death trigger that characterizes the timing of such an event into a preoperative, intraoperative, or postoperative phase

of care in order to determine the root cause that ultimately leads to death after cardiac surgery. This paper examines 34,362 patients undergoing myocardial revascularization in Michigan between January of 2008 and September of 2013. Six hundred seventy-two patients, or 2%, died and serve as the basis for this study. As presented, 84% were transfused. The transfused patients were older, more female, more emergencies, redos, on-pump, and lower preoperative and on bypass nadir hematocrits. Postoperatively, the transfused patients were ventilated longer, had more renal failure, dialysis, multisystem organ failure, total ventilation hours, reintubation, total ICU (intensive care unit) hours, and

ADULT CARDIAC

Ann Thorac Surg 2015;99:1583–90

ADULT CARDIAC

1590

PAONE ET AL RBC TRANSFUSION AND MORTALITY

overall length of stay. These higher rates of complication were present both within the total population examined and particularly the predicted risk of mortality matched population. The statistical analysis of these clinical findings, however, led the authors to suggest that “these patients died in spite of blood transfusion, not because of it” and “the role of transfusion may be secondary to other patient-related factors.” This study, which did not primarily link transfusion to adverse postoperative outcomes, is in contradistinction with the analysis we performed in our region through the Virginia Cardiac Surgery Quality Initiative. We have previously documented an analysis of 14,259 patients who underwent nonemergency, primary isolated myocardial revascularization. These patients were stratified according to two guidelines eras: preguideline and postguideline. Primary outcomes of interest were observed differences in postoperative events and mortality risk adjusted associations. In the postguideline era, intraoperative and postoperative transfusion was significantly reduced. Patients in the postguideline era demonstrated reduced morbidity with decreased pneumonia, prolonged ventilation, renal failure, new-onset hemodialysis, and composite incidence of major complications. Of particular note, after mortality risk adjustment, operations performed in the postguideline era were associated with a 47% reduction in major complications and mortality. Mortality complications were all significantly increased after intraoperative and postoperative transfusion. I raise these alternative results in order to stimulate further dialogue and scrutiny of transfusion practices and not to challenge methodologies or biostatistical analysis. With these thoughts in mind, I raise several questions. How do you reconcile these two seemingly discrepant studies? In your mind and not just as an advocate for your paper, is transfusion in cardiac surgery patients merely a surrogate for sicker patients or is transfusion the cause of adverse outcomes for mortality and morbidity? Did predicted risk of mortality fully adjust for important group differences? With the exception of marginal differences in weight and last preop[erative] hematocrit, preoperative variables were pretty evenly distributed. However, the death at discharge clearly favored the nontransfused. Have you changed your clinical transfusion practices at the Henry Ford Hospital by the result of this study? By extension, were any transfusion protocols in effect during this examination? Has there been any interest within the Michigan Quality Collaborative to adopt a transfusion protocol among participating cardiac surgical centers because of the concerns for increased mortality and morbidity in the transfused population? Thank you again for the opportunity to discuss your excellent presentation, and we all look forward to future contributions from the Michigan Quality Collaborative. DR PAONE: Thank you, Dr Speir, for your comments, for your questions, and I will also quickly add thank you for your role in helping us to develop an interest and advance our progress in this area. You will recall that early in 2009, you came to our collaborative meeting and gave a very nice presentation on blood transfusion and blood conservation, and I think that was very helpful in us moving towards a greater interest in this topic and improving our transfusion outcomes. In response to your first question, I would start by saying that advocacy for the paper should not be misinterpreted as advocacy for blood transfusion. We continue to believe that there is plenty of evidence that suggests that transfusion, whether or not you

Ann Thorac Surg 2015;99:1583–90

believe it is potentially harmful, is certainly overused and very often unnecessary. Having said that, I actually have come to believe that transfusion, at least in terms of the data analysis, is in fact a surrogate for a sicker patient population, and that has been proven by just about every report that has been presented, including from your center, our collaborative, and many others. When you compare the demographics of transfused patients to nontransfused patients, invariably the transfused population has a greater incidence of many comorbidities, lower hematocrits, and higher predictive risk of mortality, as we saw even here. And so I think in that sense transfusion is clearly a surrogate marker for a sicker population, independent of whether or not that then portends a poorer outcome specifically related to the transfusion. I would also suggest that the difference between the Virginia paper and what we have presented today isn’t necessarily all that discordant. They are two different analyses done in two different manners. You looked at comparing transfused populations to nontransfused populations, as many studies have, and propensity matched the two populations. One of the criticisms of propensity matching is that while matching the preoperative demographics, it does not account for the actual reasons that patients then go on to get transfused. That was one of the reasons why we chose to approach this analysis in the manner that we did. We started by looking at a large population of patients and filtered out those that died so as to compare transfused deaths versus nontransfused deaths. Given that all patients have died, we then compared the postoperative clinical courses. We found that there were significant differences between the two groups that portended longer follow-ups, longer hospital stays, and more comorbidities in the transfused cohort. And so rather than having the analysis suggest that by decreasing the incidence of transfusion going forward you will then improve outcomes, our suggested analysis is to say that if you can eliminate the clinical circumstances that lead to transfusion, you will then see improved outcomes. DR KEVIN D. ACCOLA (Orlando, FL): I appreciate your comments as well as Dr Speir’s commentary and discussion. I think we have all become more discretionary with utilization of blood cells, but in your analysis did you recognize any trends with regards to component therapy? Was there any association with red blood cell transfusions associated with component therapy and if so did that ha[ve] any major impact on your results? Have you developed any different strategies in your component therapy utilization in these patients? DR PAONE: I’m sorry, Kevin, I’m not sure I understood the question. DR ACCOLA: Yes, in regards to blood component therapy utilization of platelets, FFP (fresh frozen plasma), et cetera, associated with red blood cell transfusions as this may be a significant consideration in your results and conclusions. DR PAONE: We did not specifically consider the additive effect of other components. We did look at the percentage of patients who were transfused platelets, plasma, and cryoprecipitate along with red cells, but because we did not want to further complicate our intention of looking at red cell transfusion specifically we did not separately account for other component therapy in the analysis. I suspect if you considered the effect of additional component therapy, you would further document a patient group at even higher risk, with higher predicted risk, longer lengths of stay, and more complications.