Aprotinin for coronary bypass operations: Efficacy, safety, and influence on early saphenous vein graft patency

Aprotinin for coronary bypass operations: Efficacy, safety, and influence on early saphenous vein graft patency A multicenter, randomized, double-blind, placebo-controlled study The purpose of this study was to evaluate the efficacy and safety of aprotinin in a U.S. population of patients undergoing coronary artery bypass grafting. Early vein graft patency rates were assessed by ultrafast computed tomography. A total of 216 patients at five centers were randomized to receive either high-dose aprotinin or placebo during the operation; 151 patients underwent primary operation, and 65 underwent repeat procedures. Total blood product exposures in the primary group were 2.2 per patient receiving aprotinin as compared with 5.7 per patient receiving placebo (p = 0.010). The repeat group had 0.3 exposures per patient receiving aprotinin as compared with 10.7 per patient receiving placebo (p = <0.001). Consistent reductions in the percent of patients requiring donor red blood ceUs and in the number of units of platelets, fresh frozen plasma, and cryoprecipitate required were associated with the use of aprotinin in both primary and repeat groups. Mortality was 5.6 % in the aprotinin group and 3.7% in the placebo group (p = 0.517). In the primary group, clinical diagnoses of myocardial infarction were made in 8.9 % of patients receiving aprotinin as compared with 5.6 % of the patients receiving placebo (p = 0.435). In the repeat group, infarctions occurred in 10.3 % of patients receiving aprotinin and 8.3% of patients receiving placebo (p = 1.000). Secondary analysis of electrocardiograms and available enzyme data showed no significant difference in infarction rates between the treatment groups. There was no difference in c1inicaUy significant renal dysfunction. The early vein graft patency rates were 92.0 % in the aprotinin group and 95.1 % in the placebo group (p = 0.248). In this study, aprotinin was effective in reducing bleeding and blood product transfusion rates, and its use was not associated with an increase in complications. An adverse effect on early vein graft patency rates was not demonstrated, but the number of grafts assessed was insufficient for absolute conclusions in this regard. (J THORAC CARDIOVASC SURG 1994;107:543-53)

John H. Lemmer, Jr., MDa (by invitation), William Stanford, MD,b Sharon L. Bonney, MD c (by invitation), Jerome F. Breen, MDd (by invitation), Eva V. Chomka, MDe (by invitation), W. Jay Eldredge, MDf (by invitation), William W. Holt, MDg (by invitation), Robert B. Karp, MD,h Glenn W. Laub, MDi (by invitation), Martin J. Lipton, MDi (by invitation), Hartzell V. Schaff, MD,k Constantine J. Tatooles, MD,l and John A. Rumberger, PhD, MDm (by invitation), Iowa City, Iowa, West Haven, Conn., Rochester, Minn., Chicago, Ill., and Browns Mills, N.J.

From the Departments of Surgery" and Radiology," The University of Iowa College of Medicine, Iowa City. Iowa; Miles Inc., West Haven, Conn,"; Department of Diagnostic Radiology," Department of Cardiovascular Diseases,m and Section of Cardiovascular Surgery.! Mayo Clinic, Rochester, Minn.; Departments of Medicine" and Surgery,' University of Illinois at Chicago, Chicago, Ill.; Departments of Cardiology! and Surgery,' Deborah Heart and Lung Center, Browns Mills, N.J.; and Departments of Radiologys-! and Surgery," The University of Chicago, Chicago, Ill.

Statistician: Lawrence A. Schwartz, MS, Associate Director of Statistics, Statistics and Data Systems Department, Miles Inc., West Haven, Conn.

Supported by Miles Inc., West Haven, Conn.

0022-5223/94 $\.00 + .10

Consulting Statistician: Gary L. Grunkemeier, PhD, Portland, Ore. Read at the Seventy-third Annual Meeting of The American Association for Thoracic Surgery, Chicago, Ill., April 25-28, 1993. Address for reprints: John H. Lemmer, Jr., MD, Northwest Surgical Associates, 2226 NW Pettygrove, Portland, OR 97210. Copyright

:s 1994 by Mosby-Year

Book, Inc.

12/6/51763

543

544

Lemmer et al.

The risks associated with blood transfusion have generated great interest in methods, including the use of pharmacologic agents, to limit patient exposure to homologous blood products. In patients undergoing cardiac operations, intraoperative administration of aprotinin decreases postoperative blood loss and the need for blood product transfusion.l" Used extensively in Europe, aprotinin is an investigational drug in the United States, and only a few reports from this country have described experience with this drug." 8 Although aprotinin has been shown to be effective in reducing postoperative bleeding in these previous studies, concern has been raised that, as an adverse consequence of this effect, aprotinin might contribute to complications related to early thrombosis of bypass grafts; however, data in this regard have been conflicting.I-? The primary objective of this study was to evaluate the efficacy and safety of high-dose aprotinin in a placebo-controlled population of patients in the United States undergoing surgical myocardial revascularization. As an adjunct to the safety analysis, we used ultrafast computed tomography (CT) to evaluate early postoperative vein graft patency. Methods Patient characteristics. Patients undergoing isolated primary or repeat (redo) coronary artery bypass grafting (CABG) with cardiopulmonary bypass were enrolled in this study. A total of 216 patients at the fiveparticipating centers were randomized to receive either high-dose aprotinin or placebo during the operations. Double-blind conditions were maintained throughout the study. Enrolled patients were stratified as to whether they were undergoing primary (151 patients) or redo (65 patients) procedures. In addition, patients were further stratified with regard to whether they had received aspirin (within 5 days) or a nonsteroidal antiinflammatory drug (within 3 serum half-lives of the drug) before the operation. All 216 patients who received the study drug fulfilled the criteria for safety analysis; 196 fulfilled the criteria for efficacy analysis. Criteria for patient elimination from the efficacy analysis are listed in Appendix I. Of the 151 patients undergoing primary CABG, 141 (74 in theaprotinin-treated group and 67 in the placebo-treated group) fulfilled the criteria for efficacy evaluation. Of the 65 patients undergoing redo CABG, 55 (23 in the aprotinin-treated group and 32 in the placebo-treated group) fulfilled the criteria for efficacy evaluation. Patients were eliminated from efficacy analysis before the random code was broken. Study drug administration. Aprotinin (Bayer AG, Leverkusen, Germany) was supplied in a concentration of 1.4 mg/rnl (10,000 KIU/mI) in 0.9% sodium chloride solution without further additives or preservatives. An identically appearing placebo (0.9% sodium chloride solution) was also supplied. Patients randomized to receive aprotinin were given a loading dose of 280 mg, a continuous intravenous infusion of 70 rng/hr, and a dose of 280 mg added to the oxygenator prime solution. Patients in the placebo group received identical volumes of 0.9% sodium chloride solution. The continuous infusion was discontinued on the patient's arrival in the intensive care unit.

The Journal of Thoracic and Cardiovascular Surgery February 1994

Surgical technique, anticoagulation, blood conservation methods, and blood replacement policy. Surgical details of the CABG operation and perioperative aspirin administration were performed according to the usual protocol of the participating surgeons and centers. Anticoagulation during cardiopulmonary bypass was managed either by direct measurement of the heparin concentration in the patient's blood or by use of a fixed-dose heparin regimen. In the first method, the heparin loading dose was calculated before the induction of anesthesia by means of the heparin dose response test on the System Four instrument (Medtronic Hemotec, Englewood, Colo.). Additional heparin doses were determined by means of the heparin protamine titration test on the same instrument. 10 Likewise, the Hepcon System (Henotics Inc., Englewood, Colo.) was used to determine the appropriate dose of protamine to be administered after the termination of cardiopulmonary bypass. By means of the fixed-dose method, a loading heparin dose of 300 USP units /kg was administered before cannulation of the heart, and an additional dose of 150 USP units/kg was given after 90 minutes of cardiopulmonary bypass, if the period of extracorporeal circulation exceeded 90 minutes. Protamine was given after discontinuation of bypass at a dose of 1.3 mg/ I00 IU heparin administered. Intraoperatively, blood from the operative field was salvaged, processed, and reinfused in all but one center. The contents of the oxygenator were returned to the patient after the termination of bypass in all centers. In all but one center, in the postoperative period, blood collected in drainage receptacles was filtered and reinfused at specific intervals, if the quantity of shed blood was sufficient. During cardiopulmonary bypass, homologous blood was transfused if the patient's hematocrit value was less than 18% or if the clinical condition of the patient required it. Postoperatively, homologous blood was transfused if the hematocrit value was less than 21% or if the clinical condition of the patient required it. Other blood products were transfused as judged necessary by the investigator. Myocardial infarction diagnosis. In this study, no uniform definition was used for what would be classified by the investigators as a perioperative myocardial infarction. The decision to record the clinical diagnosis of perioperative myocardial infarction was the responsibility of the investigator at each individual site and was left to his or her judgment. However, because of the double-blinded nature of this study, these judgments were made without knowledge of the identity of the study drug that had been administered to the patient. So that a consistent assessment of perioperative myocardial infarction could be obtained, data from this study were submitted in a blinded fashion to the Core Electrocardiography (ECG) Laboratory at St. Louis University, headed by Bernard Chaitman, MD. The laboratory reviewed results of serum enzyme determinations and ECGs from all patients in whom a myocardial infarction or myocardial ischemia was suspected or in whom an abnormal elevation of the creatine kinase MB fraction had occurred. In this blinded review, the ECG Core Laboratory analyzed enzymatic data, clinical summaries, and preoperative and postoperative ECGs. Patients were then assigned to the following categories: (I) definite perioperative myocardial infarction as defined by a significant new Q wave, (2) definite or probable myocardial infarction on the basis of any and all information supplied, (3) definite, probable. or possible myocardial infarction on the basis of any and all information supplied, or (4) no myocardial infarction.

The Journal of Thoracic and Cardiovascular Surgery Volume 107, Number 2

Graft patency analysis. For this study, ultrafast CT (Imatron, Inc., South San Francisco, Calif.) was used in an effort to compare graft patency rates between treatment groups. I I The accuracy of this technique for vein graft patency determination, as validated by graft angiography, is 92% to 95%.12.14 By means of this protocol, CABG patency was assessed before discharge or at a follow-up visit no later than 60 days after the operation. The ultrafast CT images were first read at each center by the experienced cardiologist or cardiac radiologist (or both) participating in the study. At a later date, group meetings were held during which each of the scans was reviewed in a blinded fashion by five to seven of the participants. Technically acceptable scans were analyzed with regard to whether each graft either was patent or was not visualized. For a graft to be judged patent, the investigators agreed that the graft in question was to be visualized at two or more levels in the distribution of the vessel bypassed. Because of the inability of ultrafast CT scanning to visualize sequential branches and distal anastomoses as a result of inadequate scanner resolution, patency assessments were based on the proximal portion of the conduits soon after their exit from the aorta. Any graft that did not meet the criteria specified in the protocol for being read as patent was read as closed. Saphenous vein graft patency and internal mammary artery (IMA) graft patency were assessed separately; however, the patency results in patients undergoing primary and redo CABG operations were combined. In patients having repeat CABG, only the newly placed grafts were analyzed. Statistical methods. The methods used for determination of study sample size and data analysis are described in Appendix 2.

Results Demographic and surgical variables. Appendix 3 shows demographic variables for the 196 patients who met the criteria for efficacy analysis. No significant differences were detected between the treatment groups. Other factors, including percent with history of myocardial infarction, time from last myocardial infarction, number of previous sternotomies (for patients having redo operations), preoperative hemoglobin level, and baseline prothrombin and partial thromboplastin times were also similar. Surgical variables were similar between the treatment arms in both the primary and the redo groups (Appendix 4). There were no significant differences in these variables among those patients who underwent primary CABG. In the redo group, patients who received aprotinin had a longer mean time of total bypass and a trend toward higher frequency of IMA usage than those who received placebo. Membrane oxygenators were used in 143 (78.1%) of the 196 patients who met the criteria for efficacy analysis (72 who received aprotinin and 71 who received placebo); bubble oxygenators were used in the remainder. Blood cardioplegia was used in 68 of 97 (70.1%) patients receivingaprotinin and 66 of 99 (66.7%) receiving placebo. The remainder, with one exception, received crystalloid cardioplegia. In the one exception,

Lemmer et al.

54 5

deep hypothermia and fibrillatory arrest were used because of difficulties exposing the aorta as a result of adhesions. Blood product exposures, transfusions, and bleeding. The total number of blood product exposures for each treatment group is compared in Table I. Total number of exposures was the sum of the number of units of packed red blood cells, platelets, fresh frozen plasma, and cryoprecipitate administered to each patient. In both the primary and redo CABG groups, treatment with aprotinin significantly decreased exposure to transfused blood products. The number (and percentage) of patients who required donor red cell transfusion is shown in Table I. In both the primary and redo groups, treatment with aprotinin was associated with a smaller number of patients who required any transfusion of donor red blood cells. The mean number of red blood cell units, mean volume of donor blood cells, mean number of units of platelets, fresh frozen plasma, and cryoprecipitate transfused are also shown in Table I. For each of these variables, except for units of cryoprecipitate (p = 0.146) in redo operations, the use of aprotinin was associated with a significant decrease in transfusion requirements. The mean hourly rate of thoracic drainage during the first 8 hours after the operation and total postoperative thoracic drainage volume (from surgery until removal of the drainage tubes) were lower in those patients who received aprotinin, in both the primary and redo groups (Table II). Chest closure times in the redo operations were significantly shorter in the patients who received aprotinin than in those who received placebo, but this difference was not seen in patients undergoing primary operation. Three of the 216 (1.4%) patients in the study required exploration for bleeding during the early postoperative period. One of these patients had received aprotinin, and the other two had received placebo. All three were found to be bleeding from specific surgically correctable sites. These three patients were excluded from efficacy analysis. No patient in either treatment group required exploration for bleeding not of surgical origin (i.e., for diffuse bleeding). Miscellaneous efficacy parameters. There was no significant difference between the treatment groups (in either the primary or redo strate) regarding the change in bleeding time from before to after the operation, the maximum decrease in platelet count from the preoperative value, the mean number of days in the intensive care unit, or length of hospital stay. The mean preoperative hemoglobin level in each treatment group was similar (Appendix 3). Despite the low number of red blood cell units received by the patients in the aprotinin group, the predischarge mean serum hemoglobin level was also similar between the treatment groups (as evidenced by com-

546

The Journal of Thoracic and Cardiovascular Surgery Febru ary 1994

Lemmer et al.

Table I. Total blood product exposures Primary CABG

Mean total No. of blood product exposures per patient (SE) No. (%) who required donor RBCs Mean No. of RBC units transfused (SE) Mean volume (rnl) donor blood cells transfused (SE) Mean No. of platelet units transfused (SE) Mean No. of units of fresh frozen plasma transfused Mean No. of units of cryoprecipitate transfused

Redo CABG

Aprotinin (n = 74)

Placebo (n =67)

p Value

2.2 (0.8)

5.7 (0.8)

0.010

28 (38%)

35 (52%)

0.052

1.1 (0.3)

2.1 (0.3)

0.025

0.4 (0.8)

3.3 (0.7)

<0.001

362 (88)

606 (93)

0.023

164 (199)

931 (173)

0.005

0.8 (0.4)

1.9 (0.4)

0.044

0.4 (1.4)

4.1 (1.2)

0.017

0.3 (0.1)

0.7 (0.2)

0.002

0.4 (0.2)

2.3 (0.6)

0.004

0(0)

0.9(0.3)

<0.001

0.4 (0.4)

2.2 (1.0)

0.146

Aprotinin (n = 23)

Placebo (n =32)

p Value

0.3 (3.6)

10.7 (3.2)

<0.001

7 (30%)

23 (72%)

0.001

SE. Standard error; RBC. red blood cell. Statistical methods! and ;?,; see Appendix 2.

Table II. Thoracic drainage rates, chest closure time, and hemoglobin decrease Primary CABG Parameter Total thoracic drainage volume (rnl) (SE) Mean rate of drainage (rnl/hr) (SE) Mean closure time (min) (SE) Mean hemoglobin decrease (preoperative to discharge) (gmjdl) (SE) SE. Standard error. Statistical method

!1; see Appendix

Redo CABG

Aprotinin (n= 74)

Placebo (n = 67)

P Value

Aprotinin (n = 23)

Placebo (n = 32)

p Value

855 (78) 41 (7) 56.7 (2.3) 2.7 (0.2)

1503 (82) 87 (8) 62.7 (2.4) 2.8 (0.2)

<0.001 <0.001 0.075 0.720

1225 (326) 43 (12) 55.6 (7.0) 2.4 (0.4)

1979 (283) 78 (I I) 80.5 (6.1) 2.6 (0.4)

0.037 0.011 0.002 0.603

2.

parable mean decreases from the preoperative hemoglobin level; Table II). Mortality. Tenofthe 216 patients enrolled in the study died during or early after operation (overall mortality, 4.6%). Six (5.6%) of the 108 patients in the aprotinin group and four (3.7%) of the 108 patients in the placebo group died within 30 days after the operation or before discharge from the hospital. There was no significant difference in mortality between the two treatment groups (p = 0.517). Causes of death in both groups were predominantly related to cardiac dysfunction; five of the six deaths in the aprotinin group and three of the four deaths in the placebo group were due to low cardiac output. Autopsies were performed in three of the 10 patients who died in this study. One patient was in the aprotinin group and had required two vein grafts to three distal target sites, and both of the grafts were patent at autopsy. Likewise, all grafts (six) were patent at autopsy in the two patients in the placebo group.

When analyzed by center, operative mortality rates ranged from 3% to 13%. Myocardial infarctions: Clinical diagnosis. The clinical diagnosis of perioperative myocardial infarction was reported by the investigators in 17 (7.9%) of the 216 patients. The overall infarction rate was 7.3% (11/15 I) in the primary group and 9.2% (6/65) in the redo group. Distribution of these patients regarding treatment group and nature of operation (primary or reoperation) isshown in Table III. There was no significant difference in the prevalence of clinically apparent myocardial infarction between the treatment groups in the primary or redo strata. Myocardial infarction: Secondary analysis. The blinded analysis results of the Core ECG Laboratory at St. Louis University are shown in Table III.There was no significant difference in myocardial infarction rates between the two treatment groups. In only one category (definite myocardial infarction in patients having redo

The Journal of Thoracic and Cardiovascular Surgery Volume 107, Number 2

Lemmer et al.

547

Table III. Myocardial infarctions Redo CABG

Primary CABG Infarction classification Clinical diagnosis

Aprotinin In = 79)

7/79

Definite or probable Definite or probable or possible Statistical method

~;

In =

72)

p Value

4/72

0.435

4/79

3/72

1.000

(5.1%)

(4.2%)

(8.9%) Definite

Placebo

(5.6%)

7/79

6/72

(8.9%)

(8.3%)

8/79 (10.1%)

7/72 (10%)

0.908 0.934

Aprotinin In = 29)

Placebo In = 36)

3/29 (10.3%) 2/29 (6.9%) 3/29 (10.3%) 3/29 (10.3%)

3/36 (8.3%) 0/36 (0%) 3/36 (8.3%) 3/36 (8.3%)

p Value

1.000 0.195 1.000 1.000

see Appendix 2.

operations) was any trend toward an increase in infarctions in the aprotinin group apparent. In this category, there were definite infarctions in two of 29 patients in the aprotinin group and no definite infarctions in the 36 patients in the placebo group. The more sensitive category, "definite or probable myocardial infarctions," however, failed to confirm this trend, with three such events occurring in each treatment group. Postoperative complications. There were no statistically significant differences between treatment groups with regard to serum creatinine change, clinically significant postoperative renal insufficiency,need for postoperative hemodialysis, neurologic complications, atrial fibrillation and flutter, ventricular tachycardia and fibrillation, hypotension,and heart failure. Placement of an intraaortic balloon pump or use of another form of mechanical support (during or after the operation) for treatment of ventricular dysfunction was required in nine of 108 patients (8.3%) in the aprotinin group and in 10 of 108 patients (9.2%) in the placebo group. Among the 216 patients, there was one instance of suspected allergic reaction to the study drug. This patient had received placebo. Bypass graft patency. Ultrafast CT scans to assess graft patency were performed 7 to 60 days (mean 27 days) after the operation. Of the 216 patients randomized to the two treatment groups, 170 (78.7%) had scans that were subsequently judged (at the blinded group scan reading sessions) to be technically adequate for graft patency interpretation, based on criteria developed for vein graft assessment. Of those patients who did not have a readable postoperative scan (for whatever reason), 25 were in the aprotinin group and 21 were in the placebo group. Reasons for failure to have a readable postoperative scan were patient refusal (n = 10), patient death (n = 9), technical inadequacy for vein graft patency assessment (n = 8), patient illness (n = 5), inadequate venous access (n = 4), scan lost (n = 3), and miscella-

neous (n = 7). There were substantial differences between the study centers regarding the percent of enrolled patients who subsequently had an evaluable ultrafast CT scan (range 48% to 100%). Vein graft patency was analyzed on both a per-veingraft basis (number of patent vein grafts divided by the total number of vein grafts) and a per-patient basis (number of patients with at least one occluded vein graft divided by the number of patients with evaluable ultrafast CT scans). Of the 216 patients who received either aprotinin or placebo, 164 (75.9%) with vein grafts had ultrafast CT scansjudged at the group readings to be evaluable for vein graft patency determination (six patients had single IMA grafts only). On a per-graft basis, the overall vein graft patency rate in these patients was 93.5% (317/339 grafts). Results according to treatment group are shown in Table IV. Fourteen of the 176 vein grafts in the aprotinin group (7.9%) and eight of the 163 grafts in the placebo group (4.9%) were judged to be closed (p = 0.248). On a per-patient basis, 87.8% (144/164) had all grafts assessed as open at the time of CT scanning. In the aprotinin group, 13 of 83 (15.6%) patients had one or more closed grafts, whereas seven of 81 (8.6%) patients in the placebo group were found to have one or more closed grafts (p = 0.170). Only two patients (one in the aprotinin group and one in the placebo group) had more than one graft closed, and each of these had two grafts that were not visualized by ultrafast CT scanning. IMA graft patency was analyzed by ultrafast CT scans in 137 patients. Six ofthese patients had a single IMA and no vein grafts placed at the time of the operation. Each of the 137 patients had only one IMA graft (there were no patients with two IMA grafts). The overall IMA graft patency rate was 94.2% (129/137). The results of the patency analysis based on treatment group are shown in Table V. Fifteen surgeons performed the CABG procedures in

The Journal of Thoracic and Cardiovascular Surgery February 1994

5 4 8 Lemmer et al.

Table IV. Vein graft patency Mean No. of vein grafts per patient Open grafts/total grafts (per-graft basis) Patients with all grafts open (per-patient basis) C/, 95% confidence interval (statistical methods land

Aprotinin (n = 83)

Placebo (n = 8/)

2.1 162/176 (92.0%) (CI87.0%-95.6%) 70/83 (84.3%) (CI74.7%-91.4%)

2.0 155/163 (95.1%) (CI90.6%-97.9%) 74/81 (91.4%) (CI83.0%-96.5%)

p Value

0.248 0.170

:!; see Appendix 2).

this study. Per-vein-graft patency rates among those participating surgeons who had at least five patients with technically acceptable ultrafast CT scans ranged from 86% to 100%. Regarding IMA grafts, the individual surgeons' patency rates (of those who had at least five patients with readable scans) ranged from 50% to 100%. Five centers participated in this study. On a per-graft basis, the individual center's early saphenous vein graft patency rates ranged from 87% to 95%. For the IMA grafts, the range of patency rates was 64% to 100%. The percentage of IMA grafts among patients who subsequently had evaluable ultrafast CT scans at the different centers was variable, ranging from 75% to 93% of those patients enrolled. Although these differences existed, multiple logistic regression analysis failed to find center as a significant factor for saphenous vein graft closure.

Discussion The effectiveness of aprotinin in decreasing the need for blood transfusion in patients undergoing cardiac procedures has been documented by multiple European reports. 1-6, 9 It has been estimated that by 1991 more than 10,000 patients who had CABG procedures had received the drug at high dose.15 In association with the widespread and frequent use of aprotinin, a clear pattern of serious adverse effects has not emerged. In this study, the use of high-dose aprotinin during primary and repeat CABG procedures markedly decreased the exposure of patients to homologous blood product transfusion. For patients undergoing first-time CABG, the use of aprotinin was associated with a 60% reduction in patient exposures to blood products. For those having a second operation, the reduction in exposures was tenfold. This study therefore confirms the results of earlier studies conducted in Europe, 1-6,9 and the one CABG study from the United States reported by Cosgrove and associates," in establishing the efficacy of aprotinin in improving hemostasis in CABG operations. The frequency of complications of any type reported by investigators in this study was not significantly different in the two treatment groups. This failure to demonstrate a pattern of adverse events associated with the use of

aprotmm is consistent with the European experience, including the single-center results of 902 patients reported by Dietrich and associates.P In our study, there was no difference between treatment groups in the prevalence of myocardial infarction as a clinical diagnosis or by blinded review of available relevant data. The prevalence of "definite" myocardial infarctions in the aprotinin redo group (6.9%) was comparable with the reported rate of this complication in larger series of patients undergoing redo CABG. 16-18 Likewise, in the more sensitive categories of "definite or probable" and "definite or probable or possible" myocardial infarctions, there was no difference between the treatment groups. Thus this study did not identify an increased risk of myocardial infarction in patients who received highdose aprotinin during CABG. One study, conducted by Cosgrove and associates," concluded that aprotinin use may be associated with an increased prevalence of perioperative myocardial infarctions in patients undergoing redo CABG procedures. In that study, using high- and low-dose aprotinin and placebo in 169 patients, nine of 57 patients (15.8%) receiving high-dose aprotinin had Q-wave infarctions as compared with four of 56 patients (7.1%) receiving placebo (p = not significant). One difference in methods between the Cosgrove study and our study was the management of intraoperative anticoagulation. In the former, the patients were initially given 300 USP units of heparin per kilogram of body weight, and additional heparin was given during cardiopulmonary bypass if the activated clotting time (ACT) fell to less than 400 seconds. Data regarding the length of time of cardiopulmonary bypass and total heparin units administered were not reported. In our protocol, the ACT was not used as the measure of anticoagulation during cardiopulmonary bypass. In most patients, blood heparin concentrations were measured and further heparin was administered when indicated on that basis without regard to the patient's measured ACT; that is, even if the ACT was above 400 seconds, if the Hepcon System indicated that further heparin was indicated, it was given to the patient. For patients in whom the heparin concentration was not measured, additional doses

The Journal of Thoracic and Cardiovascular Surgery Volume 107, Number 2

Lemmer et al.

549

Table V. IMA graft patency Parameter

Aprotinin (n = 83)

Placebo (n = 8 I)

p Value

Mean No. of IMA grafts per patient Open grafts/total grafts (per-patient basis)

1.0 67/73 (91.7%) (CI83.0%-96.9%)

1.0 62/64 (96.8%) (CI89.2%-99.6%)

1.000 0.283

C/, 95% confidence interval (statistical method

~;

see Appendix 2.

were administered at fixed intervals, again without regard to the measured ACT. Aprotinin, in the presence of heparin, has been shown to interfere with the accuracy of the commonly used celite-activated ACT (Hemochron; International Technidyne Corp, Edison, N.J.), causing prolongation independent of heparin.f 18-22 Thus it is possible that the use of celite-activated ACT to monitor heparin administration in conjunction with the use of aprotinin might result in inadequate heparinization during cardiopulmonary bypass. Because of these findings, a subsequent report has recommended that an ACT of greater than 750 seconds be maintained.P It is unknown whether this difference in technique might have contributed to the results reported by Cosgrove and colleagues.' The early vein graft patency rates in both treatment groups of this study, 92.0% in the aprotinin group and 95.1o/c in the placebo group, compare favorably with studies in which the patients undergoing CABG (who did not receive aprotinin) underwent early postoperative coronary arteriography. In these studies, early (8 to 60 day) patency rates range from 82% to 94%.24-27 Thus, according to ultrafast CT assessment, aprotinin administration did not appear to have an adverse effect on early vein graft patency as compared with reported angiographically determined graft patency rates. Only one previous investigation has been directed at determining the effect of aprotinin on bypass graft patency. Bidstrup, Underwood, and Sapsford? reported their results in 90 randomized patients in whom magnetic resonance imaging was used to assess patency of 269 total vein grafts at a median of 9 days after operation. In that study, 126 of 131 (96.2%) grafts in patients who received aprotinin were patent as compared with 134 of 138 (97.1%) grafts in the placebo group. There was no difference in graft patency between the two treatment groups. In our study, the proportion of patent vein grafts was lower in the group that received high-dose aprotinin than in the placebo group (92.0% versus 95.1%). The difference in patency rates was not significant and the 95% confidence intervals overlap widely (Table IV). A p value of 0.248 indicates that the observed difference may have been due to chance. The number of grafts analyzed (339) was, however, too small for a definite conclusion regard-

ing the effect of aprotinin on graft patency. If this difference had persisted, more than 1000 total grafts would have been required before a statistically significant difference (with p < 0.05) would have been evident. Evaluation of early IMA graft patency found the percentage of patent IMA grafts to be lower among those patients who received aprotinin than among the placebo group (91.7% versus 96.8%), although this difference did not reach statistical significance (p = 0.283) and the number of IMA grafts (137) was too small for definite conclusions to be drawn. It is important to note several factors regarding the IMA patency data presented in this report. First, validation of the accuracy of ultrafast CT for the determination ofIMA graft patency has not been well demonstrated. The ultrafast CT validation studies cited previously were directed toward vein graft patency. In these previous studies, only 23 total IMA grafts were evaluated. 13, 14 Thus the true sensitivity and accuracy of ultrafast CT for the determination of IMA graft patency are unknown. Second, in the blinded group meetings to review the ultrafast CT scans, the determination of technical adequacy of each scan was based on criteria used to analyze vein graft patency. At centers experienced in using this technique for IMA visualization, it has been found that an increased number of cephalad images are required to confirm the presence of a patent IMA graft. This was not performed in all centers. Thus the scans, although technically adequate for vein graft visualization, were not always suitable for IMA graft visualization. Third, small metallic clips, frequently used during the preparation of the IMA graft, significantly interfere with ultrafast CT graft visualization. Because our protocol specified that any graft not determined with certainty to be open was to be called closed, open grafts may have been misjudged to be closed because of interference from the surgical clips. Fourth, the variation in IMA graft patency rates among participating surgeons was great, and the distribution of evaluable patients was not equal. Despite these caveats, the difference in IMA graft patency rates was not demonstrated to be significantly different between treatment groups in this study. We are reluctant, however, to draw definitive conclusions regarding the effect of aprotinin on IMA graft patency for the reasons cited herein. In the graft patency study of

550

Lemmer et al.

Bidstrup, Underwood, and Sapsford,? all 46 IMA grafts (in the aprotinin and placebo groups) were found to be open at the time of magnetic resonance imaging, approximately I week after operation. Limitations of the study. In retrospect, this study contains several possible weaknesses: 1. Criteria for the clinical determination of perioperative myocardial infarctions were not uniform. Because myocardial infarction was not suspected to be a sideeffect of aprotinin use, this study was not designed to investigate rigorously the frequency of this complication. 2. Technically adequate postoperative ultrafast CT scans were not obtained in some patients. Bias may have been introduced by the fact that graft patency assessments were not obtained in over 20% of the patients and by the considerable center variation in this regard. 3. Ultrafast CThas limitations. Validation of ultrafast CT with coronary arteriography has found its overall reading accuracy for open and closed vein grafts to be 92% to 96%; however, the reading accuracy for severely stenotic (but not totally occluded) vein grafts is lower, about 78%.12 Because the prevalence and distribution of partially obstructed vein grafts in this study is not known, this lower reading accuracy could be of importance. 4. Ultrafast CT findings regarding IMA graft patency determinations were not reinforced angiographically. 5. The number of grafts was insufficient. This study was designed on the basis of sample size calculations regarding aprotinin efficacy, not on graft patency determinations. Although the difference in graft patency rates observed between treatment groups was small and did not reach statistical significance, the study did not have sufficient power to detect a difference of the observed magnitude in occlusion rates at the a level of 0.10 that was specified in the protocol as the level of significance that would be used in the graft patency analysis.

Conclusions On the basis of the results of this study of a U.S. population of patients undergoing CABG, we offer the following conclusions: 1. Aprotinin was highly efficacious in reducing patient exposure to transfused homologous blood products and in reducing postoperative bleeding. This reduction was particularly striking among those patients undergoing repeat operations. 2. The use of aprotinin for CABG operations was safe. It was not associated with a significant increase in perioperative complications, including myocardial infarction, clinically evident renal toxicity, or any other apparent thrombotic events. 3. An increase in early vein graft closure was not

The Journal of Thoracic and Cardiovascular Surgery February 1994

demonstrated. Although a weak trend in this direction may have been present, the number of grafts was insufficient to allow for absolute conclusions to be drawn, and further investigations in this regard may be indicated. If such an effect is present, it would appear to be weak, inasmuch as the vein graft patency rate of the aprotinintreated group was as high as or higher than those reported in patients who did not receive the drug in angiographically controlled studies. 4. Although no definite effect was demonstrated, conclusions from this study regarding the effect of aprotinin on IMA graft patency may not be valid because the ability of ultrafast CT to determine IMA graft patency has not been confirmed. Participating surgeons: Mark S. Adkins, MD, Douglas M. Behrendt, MD, Javier Fernandez, MD, Vincent A. Kucich, MD, Bryan K. Lee, MD, Lynn B. McGrath, MD, and James Robert F. Yario, MD. Study Coordinators: Bridget M. Bailey, Christopher M. Glenn, Rosalia C. Gonzalez, Deborah A. Ilkowski, Marylin Krachmer, Judy Roebbeke, and Sara J. Vance. We also wish to acknowledge the important contributions of Melvin L. Marcus, MD (deceased, 1989), who was instrumental in the development of this investigation.

REFERENCES 1. Royston D, Taylor KM, Bidstrup BP, Sapsford RN. Effect of aprotinin on need for blood transfusion after repeat open heart surgery. Lancet 1987;2:1289-91. 2. Bidstrup BP, Royston D, Sapsford RN, Taylor KM. Reduction in blood loss and blood use after cardiopulmonary bypass with high dose aprotinin (Trasylol). J THORAC CARDIOVASC SURG 1989;97:364-72. 3. Dietrich W, Spannagl M, Jochum M, et al. Influence of high-dose aprotinin treatment on blood lossand coagulation patterns in patients undergoing myocardial revascularization. Anesthesiology 1990;73:1119-26. 4. van Oeveren W, Harder MP, Roozendaal KJ, Eijsman L, Wildevuur CRH. Aprotinin protects platelets against the initial effect of cardiopulmonary bypass. J THORAC CARDIOVASC SURG 1990;99:788-97. 5. Harder MP, Eijsman L, Roozendaal KJ, van Oeveren W, Wildevuur CRH. Aprotinin reduces intraoperative and postoperative blood loss in membrane oxygenator cardiopulmonary bypass. Ann Thorac Surg 1991;51:936-41. 6. Dietrich W, Barankay A, Hahnel C, Richter JA. High-dose aprotinin in cardiac surgery: three years' experience in 1784 patients. J Cardiovasc Vase Anesth 1992;6:324-7. 7. Cosgrove DM III, Heric B, Lytle BW, et al. Aprotinin therapy for reoperative myocardial revascularization: a placebo-controlled study. Ann Thorac Surg 1992;54: 1031-8. 8. D'Arnbra MN, Akins CW, Blackstone EH, et al. Aprotinin in primary cardiac valve replacement reduces bleeding, increases creatinine. Circulation I992;86(Suppl):I495. 9. Bidstrup BP, Underwood SR, Sapsford RN. Effect of

The Journal of Thoracic and Cardiovascular Surgery Volume 107, Number 2

aprotinin on aorta-coronary bypass graft patency, J THORAC CARDIOVASC SURG 1993;105:147-53, 10. BowieJE, Kemna GO. Automated management of heparin anticoagulation in cardiovascular surgery. Proc Am Acad Cardiovasc Perf 1985;6:1-10. II. Boyd DP, Couch JL, Napel SA, Peschmann KP, Rand RE. Ultrafast cine-CT for cardiac imaging. Where have we been? What lies ahead? Am J Card Imag 1987;1:175-85. 12. Stanford W, Rooholamini SA, Rumberger J, et al. Evaluation of coronary bypass graft patency by ultrafast computed tomography. J Thorac Imag 1988;3:52-5. 13. Bateman TM, Gray RJ, Whiting JS, et al. Prospective evaluation of ultrafast cardiac computed tomography for determination of coronary bypass graft patency. Circulation 1987;75:1018-24. 14. Stanford W, Brundage BH, MacMillan R, et al. Sensitivity and specificityof assessing coronary bypass graft patency with ultrafast computed tomography: results of a multicenter study. J Am Coil CardioI1988;12:1-7. 15. Royston D. High dose aprotinin therapy: a review of the first five years' experience. J Cardiothorac Vase Anesth 1992;6:76-100. 16. Lytle BW, Loop FD, Cosgrove OM, et al. Fifteen hundred coronary reoperations: results and determinants of early and late survival. J THORAC CARDIOVASC SURG 1987; 93:847-59. 17. Salomon NW, Page US, Bigelow JC, Krause AH, Okies JE, Metzdorff MT. Reoperative coronary surgery: comparative analysis of 6591 patients undergoing primary bypass and 508 patients undergoing reoperative coronary artery bypass. J THORAC CARDIOVASC SURG 1990;100: 250-60. 18. Foster ED. Reoperation for coronary artery disease. Circulation (Suppl V) I985;72(Suppl):59-64. 19. WangJ-S, LinC-Y, HungW-T, Thisted RA, KarpRB. In vitro effects of aprotinin on activated clotting time with different activators. J THORAC CARDIOVASC SURG 1992;104: 1135-40. 20. Wang J-S, Lin C-Y, Hung W-T, Karp RB. Monitoring of heparin-induced anticoagulation with kaolin-activated clotting time in cardiac surgical patients treated with aprotinin. Anesthesiology 1992;77:1080-4. 21. Royston 0, Bidstrup BP, Taylor KM, Sapsford RN. Reduced blood loss following open heart surgery with aprotinin (Trasylol) is associated with an increase in intraoperative activated clotting time. J Cardiothorac Anesth 1989;3(Suppl 1):80. 22. Najman OM, Walenga JM, Fareed J, Pifarre R. Effects of aprotinin on anticoagulant monitoring: implications in cardiovascular surgery. Ann Thorac Surg 1993;55:662-6. 23. Hunt BJ, Segal H, Yacoub M. Aprotinin and heparin monitoring during cardiopulmonary bypass. Circulation I992;86(Suppl):II41 0-2. 24. Goldman S, Copeland J, Moritz T, et al. Improvement in early saphenous vein graft patency after coronary artery bypass surgery with antiplatelet therapy: results of a Veterans Administration Cooperative Study. Circulation 1988; 77:1324-32.

Lemmer et al.

55I

25. Goldman S, Copeland J, Moritz T, et al. Starting aspirin therapy after operation: effects on early graft patency. Circulation 1991;84:520-6. 26. Bourassa MG, Fisher LD, Campeau L, Gillespie MJ, McConnery M, Lesperance J. Long-term fate of bypass grafts: the Coronary Artery Surgery Study and Montreal Heart Institute experiences. Circulation 1985;72:(SuppI):V71-8. 27. Sanz G, Pajaron A, Alegria E, et al. Prevention of early aortocoronary bypass occlusion by low-dose aspirin and dipyridamole. Circulation 1990;82:765-73.

Discussion Ben P. Bidstrup (London, England). Those of you who were present at this meeting last year will recall that I presented a single-center study in which saphenous vein graft patency was examined somewhat earlier than Dr. Lemmer's group. Ninety patients underwent magnetic resonance imaging about 9 to 10 days after the operation. This method has similar sensitivity and specificity as ultrafast cine-CT. Although the numbers of patients were somewhat smaller, bearing in mind that this was a singlecenter study, there was no significant difference between the prevalence of graft occlusions in the two groups treated with either aprotinin or placebo. There is a notable effect of surgeon on graft patency. The EPP AC study (Etude de la permeabilite des pontages aortocoronaires) from France involved 18 centers and recruited some 400 patients. A substantial variation in patency of the saphenous vein grafts was noted, with half the centers having patency rates of 90% or better, a third of them having patency rates of 80% to 90%, and three of the centers having patency rates less than 80% and actually going down to 50%. Would Dr. Lemmer care to comment on the effect of surgeon or center on the rate of vein graft patency? The final point I'd like to make concerns the control of anticoagulation. This has caused considerable discussion at most of the presentations that have been made about the use of aprotinin, especially in North America. We have noticed for a long time that the ACT as measured with the Hemochron device (International Technidyne, Corp., Edison, N.J.), which uses a celite activator, is markedly prolonged in the presence of aprotinin. Because of this, we and others have recommended that the ACT be maintained above about 750 seconds in the presence of aprotinin to ensure adequate heparinization. Recent work has shown that in vitro at least, this prolongation is less marked when kaolin is used as the activator such as in the Hemotec device (Medtronic Hemotec, Englewood, Colo.) and the Hemochron FfK-ACT tubes. You showed one slide about anticoagulation at the beginning of your talk. Would you please elaborate on it, especially in light of the fact that you used the Hepcon HMS system for determining heparin levels? Dr. Safuh Attar (Baltimore, Md.). This study points out the increased potential of thrombosis of bypass grafts during CABG operations. The report of Dr. Cosgrove and associates at the Cleveland Clinic, confirming the increased prevalence of thrombosis of bypass grafts as well as the presence of thrombi in the heart, lungs, brain, and kidneys, points out the need for further research. Other areas of controversy involving aprotinin is its application in children. Recent reports from Munich by Dr. Dietrich point out its effectiveness in reducing bleeding in 60 children

552

Lemmer et al.

undergoing cardiac operations, whereas Dr. Balls from Giessen, Germany, found no effect on hemostasis, nor in the use of blood products. I should also mention the reported deaths from anaphylaxis after its use. Therefore the risks of aprotinin should be weighed against its benefits before its use is recommended. Three areas should be explored: ( I) The mechanism of action of aprotinin is still controversial, especially its effect on platelets; (2) once the mode of action of aprotinin is determined, then the cause of bleeding in cardiac operations should be investigated. Dr. Desmotis, an anesthesiologist from Washington University in St. Louis, reported in our symposium on transfusion practices in cardiac surgery about the significant decrease in the use of blood and blood products by on-site determination of coagulation parameters in the operating room, specifically prothrombin time, partial thromboplastin time, platelets, and fibrinogen. Only then can the proper therapy be matched to the coagulation defect. I have one question to ask the authors: Have you investigated the effect of aprotinin on ACT during cardiac operations, and should heparin administration be altered when aprotinin is used? Dr. Lemmer. I appreciate Mr. Bidstrup's previous work very much. He directed the only previously published study with regard to the effect of aprotinin on bypass graft patency. His contributions in this area are widely known and important. I was asked about the influence of surgeon and center on the graft patency rates. With regard to surgeons in this study, the vein graft patency rates ranged from 86% to 100%. For the centers, the patency rates ranged from 87% to 95%. Multiple logistic regression analysis did not show center to be a significant factor predictive of vein graft closure. In this study, the protocol specified that anticoagulation during bypass was to be performed without regard to the celite-activated ACT. It has been demonstrated by a number of laboratories in this country and in Europe, in in vivo and in vitro experiments, that the use of heparin and aprotinin together prolongs the ACT. This prolongation is probably independent of the heparin effect. The possibility therefore exists for the situation in which the ACT is long enough to be considered therapeutic but, in fact, does not reflect adequate anticoagulation. This situation occurs in the presence of both aprotinin and heparin. Therefore, in this study we used the Hepcon device which provides a measurement of the patient's heparin level. Generally, we keep the blood heparin level above 2.5 to 3.0 mg/kg of patient weight. Some centers used a fixed regimen dose, in which heparin was routinely given after every 90 minutes of bypass no matter what the ACT result was. If I understood correctly, Dr. Attar said that this study supports the concept that aprotinin causes thrombosis ofvein grafts. I hesitate to agree with his assessment. This study does not conclusively demonstrate an effect of aprotinin on vein graft patency. It is inconclusive. There were not enough grafts in the study to reach the statistical power needed to come to a definite conclusion in either direction. Multiple studies are now underway to try to answer these questions. One investigation, in which my group is participating, involves twenty centers with rigorous examination of postoperative ECGs and myocardial enzyme determinations in patients who have undergone CABG operations. Another multiinstitutional study is currently undertaking the huge task of subjecting each patient to graft arteriography shortly after the opera-

The Journal of Thoracic and Cardiovascular Surgery February 1994

tion. I hope these important questions can be answered conclusively in the near future.

Appendix 1: Criteria for elimination from efficacy analysis I. Patients for whom the random code had been broken at the discretion of the investigator (two patients) 2. Patients who did not receive the full dose of study medication because of an adverse experience, or in whom administration of the study medication was deliberately discontinued before the completion of the operation (two patients) 3. Patients who died intraoperatively or within 6 hours after the operation (two patients) 4. Patients who required mediastinal exploration for excessive bleeding from an anatomically localized site, that is, surgical bleeding (three patients) 5. Patients who required continued heparinization after the operation because of placement of a mechanical cardiac support or membrane oxygenator device (three patients) 6. Patients who had major postoperative bleeding from the gastrointestinal tract or other site unrelated to the surgical procedure (three patients) 7. Patients who had to undergo postoperative laparotomy for abdominal complications (two patients) 8. Patients who received donor blood products in the prime of the cardiopulmonary bypass circuit or before the operation after the blood sample for preoperative complete blood count had been drawn (three patients)

Appendix 2: Statistical methods Determination of study sample size. The primary variable to be analyzed in this study was the efficacy of aprotinin with regard to transfusion requirements. In the design of the study, we estimated that for primary CABG 60% of patients given placebo would require donor blood, and a 50% reduction (to 300/0) was considered clinically meaningful. Sample size calculations determined that with 65 patients per treatment group, a two-tailed test with an a level of 0.05 would yield 90% power of detecting a difference between the aprotinin group and the placebo group in a multicenter study, given a true difference of at least that magnitude. For redo operations, we estimated that 800/0 of patients given placebo would require donor blood, and a 50% reduction (to 40%) was considered clinically meaningful. Sample size calculations determined that with 35 patients per treatment group, a two-tailed test with an alpha of 0.05 would yield 90% power of detecting a difference between the aprotinin group and the placebo group in a multicenter study, given a true difference of at least that magnitude. Statistical methods. With the exception of the analysis of the graft patency rates, all statistical tests were two-tailed and were performed with an a level of 0.05. In this report, p values are presented with underlined numerals corresponding to the particular statistical method used in their derivation. Explanations of these methods are given below, along with the relevant underlined numeral appended in parentheses. The primary efficacy variable (percent of patients requiring any blood transfusion) was analyzed by a Mantel-Haenszel test adjusting for center U). Complication rates, laboratory abnormality incidence rates, and categorical values not tabled were analyzed by either Fisher's exact test or a X2 test. Fisher's exact tests were used if at least one fourth of the cells had expected values ofless than 5; otherwise, X2 tests were used (1). Tabled categorical variables other than complication rates and laboratory abnor-

The Journal of Thoracic and Cardiovascular Surgery Volume 107, Number 2

Lemmer et al.

553

Appendix 3. Demographic variables ofpatients fulfilling criteria for efficacy analysis Primary CABG

Redo CABG

Aprotinin Clinical parameter Mean age (yr) Male (%) White (%) Mean weight (kg) History of myocardial infarction (%) History of hypertension (%) Mean NYHA functional class (I-IV) (SE) No. (%) with LVEF <50% Preoperative hemoglobin

Aprotinin

Placebo

(n = 67)

(n = 74)

Placebo

(n = 23)

(n = 32)

62 82 80 83 54

64 76 79 82 54

66 91 96 85 57

65 91 94 86 69

61 2.9

63 3.0

48 2.9

41 3.1

26/74 (35%) 13.7

28/67 (42%) 13.5

7/23 (30%) 13.9

11/30 (37%) 13.2

(gm/dl) NYlfA, New York Heart Association; SE, standard error; LVEF,left ventricular ejection fraction. There were no differences in the above parameters between

the treatment groups in either the primary or redo CABG stratum (statistical methods 1, and

g; see Appendix 2).

Appendix 4. Surgical variables ofpatients fulfilling criteria for efficacy analysis Primary CABG

Redo CABG

Aprotinin

Aprotinin

Placebo

Parameter

(n = 74)

Placebo

Received preop. aspirin Mean No. (SE) of grafts (vein and IMA) No. of patients with IMA grafts (%) Mean volume (mI) of cardioplegic solution (SE) Mean total CPB minutes (SE) Mean No. of heparin units during first run of CPB Mean duration (hr) of operation (SE)

61% 3.2 (0.1)

63% 3.0 (0.1)

57% 2.8 (0.2)

53% 2.4 (0.2)

65/74 (88%)

54/67 (81%)

19/23 (83%)

21/32 (66%)

1910(77)

1730 (81)

2033 (104)*

1363 (81)*

99.1 (3.6)

98.2 (3.8)

119.7 (l1.0)t

91.2 (8.6)t

34,496

35,160

22,690

25,405

3.8 (0.1)

3.8 (0.1)

4.2 (0.3)

4.7 (0.2)

(n = 67)

(n = 23)

(n = 32)

SE, Standard error; CPB, cardiopulmonary bypass.

'Difference between the two redo treatment groups is significant (p < 0.001). tDifference between the two redo treatment groups is significant (p = 0.048). For all other comparisons, p > 0.05 (statistical methods 1, and ~; see Appendix 2).

mality incidence rates were analyzed by Mantel-Haenszel tests adjusting for center (1), except for graft occlusion rates, which were analyzed differently because of the sparseness of the data. The analysis of per-patient occlusion rates used a MantelHaenszel test adjusting for stratum (either primary or redo group) (1). The analysis of per-graft occlusion rates used a Mantel-Haenszel test adjusting for number of grafts within patients (4). An a levelof 0.10 was used for the analysis of graft occlusionrates. Because of gross departures from normality in certain variables (the number of units of donor blood required, the number of milliliters of donor blood required, the number of units of fresh frozen plasma required, and the number of units of cryoprecipitate required), these variables were analyzed by a twoway analysis of variance of ranked data, with the data ranked over all of the centers ~). A standard two-way analysis ofvariance model was used to analyze all other continuous efficacy

variables (Q). The initial analysis of variance model included the effects of drug, center, and drug-by-center interaction. For a center to be included in the interaction model, there had to be data for at least two patients per drug group. If there was no significant drug-by-center interaction, the main-effects model was used. For a center to be included in the main-effects model, there had to be data for at least one patient per drug group. This assured that each center had evaluated both treatments (in the main-effects model) and protected the least-square means from being unduly affected by outliers. All means presented for these variables are least-squares means from the aforementioned models. Data from the five centers were compiled and analyzed, as described earlier, by the Statistics and Data Systems Department of Miles Inc. (project statistician Lawrence A. Schwartz, MS). Independent statistical consultation was provided by Gary L. Grunkemeier, PhD.