Secondary Logo

Journal Logo


Aprotinin; friend or foe? A review of recent medical literature

Royston, D.1; van Haaften, N.1; De Vooght, P.1

Author Information
European Journal of Anaesthesiology (EJA): January 2007 - Volume 24 - Issue 1 - p 6-14
  • Free



A recent article [1] and accompanying editorials [2,3] in the New England Journal of Medicine (NEJM) have questioned the safety of using aprotinin in patients having heart surgery. A second article [4] has also suggested an increase in renal events in patients given aprotinin when compared to those where tranexamic acid was used. The present review will focus principally on the first of these articles in relation to previously published data and experience.


There are over 30 publications from randomized placebo-controlled studies of the use of a high dose aprotinin regimen, as described in 1987 [5], in a variety of heart surgeries. These studies, together with two meta-analyses [6,7] have shown that aprotinin is unique in reducing transfusions and the rates of return to the operating theatre to control bleeding after heart surgery.

For the Cochrane review [6] 61 trials evaluated aprotinin, of which 55 were in patients undergoing cardiac surgery with 3814 patients randomized to receive aprotinin and 2755 patients randomized to a control group. Relative risk (RR) of transfusion was reduced by 30% (RR 0.71; 95% CI 0.66–0.77; P < 0.0001). In 27 studies evaluated, 29 of 1758 patients were returned to theatre for re-exploration for bleeding in aprotinin treated patients compared with 58 of 1152 patients included in the control arms of these studies (RR 0.40; 95% CI 0.25–0.66; P = 0.0003). Sedrakyan and colleagues [7] performed a meta-analysis of 35 prospective, randomized controlled trials (n = 3879) that evaluated clinical outcomes in isolated coronary artery bypass grafting (CABG) patients administered aprotinin. Aprotinin reduced transfusion requirements by about 40% (RR 0.61; 95% CI 0.58–0.66) compared to placebo. Aprotinin is the only drug with this Class A Level 1 evidence for efficacy.

The absolute mechanism of action of aprotinin is unknown. However it is clear that this agent is also unique by reducing transfusion burden in a wide variety of surgeries in addition to cardiac procedures such as major orthopaedic [8], thoracic [9] and hepatic resection [10] and transplantation [11].

Unfortunately in the article critical of aprotinin [1] there are no transfusion data. The only efficacy data shown are mean and standard deviations of ‘estimated blood loss’ (this was not defined). This estimated loss was reported as not showing a statistically significant difference between treatment groups but the data showed wide variability. For example, in the patients allocated to tranexamic acid, the standard deviation of the mean (741 mL) was greater than the mean loss (676 mL).


Individual studies conducted principally for regulatory approval.

Because aprotinin had no license for use in any indication within the USA the Food and Drug Administration required certain pivotal studies to determine both efficacy and safety. These studies included eligible data from 1723 patients. The results of those studies have been published separately [12-16] and as grouped data [17].

These data showed no increase in the incidence of ischaemic events such as myocardial infarction and a possible decrease in the incidence of stroke. Also published [18] is an analysis of pooled data (n = 2283) from all sponsored US clinical trials in patients undergoing primary or repeat coronary revascularization and valvular surgery. This demonstrated a significant reduction in the incidence of stroke in patients receiving full-dose aprotinin compared to placebo, 1.0% vs. 2.4% (P = 0.027).

The largest and most comprehensive study on saphenous graft patency was the International Multicentre Graft Patency Experience (IMAGE) trial [16]. Patients having isolated first time CABG through a median sternotomy were randomly allocated to receive high dose aprotinin [5] (n = 436) or placebo (n = 434). There was no observed difference in mortality (aprotinin 1.4% vs. placebo 1.6%), patency of internal mammary artery bypass grafts (aprotinin 98.2% vs. placebo 98%), or definite myocardial infarction/ischaemia (MI) (aprotinin 2.9% vs. placebo 3.8%).

Patients had angiography to assess graft patency at a mean of 10.8 days (median 7 days; maximum 91 days) after surgery. Angiograms were read by an independent group of cardiologists blinded to the treatment group. Overall 90 patients had an occluded vein graft. Of these 22 had both an occluded vein graft and MI (11 aprotinin, 11 placebo). Only 6 patients in the study had more than one occluded graft. Saphenous vein graft occlusion occurred more often in patients allocated to aprotinin (15.4%) than to placebo (10.9%) (Risk ratio 1.5; 90% CI 1.1–2.1; P = 0.03). It is recognized that early vein graft patency is affected by many factors related to the conduit used (artery vs. vein), the target vessel (position, small size and poor distal vessel quality) and patient characteristics (female gender and use of anti-aggregatory therapy such as aspirin). When these factors were included in the analysis, the effect of aprotinin therapy was removed as a risk for vein graft closure.

With regard to renal impairment, in the studies used for regulatory approval a proportion of patients develop a short-lived postoperative increase in serum creatinine following cardiac surgery. The definition used of an ‘abnormal rise’ was >44 μmol L−1 (0.5 mg dL−1) above the baseline value. This occurred in about 10–15% of patients having primary revascularization surgery [19] with no statistically significant differences between aprotinin treated and placebo groups. Plasma creatinine tended to be higher in aprotinin treated patients when compared with the placebo population from about day 4 postoperatively [19]. However, the values were not usually outside the normal range and had returned to baseline values by the 14th postoperative day. In patients with a creatinine ≥ 124 μmol L−1 (≥1.4 mg dL−1) preoperatively, the incidence of a rise of ≥44 μmol L−1 (0.5 mg dL−1) was 18.5% of patients with aprotinin and 15% in the placebo patients [19]. Two patients in the aprotinin treated group had a rise in creatinine of >177 μmol L−1 (2.0 mg dL−1). These patients were found to have also received aminoglycoside antibiotics in the postoperative period. A formal study of this potential interaction [20] showed that high dose aprotinin therapy alone had no significant effect on postoperative creatinine or Cystatin C (a sensitive marker of tubular injury) concentrations. However this became significant when aprotinin was combined with vancomycin and 48 h of gentamycin therapy suggesting that aprotinin and aminoglycosides should be used together with caution. The incidence of an abnormal rise in creatinine was higher (about 15–20%) in those patients having repeat surgery through a prior sternotomy [12,14] and again there was a trend for the duration of this rise to be more prolonged. However this rise was again not statistically significant between treatment and placebo groups and in one study was also associated with a non-significant increase in creatinine clearance [12].

A significant rise in creatinine by the third postoperative day is however more likely to be observed with higher risk surgery such as after implantation of a Left Ventricular Assist Device (LVAD) [21]. In this report a significant rise in creatinine (>0.5 mg dL−1) was observed in nearly 40% of 42 aprotinin treated patients compared with 14% in the 100 non-aprotinin treated patients (P < 0.01). Of interest this rise in creatinine was not mirrored by changes in blood urea nitrogen estimations. This short-lived biochemical abnormality did not translate into an increase in the incidence of clinically significant adverse events. Indeed the use of aprotinin was associated with a significantly lower perioperative (7 day) mortality of 11.9% compared to 28% in those not given aprotinin therapy (P = 0.04).

The risk of renal failure requiring continuous support/dialysis associated with aprotinin use appears to be extremely low in more common clinical practice [14,19,22]. The study reported by Lemmer and colleagues [19] specifically on the renal effects of aprotinin showed that clinically important renal insufficiency was infrequent (1 of 108 treated and 1 of 108 placebo patients required new dialysis). Compared to the placebo group, no difference was found in serum electrolytes, blood urea nitrogen, urinalysis findings or in the frequency of abnormal creatinine clearance rates. These data have also been shown in studies with lower numbers of participants [23,24].

Meta-analyses of efficacy and safety data.

The Cochrane Collaboration [6] has published an evidence-based review of randomized controlled trials in adults scheduled for non-emergency surgery conducted to assess the effects of aprotinin, aminocaproic acid and tranexamic acid on perioperative red blood cell (RBC) transfusion. Other outcomes measured included reoperation due to bleeding, mortality and postoperative complications (non-fatal MI, stroke, deep vein thrombosis, pulmonary embolism, any thrombosis, renal failure). Of the 89 trials that met inclusion criteria, 61 evaluated aprotinin, of which 55 were in patients undergoing cardiac surgery with 3814 patients randomized to receive aprotinin and 2755 patients randomized to a control group.

In 28 studies that reported data on mortality, there was a non-significant reduction RR of 13% (RR 0.87; 95% CI 0.63–1.19) in those participants treated with aprotinin (n = 2828) compared to control (n = 2085).

In 20 trials that reported data on non-fatal MI, the pooled RR of sustaining a non-fatal MI in those participants treated with aprotinin (n = 1871) compared to control (n = 1117) was 0.97 (95% CI 0.69–1.36).

In 8 trials that reported data on stroke, the pooled RR of developing a stroke in those participants treated with aprotinin (n = 605) compared to control (n = 373) was 0.43 (95% CI 0.16–1.19). Aprotinin therapy was therefore associated with a 57% reduction in RR of stroke but this did not reach statistical significance.

Of the 13 trials that reported data on renal failure/renal dysfunction, the risk of developing renal failure/renal dysfunction in those participants treated with aprotinin (n = 2210) compared to control (n = 1566) was not significantly elevated (RR 1.19; 95% CI 0.79–1.79; P = 0.4).

Sedrakyan and colleagues [7] performed a meta-analysis of 35 published, prospective, randomized controlled trials (n = 3879) that evaluated clinical outcomes (transfusion, mortality, MI, renal failure, stroke, atrial fibrillation) in CABG patients administered aprotinin. Aprotinin reduced transfusion requirements by about 40% (RR 0.61; 95% CI 0.58–0.66) compared to placebo.

The analysis showed no increased risk of MI (28 trials, n = 3555) in which 96/2024 aprotinin patients developed an MI vs. 77/1531 with placebo (RR 0.85; 95% CI 0.63–1.14) or mortality (32 trials, n = 3779) in which 53/2149 aprotinin treated patients died against 39/1630 with placebo (RR 0.96; 95% CI 0.65–1.40).

Of note, in aspirin non-users (9 trials, n = 776), 6 of 440 patients developed an MI in aprotinin treated and 13/336 placebo treated patients (RR 0.40; 95% CI 0.17–0.92; P < 0.05).

Stroke was evaluated in 18 trials (n = 2976). Aprotinin therapy was associated with a 47% reduction in RR of stroke (RR 0.53; 95% CI 0.31–0.90; P < 0.05).

In the 17 trials that evaluated renal failure/dysfunction included in this analysis aprotinin therapy was not associated with increased or decreased risks of renal failure (RR 1.01; 95% CI 0.55–1.83).

The interpretation of the data presented in the randomized, blinded, placebo-controlled studies and two meta-analyses was that high dose aprotinin therapy significantly reduced transfusion burden and need for re-exploration, had no statistically significant effect on mortality, MI or renal failure and may be associated with a statistically significant reduction in the incidence of stroke.

Multicentre Studies of Perioperative Ischaemia (McSPI) Research Group and the Ischaemia Research and Education Foundation (IREF) publications.

This group has collected two large datasets. The McSPI epidemiology 1 dataset collected data from 2417 patients having surgical myocardial revascularization in 24 centres in the USA. These data were collected between 1991 and 1993. The epidemiology 2 database collected data between 1996 and 2000 from patients having myocardial revascularization in 70 centres in 17 countries. The database captured data from 5436 patients. Of these 5065 patients data were included in the final dataset. These data have been analysed and published in a number of articles [1,25-29]. The publication that has caused most attention is the most recent article published in the NEJM [1] that purports to show the use of aprotinin therapy is associated with a significantly increased risk of MI and heart failure, stroke and encephalopathy in patients having primary isolated coronary revascularization but NOT in those having complex surgery (the licensed indication for the drug). In addition the article claims to show a doubling of the risk for renal dysfunction and need for dialysis in those patients having primary low risk surgery and also in those having complex surgery. The abstract of the article received massive publicity in the media in the USA and has excited the interest of large numbers of malpractice groups.

It cannot be over-emphasized that the McSPI database analysis is NOT from a randomized study but is a retrospective analysis of observational data that were collected prospectively. Selection bias is a major problem with this type of study design. The multiple publications from the same database raise some intriguing issues related specifically to selection bias in analyses from observational studies and the possible effects of that bias.

As the first example, the legend to Figure 3 in the dataset published in 2002 showing a signifi-cant benefit of aspirin [25] states ‘Antifibrinolytic therapy entailed the use of aprotinin (in 1578 patients), aminocaproic acid (in 1258 patients), tranexamic acid (in 951 patients) or desmopressin (in 61 patients). There were 189 patients who received more than one of these drugs, and 6 patients had missing data on antifibrinolytic therapy’. A total of 1363 patients were stated to have received no antifibrinolytic therapy. These data were from 5022 patients who survived 48 h after surgery (43 died in the early postoperative period for a total n of 5065 patients). Figure 1 of the article published in 2006 [1] from the same starting dataset of 5065 patients shows aprotinin was used in 1295 (Δ = minus 283 (17.9%) compared to 2002), aminocaproic acid in 822 (Δ = minus 436 (34.7%)), tranexamic acid in 883 (Δ = minus 68 (7%)) and no antifibrinolytics in 1374 (Δ = plus 11 (0.8%)). According to this figure, were excluded data from a total of 691 patients (226 patients receiving multiple drugs (Δ = plus 37 (19.6%)), 17 (plus 9) without data and 448 who were said to have received an inadequate dose of drug although this criteria is not defined). There are obvious major differences in the data selected when showing a purported benefit of aspirin compared to those showing a deleterious effect of aprotinin even though the original dataset is the same. The potential effect of this in interpreting these articles is discussed further in a subsequent section.

Figure 1.
Figure 1.:
Non-risk adjusted renal dysfunction rates [ 1,30,32-34 ].

The second issue relates to the effect of selection and exclusion bias and is a consequence of data not being from a randomized study. Allocation of the patient to a specific treatment was at the discretion of the clinicians involved. For this reason it is clear that aprotinin was more likely to be administered if the patient had a perceived higher risk of bleeding and transfusions. The 2006 NEJM article acknowledges that the patients who received aprotinin were at considerably higher risk for worse outcomes in addition to the transfusion risk. Table 1 of the 2006 article lists and compares 27 different preoperative characteristics. For example, reoperation through a prior sternotomy was 5 fold more common in the aprotinin treated patients compared with patients who received no drug, and 3 fold more than those receiving tranexamic acid. Concurrent valve surgery with myocardial revascularization was 2 fold more common with aprotinin compared to no drug or tranexamic acid treated patients. A history of carotid disease was reported about 50% more often in the aprotinin treated than the other two groups. Notably some factors are missing from the list of preoperative risk factors such as patient age, and times of cardiopulmonary bypass (CPB), use of aspirin and the influence of country and centre. This is more surprising given that not only have these data been published by the same authors elsewhere but also these excluded factors have been separately shown in McSPI group publications to have a highly significant impact on renal failure [30], mortality, heart failure and MI [26] and these four morbidities together with neurological outcomes [25].

Table 1
Table 1:
Aprotinin usage in McSPI and subsequent mortality outcomes.

Another less obvious example of bias in the analysis is the number of patients classified as having MI. In the 2002 publication this was diagnosed in 185 patients. This rises to over 600 in the 2006 manuscript. This enormous difference is because electrophysiological data were used only in the latter analysis, the enzyme data (creatine kinase MB isoform concentration) used in the earlier article were excluded. It is impossible to determine from the 2006 manuscript if this exclusion bias altered the interpretation and conclusion of an increase in MI in aprotinin treated patients having primary surgery.

More important is the data for mortality, which has a universally accepted definition. There were 164 deaths in 5065 patients reported in 2002 [25]. This falls to about 110–115 of the 4374 patients included in the new analysis [1]. This suggests that about 50 deaths occurred in the 691 patients were excluded from the most recent analysis. The mortality rate in the excluded patients was therefore about 3 times higher than the overall rate of 2–3%. Most clinicians would consider that exclusion of about 30% of the overall deaths must lead to some bias in the conclusions especially as nearly 50% of these excluded patients had received epsilon-aminocaproic acid and were therefore patients predominantly treated in North American centres. This last figure is based on the differences between numbers of drug treated patients in the two manuscripts discussed above.

Nonetheless the authors claim to overcome the apparent selection bias by the use of two statistical methods – multivariate regression analysis and comparison using propensity scores. While the use of these methods to control selection bias is valid, it does not completely remove the bias. Therefore, it is necessary to examine in some detail the assumptions for statistical methods and how the data has been handled.

It is obvious that discussion of the data analysed without and with risk adjustment leads to large differences in how it could be interpreted. In the text of the 2006 article the authors wrote ‘Among the 3013 patients undergoing primary surgery, aprotinin, but not aminocaproic acid or tranexamic acid, was associated with the risk of death (2.8% vs. 1.3%, P = 0.02)’. They follow this with similar comparisons showing worse outcomes for cardiovascular, cerebrovascular and renal end-points in aprotinin treated patients. But in the tabulated results, there are completely different numbers. The P value for the risk of death with aprotinin vs. control in primary patients is 0.22. The risk for death (P = 0.66), cardiovascular (P = 0.67) and cerebrovascular (P = 0.41) events after complex surgery (the licensed indication for use) are not significant statistically. It is difficult to come to a coherent theory as to why the use of aprotinin alone demonstrated significantly increased ischaemic events only in patients having primary surgeries if this effect is as a direct effect of the drug.

Multivariate analysis is used to identify the contributions of multiple factors (confounders) both individually and in combination to one or more outcomes. Logistic regression is used when the outcome variable is dichotomous, i.e. it can take on only two different values. The goal of regression analysis is to correctly predict outcome using the simplest possible model. The crucial step in this process is the creation of a mathematical model, with all the predictor variables. The data in this study has been analysed using a technique called stepwise regression, where a so-called fully saturated model (i.e. one in which there are as many parameters as unknowns) is created followed by stepwise reduction of data, with retesting at each stage. The authors have used exploratory testing, which is used when there is no a priori hypothesis about the model, and the contributions of the factors or the nature of the model is unknown.

Propensity score is the propensity towards (probability of) belonging to one group and not the other. For example, in a population with 1 male and 9 females the propensity score for being a female would be 0.9. Propensity scores are used to handle selection bias in observational studies by matching patients based on specific characteristic propensity scores. However, if the number of observed characteristics is high as in this study, matching is not practical especially if there is a large selection bias (those patients with a perceived higher bleeding risk are more likely to receive aprotinin). In this circumstance a composite propensity score is created using all the characteristics, and matching is done on the basis of this composite score. Propensity scores were created using multivariable regression analysis on the data. By using regression analysis, the probability of a patient receiving aprotinin is identified given the clinical risk factors (in the methods section the article states 45 factors were identified as determining treatment but they are not defined). The model also needs to be balanced for the covariates. When the propensity score has been created, it can be used to standardize the two comparison groups. Propensity score can be used in three ways: matching, stratification or as one of the factors in regression analysis to compare outcomes. This is what the recent study has used. The caveat, which cannot be over-emphasized is the use of propensity score in this way assumes the model included ALL covariates and confounders.

The key assumption with these techniques is therefore the validity of the model.

The interpretation of results from this multivariable model depends on both the choice of factors and the order in which these factors were added or subtracted within this model. The potential problem of interpretation from such modelling is highlighted when considering the risk factors for developing renal dysfunction.

We do know that a large amount of data was collected. Despite this when the model for developing renal dysfunction/failure was created, an additional 410 patients data were excluded as their dataset was incomplete for one or more factors used in developing a propensity score to include in the model. Overall the recent NEJM article has excluded nearly 20% of the patients presented in the original published analysis of the McSPI database.

In addition, how the predictive factors in the model for renal impairment were chosen is not known. A number of critically important issues stem from the choice of factors in the model. First the use of aspirin does not feature as a protective factor for renal dysfunction in the 2006 analysis [1]. This is despite the same dataset being used to demonstrate a 60% reduction in renal dysfunction associated with aspirin use in the prior McSPI publication in the NEJM [25]. This leads the reader to question if there was a difference in confounding variables used in the model developed for the 2006 analysis compared with those developed for prior publications by the McSPI group. Obviously the multivariable and logistic regression models used previously cannot apply if criteria have been changed without obvious consistency and in an apparently idiosyncratic manner. The alternate explanation is that the exclusion of data from over 1000 patients for the 2006 analysis compared with the earlier publication removed the effect of aspirin as a protective agent on renal outcomes. In this case this questions the validity of the conclusion from the McSPI group that aspirin therapy has a universal organ protective action after heart surgery.

Second, the renal failure model identified use of fresh frozen plasma (FFP) as a risk factor with odds ratios almost identical to that for the aprotinin risk without and in the presence of propensity adjusted covariates. What conclusions are we to draw from this? Are we to infer that the use of FFP somehow affects renal outcome as a transfusion associated reaction? Nobody doubts the existence of transfusion related acute lung injury (TRALI) but have we missed FFP related renal injury? Is it not more likely that FFP was used in patients with more bleeding and thus there was a worse renal outcome because of relative reductions in renal perfusion possibly associated with hypovolaemia? Unfortunately, we cannot answer this question without knowing the steps the authors used in the creation of the model. If the use of FFP was associated with bleeding and not an immune mediated nephrotoxic action, then is it reasonable to infer that these patients were more likely to receive aprotinin, and any association between aprotinin and renal outcomes were confounded by the perfusion defects? This problem, termed colinearity, occurs when variables are so highly correlated that minor changes in one independent variable will have a major impact on the weighting of others. This could be a critically important issue in interpreting the conclusions from this article. Without the actual data together with some knowledge of how it was manipulated must lead the reader to assume that there are significant errors to be considered.

A further problem with the data analysis is that the models generated and their analysis appears to give equal weight to data from low volume programmes as those from high volume institutions. In addition the data shown in Table 1 taken from the authors prior publication [26] shows marked differences between countries as to the contributions made by these centres and the ischaemic outcomes by participating country.

Simple mathematical manipulation from these data and those presented in the 2002 McSPI article [25] reveal that 80 deaths occurred in the remaining 1885 patients not included in Table 1 and eligible for analysis. This gives a mortality rate of 4.2% in this cohort. These 1885 patients had their surgery in one of 20 centres in 13 countries (an average n of 94 patients contributed per centre).

There are also large differences between countries in approaches to therapy (crucial to the propensity assumptions). Two examples illustrate this problem. First during the time of data collection, epsilon-aminocaproic acid could not be given to human beings in the UK or Germany and this agent has been withdrawn from the market by the manufacturer in Canada. Likewise aprotinin was not available for human use in Italy.

Second, and related to need for renal support, a recent article from Germany [31] reported on data from over 8000 patients having primary isolated myocardial revascularization at that centre. The authors showed that their rate of renal replacement therapy (haemofiltration) with normal renal function preoperatively was 4.8%. The incidence quoted in patients with abnormal preoperative renal function was 15.8%. These figures may seem high to many readers but they reflect a trend in many centres to aggressively correct fluid and electrolyte balance using haemofiltration prior to biochemical evidence of overt renal impairment. Could this in part explain why nearly one in 5 patients who received high dose aprotinin therapy were reported to have developed ‘renal failure’ and one in 20 required dialysis?

Finally, when considering exclusion bias in the analysis, the article has a figure showing a statistically significant dose-dependent effect of aprotinin therapy on renal and cardiovascular outcome. These data were from 596 patients given aprotinin therapy. The figure does not differentiate between primary and complex surgery. It is left to the reader to speculate if this neat relationship would still be apparent if the data from the other 699 patients who received aprotinin had been included.

Notwithstanding the problems of interpreting how the data were analysed, the fundamental message (to lawyers and others that may have only read the abstract) is that aprotinin is nephrotoxic. Let us specifically examine these data and the conclusions drawn.

Studies have been reported from various countries that have investigated the relationship between evidence of postoperative biochemical renal impairment and preoperative risk. These articles can be used to show the likely proportion of patients with postoperative renal impairment based on a large rise in plasma creatinine or a rise above an absolute value. Data preented from Italy [32] (n = 2009, criteria = creatinine >177 μmol L−1), Portugal [33] (n = 2445, criteria = creatinine rise of >80 μmol with maximum value >177 μmol), Finland [34] (n = 815, criteria = 2-fold rise or >100 μmol L−1 rise), Duke University Medical Centre in the USA [35] (n = 2672, criteria = rise of >88 μmol L) and the original McSPI epidemiology 1 dataset [30] that reported on 2222 patients using the same criteria as in the current article of either a rise of >62 mmol L−1 or an absolute value of >177 μmol L−1. If the data from these studies are now plotted together with the data shown in the 2006 NEJM article for non-risk adjusted renal dysfunction rates then the following figure is produced (Fig. 1).

This figure appears to show that the increased risk of renal dysfunction attributed to aprotinin use was more because of a significant reduction in the value for the no drug groups than previously published by McSPI and other groups. Instead of trying to explain this, the authors discuss data from studies in rats from the 1960s to the 1980s. Cellular effects of aprotinin upon renal cells in rats tested without CPB or anticoagulation in the 1960–1980s may have little relevance to the current argument. A more contemporary study is reported from Stanford University in 183 patients who survived more than 24 h after surgery that included a period of deep hypothermic circulatory arrest. The article [36] failed to show any association (P = 0.95) between renal failure and the use of aprotinin in this setting. What has surprised many observers is that Mangano and colleagues did not discuss these results despite funding by the IREF/McSPI system and with the same statistician conducting the statistical analysis of this and the 2006 NEJM article.

Data from a single centre analysis.

A second article also appeared early in 2006 [4] that has been quoted as suggesting aprotinin therapy was associated with significant renal impairment. This analysis included data from over 10 000 patients having cardiac surgery at the Toronto General Hospital. The study was not specifically designed to investigate safety issues but was specifically intended to show non-inferiority of tranexamic acid when compared to aprotinin using transfusions as the end-point. A total of 10 294 patients received tranexamic acid and 586 received high dose aprotinin. The authors accepted that aprotinin was used in much higher risk patients and also used a propensity score based on transfusion risk to match the groups. After this matching a comparison was made between 449 patients allocated to each treatment. In other words a match could not be found for 137 patients (23%) given aprotinin based on their very high transfusion risk. Of the patients eventually included over 70% had complex surgery, 50% were reoperations through a prior sternotomy, 12% had a period of circulatory arrest, 16% were having emergency surgery and 6% had endocarditis. This patient population is obviously different to that reported from the McSPI database analysis. Acute renal failure was defined as a new requirement for dialysis support; acute renal dysfunction was defined as a greater than 50% increase in creatinine concentration during the first postoperative week to more than 100 μmol L−1 (1.13 mg dL−1) in female and greater 110 μmol L−1 in male (1.24 mg dL−1). The reader will note that these values are within the laboratory normal range. One hundred and seven of 449 aprotinin treated patients (24%) and 75 of the 449 patients given tranexamic acid (17%) fulfilled these criteria (P = 0.01). However when considering the need for new dialysis in those patients with normal preoperative renal function the data showed this was necessary in 11 of 339 patients allocated to aprotinin and 6 of 323 patients in the tranexamic acid group (P = 0.3). Corresponding data for those with abnormal preoperative renal function were 14 of 110 aprotinin and 6 of 186 tranexamic acid (P = 0.1). The observation from these very high risk patients of a significantly greater number with a rise in plasma creatinine in the early postoperative period but without a statistically increased risk for dialysis is in keeping with prior publications [19] and also three meta-analyses where renal function was a specific end-point [6,7,37].

So, what should the members of the caring team conclude or do in light of these recent publications? The McSPI/IREF data are presented by well-known authors and published in a reputed journal so what conclusions can be drawn? Because of the critical underlying problems of selection bias, which was not completely removed by the statistical methods, we can only conclude that aprotinin administration prior to cardiac surgery and raised biochemical markers of renal dysfunction are correlated. The data presented can never be used to show causality. Raised biochemical markers were also shown in prior studies submitted to the FDA and it was concluded this had little if any clinical relevance. The conclusions of the NEJM paper regarding renal dysfunction should therefore be considered only as a hypothesis and not as a definitive answer. The total of the data based associations in this study should be digested and compared with over 500 papers on cardiac surgery and the use of aprotinin. Everyone should read the article carefully and in light of what data are present and what are missing, ask themselves whether they agree or support the conclusions. The FDA have not withdrawn the drug from use during their investigations nor have they suggested that clinicians should stop using the drug if they consider that the risks of transfusion and bleeding in a particular patient scenario are high. In the meantime it is recognized that transfusions alone confer an increased immediate [38,39] and long-term risk [40-42] to the patient. The risks of death, perioperative infection, respiratory and renal failure, length of intensive care unit (ICU) and hospital stay all worsen with more transfusion [39]. No transfusion data are presented in the NEJM paper.

If the NEJM article is incorrect and patients suffer increased transfusions leading to renal failure, and death because aprotinin therapy was withheld due to a fear of litigation what debt is owed to our patients for such information?


1. Mangano DT, Tudor IC, Dietzel C. The risk associated with aprotinin in cardiac surgery. New Engl J Med 2006; 354(4): 353–365.
2. Hunter D. First, gather the data. New Engl J Med 2006; 354(4): 329–331.
3. Vlahakes GJ. The value of phase 4 clinical testing. New Engl J Med 2006; 354(4): 413–415.
4. Karkouti K, Beattie WS, Dattilo KM et al. A propensity score case-control comparison of aprotinin and tranexamic acid in high-transfusion-risk cardiac surgery. Transfusion 2006; 46(3): 327–338.
5. Royston D, Bidstrup BP, Taylor KM, Sapsford RN. Effect of aprotinin on need for blood transfusion after repeat open-heart surgery. Lancet 1987; 2(8571): 1289–1291.
6. Henry DA, Moxey AJ, Carless PA et al. Anti-fibrinolytic use for minimising perioperative allogeneic blood transfusion. Cochrane Database Syst Rev 2001 2001(1): CD001886.
7. Sedrakyan A, Treasure T, Elefteriades JA. Effect of aprotinin on clinical outcomes in coronary artery bypass graft surgery: a systematic review and meta-analysis of randomized clinical trials. J Thorac Cardiovasc Surg 2004; 128(3): 442–448.
8. Samama CM, Langeron O, Rosencher N et al. Aprotinin versus placebo in major orthopedic surgery: a randomized, double-blinded, dose-ranging study. Anesth Analg 2002; 95(2): 287–293, table of contents.
9. Bedirhan MA, Turna A, Yagan N, Tasci O. Aprotinin reduces postoperative bleeding and the need for blood products in thoracic surgery: results of a randomized double-blind study. Eur J Cardiothorac Surg 2001; 20(6): 1122–1127.
10. Lentschener C, Benhamou D, Mercier FJ et al. Aprotinin reduces blood loss in patients undergoing elective liver resection. Anesth Analg 1997; 84(4): 875–881.
11. Porte RJ, Slooff MJ. Aprotinin: safe and effective in all patients undergoing orthotopic liver transplantation? Liver Transpl 2001; 7(9): 808–810.
12. Cosgrove Dd, Heric B, Lytle BW, et al. Aprotinin therapy for reoperative myocardial revascularization: a placebo-controlled study. Ann Thorac Surg 1992; 54(6): 1031–1036.
13. Lemmer Jr JH, Stanford W, Bonney SL et al. Aprotinin for coronary bypass operations: efficacy, safety, and influence on early saphenous vein graft patency. A multicenter, randomized, double-blind, placebo-controlled study. J Thorac Cardiovasc Surg 1994; 107(2): 543–551.
14. Levy JH, Pifarre R, Schaff HV et al. A multicenter, double-blind, placebo-controlled trial of aprotinin for reducing blood loss and the requirement for donor-blood transfusion in patients undergoing repeat coronary artery bypass grafting. Circulation 1995; 92(8): 2236–2244.
15. Lemmer Jr JH, Dilling EW, Morton JR et al. Aprotinin for primary coronary artery bypass grafting: a multicenter trial of three dose regimens [see comments]. Ann Thorac Surg 1996; 62(6): 1659–1667.
16. Alderman EL, Levy JH, Rich JB et al. Analyses of coronary graft patency after aprotinin use: results from the International Multicenter Aprotinin Graft Patency Experience (IMAGE) trial. J Thorac Cardiovasc Surg 1998; 116(5): 716–730.
17. Royston D, Levy JH, Fitch J et al. Full-dose aprotinin use in coronary artery bypass graft surgery: an analysis of perioperative pharmacotherapy and patient outcome. Anesth Analg 2006; 103: 1082–1088.
18. Smith PK, Muhlbaier LH. Aprotinin: safe and effective only with the full-dose regimen [editorial; comment]. Ann Thorac Surg 1996; 62(6): 1575–1577.
19. Lemmer Jr JH, Stanford W, Bonney SL et al. Aprotinin for coronary artery bypass grafting: effect on postoperative renal function. Ann Thorac Surg 1995; 59(1): 132–136.
20. Mercieri M, Mercieri A, Tritapepe L et al. High-dose aprotinin with gentamicin–vancomycin antibiotic prophylaxis increases blood concentrations of creatinine and cystatin C in patients undergoing coronary artery bypass grafting. Br J Anaesth 1999; 82(4): 531–536.
21. Goldstein DJ, Seldomridge JA, Chen JM et al. Use of aprotinin in LVAD recipients reduces blood loss, blood use, and perioperative mortality. Ann Thorac Surg 1995; 59(5): 1063–1067.
22. Bidstrup BP, Harrison J, Royston D, Taylor KM, Treasure T. Aprotinin therapy in cardiac operations: a report on use in 41 cardiac centers in the United Kingdom. Ann Thorac Surg 1993; 55(4): 971–976.
23. Blauhut B, Gross C, Necek S, Doran JE, Spath P, Lundsgaard-Hansen P. Effects of high-dose aprotinin on blood loss, platelet function, fibrinolysis, complement, and renal function after cardiopulmonary bypass [see comments]. J Thorac Cardiovasc Surg 1991; 101(6): 958–967.
24. Schweizer A, Hohn L, Morel DR, Kalangos A, Licker M. Aprotinin does not impair renal haemodynamics and function after cardiac surgery [see comments]. Br J Anaesth 2000; 84(1): 16–22.
25. Mangano DT. Aspirin and mortality from coronary bypass surgery. New Engl J Med 2002; 347(17): 1309–1317.
26. Ott E, Moehnle P, Tudor I, Hsu P, Mangano D. CABG-Surgery in Europe and North America: Timelines and Outcomes. Anesthesiology 2003; 99: A254.
27. Mathew JP, Fontes ML, Tudor IC et al. A multicenter risk index for atrial fibrillation after cardiac surgery. JAMA 2004; 291(14): 1720–1729.
28. Nussmeier NA, Mora-Mangano C, Fontes M, Schwann NM, Mangano DT. Hormone replacement therapy is safe in women undergoing coronary artery bypass grafting. Tex Heart Inst J 2005; 32(4): 507–514.
29. Drenger B, Gozal Y, Fontes M, Mangano D. The association between ACE inhibitors pattern of use and patient outcome after CABG surgery. Anesthesiology 2006; 105: A1737.
30. Mangano CM, Diamondstone LS, Ramsay JG, Aggarwal A, Herskowitz A, Mangano DT. Renal dysfunction after myocardial revascularization: risk factors, adverse outcomes, and hospital resource utilization. The Multicenter Study of Perioperative Ischemia Research Group. Ann Intern Med 1998; 128(3): 194–203.
31. Bucerius J, Gummert JF, Walther T et al. On-pump versus off-pump coronary artery bypass grafting: impact on postoperative renal failure requiring renal replacement therapy. Ann Thorac Surg 2004; 77(4): 1250–1256.
32. Zanardo G, Michielon P, Paccagnella A et al. Acute renal failure in the patient undergoing cardiac operation. Prevalence, mortality rate, and main risk factors. J Thorac Cardiovasc Surg 1994; 107(6): 1489–1495.
33. Antunes PE, Prieto D, Ferrao de Oliveira J, Antunes MJ. Renal dysfunction after myocardial revascularization. Eur J Cardiothorac Surg 2004; 25(4): 597–604.
34. Kuitunen A, Vento A, Suojaranta-Ylinen R, Pettila V. Acute renal failure after cardiac surgery: evaluation of the RIFLE classification. Ann Thorac Surg 2006; 81(2): 542–546.
35. Conlon PJ, Stafford-Smith M, White WD et al. Acute renal failure following cardiac surgery. Nephrol Dial Transpl 1999; 14(5): 1158–1162.
36. Mora Mangano CT, Neville MJ, Hsu PH, Mignea I, King J, Miller DC. Aprotinin, blood loss, and renal dysfunction in deep hypothermic circulatory arrest. Circulation 2001; 104(12 Suppl 1): I276–I281.
37. Brown J, Birkmeyer N, O'Connor J. Aprotinin in cardiac surgery. New Engl J Med 2006; 354(18): 1953–1957.
38. 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(8): 1749–1756.
39. Koch CG, Li L, Duncan AI et al. Transfusion in coronary artery bypass grafting is associated with reduced long-term survival. Ann Thorac Surg 2006; 81(5): 1650–1657.
40. Engoren MC, Habib RH, Zacharias A, Schwann TA, Riordan CJ, Durham SJ. Effect of blood transfusion on long-term survival after cardiac operation. Ann Thorac Surg 2002; 74(4): 1180–1186.
41. Kuduvalli M, Oo AY, Newall N et al. Effect of peri-operative red blood cell transfusion on 30-day and 1-year mortality following coronary artery bypass surgery. Eur J Cardiothorac Surg 2005; 27(4): 592–598.
42. 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; 94(6): 1608–1616.


© 2007 European Society of Anaesthesiology