Pro: Aprotinin Has a Good Efficacy and Safety Profile Relative to Other Alternatives for Prevention of Bleeding in Cardiac Surgery

Two recently published reports have evoked partisan debate and regulatory review on the safety of the widely used protease inhibitor, aprotinin (Bayer Pharma, West Haven, CT) (1,2). These retrospective observational studies claimed significant increases in several adverse postoperative events after cardiac surgery, without important reduction in blood loss or other benefits. However, several earlier studies have documented the efficacy of aprotinin in decreasing blood loss and transfusion exposure, and have found an acceptable safety profile, even a reduction in the incidence of stroke without alteration in renal function or increased risk of adverse thrombotic events (3–6). The dichotomy between two nonrandomized, observational clinical studies published in well-regarded journals, and many well-conducted randomized trials is, as yet, unresolved, and may depend on details of statistical analyses. Because of this controversy, we wish to summarize the large body of prior literature demonstrating the efficacy and safety of aprotinin, review potential weaknesses of the two recent observational studies, point out remaining “holes” that require future study, and summarize the ongoing BART study (Blood conservation using Antifibrinolytics: Randomized Trial in high-risk cardiac surgery), which is the largest randomized trial comparing all the three antifibrinolytic drugs. APROTININ HAS A FAVORABLE EFFICACY PROFILE Aprotinin has been shown to reduce bleeding during and after cardiac surgery when compared with placebo (5,6). Randomized trials of aprotinin versus placebo performed using a transfusion protocol show a 32% reduction in transfusion exposures (95% confidence interval, CI 26%–39%) (6). Without a transfusion protocol, the effect of aprotinin is attenuated (23% reduction), but is still significant and important (95% CI 4%–38%) (6). By contrast, the effect on bleeding of “half-dose” aprotinin approaches “full-dose” aprotinin (odds ratio, OR = 0.68, 95% CI 0.62%–0.75%) (6). The effect of aprotinin on the number of patients transfused, when used solely in the cardiopulmonary bypass (CPB) prime solution, is small (OR = 0.85, 95% CI 0.73%–0.99%) (6). The evidence for a greater reduction in blood loss with aprotinin compared to other antifibrinolytics, such as ε-aminocaproic acid (EACA) or tranexamic acid (TEA), is less robust. This is because drug companies are only required by the Food and Drug Administration to show efficacy when compared with placebo and not compared with currently available drugs. Ten trials involving 1707 patients have examined aprotinin versus TEA (7). Overall, these trials showed a reduction in mean blood loss with aprotinin compared to TEA, which was statistically different, but with limited clinical significance (106 mL). This observation is not surprising as most studies were performed in populations at low risk for bleeding, such as those undergoing primary coronary artery bypass (CABG) surgery, where the relative efficacy of these drugs may be quite different from redo median sternotomy. Most importantly, aprotinin resulted in a significant 60% reduction in reoperation for bleeding (CI 34%–75%) compared to TEA and EACA (7). This is a very important beneficial clinical effect from an antifibrinolytic, given the significant morbidity associated with re-sternotomy. Although aprotinin is indicated for the prevention of perioperative bleeding, there is also evidence that it reduces the frequency of the most devastating adverse event–stroke (5,6). Even though the risk of stroke is relatively low (approximately 3%), the personal and social cost far outweighs the cost of the original surgical procedure. Additionally, no other therapy, including TEA and EACA, has been shown to reduce the risk of stroke. Overall, the stroke rate in aprotinin-treated patients was reported to have been reduced by 47% (CI 10%–69%), as evaluated in 18 trials with 2976 patients (5). The mechanism for stroke reduction could either be a direct effect, perhaps because of antiinflammatory properties of the drug, or an indirect effect because of a reduction in platelet transfusion. No previous trials have had stroke as a primary outcome, although an application for such an essential study has recently been submitted to the National Institutes of Health (Bennet-Guerrero E., personal communication, 2006). APROTININ HAS AN ACCEPTABLE SAFETY PROFILE As with any therapy, the benefit of aprotinin administration must be weighed against the potential risks of receiving aprotinin (i.e., safety) and the possible loss of benefit from not receiving aprotinin. Hypersensitivity Reactions The incidence of adverse hypersensitivity reactions to aprotinin on initial exposure is <0.1% (8,9). The overall incidence of adverse reactions to aprotinin upon re-exposure is 2.7%; 5% if re-exposure occurs in <6 months and 0.9% if more than 6 months. Although unquantified against the risks of not receiving aprotinin, the risk of re-exposure to aprotinin at least 6 months after initial exposure is likely less than the risks of avoiding aprotinin in reoperative higher-risk cardiac surgery (10). Thrombosis Anecdotal incidents of thrombosis have been reported with all antifibrinolytic drugs (3,11,12). There has been speculation that these drugs have “prothrombotic” effects which may lead to adverse events, such as catheter thrombosis, graft occlusion, stroke, and myocardial infarction (MI) (3,13). However, aprotinin also inhibits prothrombotic pathways. Aprotinin’s nonspecific serine protease activity inhibits plasmin, thrombin, trypsin, and tissue and plasma kallikrein. Both directly and indirectly, aprotinin prevents activation of the tethered-ligand platelet thrombin receptors (PAR-1,3,4) (14) and the platelet glycoprotein receptors (GPIbIIa and GPIIbIIIa). Aprotinin also has well delineated anticoagulant effects mediated by inhibition of the intrinsic pathway (15). In addition, aprotinin inhibits activated protein C, thus inhibiting degradation of activated factor V (16) endothelial nitric oxide (17) and eicosanoid synthesis (18). Although prothrombotic mechanisms have been invoked as being a cause for an undefined prothrombotic tendency after its use during cardiac surgery, such an effect has not been demonstrated in clinical trials, meta-analyses, or clinical data submitted for regulatory approval. Early reports of increased graft and other thromboses may have been due to inadequate heparinization in the presence of aprotinin, and the consequent misinterpretation of an “adequate” activated clotting time. The best example of this effect is the site-specific difference in vein graft patency in the IMAGE trial (19). The overall differences in graft patency in the aprotinin group were seen only at two of the 13 sites, where heparinization may have been inadequate because of misinterpretation of aprotinin-induced activated clotting time prolongation. Graft patency rates in the 11 United States (US) sites were not different between groups (9.4% vs 9.5%; P = 0.72). In complete contrast, several randomized trials (20–25) and meta-analyses (5,6) of aprotinin versus placebo have failed to find an increased risk of graft thrombosis or enzyme-diagnosed MI (Table 1). In fact, the evidence to suggest that full-dose aprotinin may prevent cerebrovascular events is directly opposite to what would be expected from a prothrombotic drug (26).Table 1: Selected Summary of Meta-analyses and Studies Using Propensity ScoringAnother explanation for the potential association of aprotinin with thrombosis in some studies may involve inappropriate coadministration of antifibrinolytic drugs. Animal studies have demonstrated increased fibrin deposition in the kidney with combined administration of two antifibrinolytic drugs. Three observational studies have reported coadministration of aprotinin and EACA in a cardiac surgical population (3,27,28) with increased risk of renal dysfunction, MI, or death. Coadministration of antifibrinolytics was likely driven by bleeding during surgery, thus increasing overall risk of adverse events. A similar clinical scenario may have been present in many patients reported by Mangano et al. (1) who had received both aprotinin and another antifibrinolytic. APROTININ AND RENAL FAILURE Aprotinin and the lysine analogs are excreted by the kidney (29,30). Aprotinin is taken up by the epithelial cells of the proximal tubules, metabolized to small peptides or amino acids, and eliminated in urine (31). Aprotinin clearance is reduced and its half-life is prolonged in patients with renal dysfunction undergoing CPB (32). By contrast, EACA and TEA are concentrated and rapidly excreted unchanged in the urine. Animal studies have shown an inconsistent effect of aprotinin upon renal function that is difficult to translate to human cardiac surgery with CPB, especially because many of the animal studies used very large doses of aprotinin. Aprotinin produces kallikrein inhibition (33), which can inhibit the release of lysyl-bradykinin with attenuation of renal vasodilation. In normovolemic animals, aprotinin has no effect on renal hemodynamics or filtration (34). Small human studies have examined detailed biomarkers of inflammation and renal injury, while large multicenter trials have looked for broad population-based evidence of renal dysfunction, as evidenced by increases in creatinine or new requirement for dialysis. Full-dose aprotinin has no effect on renal plasma flow, glomerular filtration rate, electrolyte excretion, and eicosanoid excretion in patients undergoing CABG with CPB (35). However, in an earlier study, patients undergoing valve surgery, full and half-dose aprotinin compared to placebo were associated with a higher incidence of transient renal dysfunction or dialysis when compared with those who received placebo (36). However, some well-known important predictors of renal dysfunction, such as diabetes, were not equally distributed across all three groups in this study. In addition, an independent predictor of renal events was the preoperative use of angiotensin converting enzyme (ACE) inhibitors. Another study has suggested that there may be a deleterious interaction of ACE inhibitors and aprotinin upon renal function, when neither drug alone has any effect (37). This interaction may be especially relevant to the current debate, because ACE inhibitors are more frequently prescribed in patients with heart failure, an independent risk factor for renal failure. It would be valuable to examine this specific interaction in existing datasets from other well-conducted randomized trials. Overall, there is as yet no convincing published evidence from randomized trials or meta-analyses that aprotinin alone is a risk factor for significant renal dysfunction or failure. In the surgical population at highest risk of renal failure, those undergoing deep hypothermic circulatory arrest, aprotinin has not been associated with renal dysfunction or failure (38). The overlapping, but well-performed meta-analyses of Henry et al. (6) and Sedrakyan et al. (5) fail to show any adverse effect of aprotinin on renal, or other organ function (Table 1). RECENT OBSERVATIONAL STUDIES USING PROPENSITY SCORING Two recent observational studies have intensified the discussion about the safety of antifibrinolytics in cardiac surgery (1,2). Using propensity-adjusted multivariable logistic regression from observational data on 4374 patients, Mangano et al. (1) reported an association between use of aprotinin and adverse renal, cardiac, and neurologic outcomes (Table 1). No such association was identified with either EACA or TEA. All three drugs reduced bleeding in patients when compared with those not receiving an antifibrinolytic. Karkouti et al. (2) compared 449 aprotinin-treated patients with 449 propensity-matched TEA-treated patients drawn from a single center database of more than 10,000 patients. They found an association of aprotinin use with renal dysfunction alone (Table 1). Propensity matching can be a useful technique for minimizing differences between populations to examine the effect of a nonrandomly administered therapy. It is beginning to be widely used in the surgical literature because many therapeutic interventions cannot, for ethical reasons, be randomly assigned (39). Although propensity matching studies may provide evidence of association, at best, post hoc propensity matching studies should be considered as hypothesis generating. Thus, it is not surprising that a recent study of highly cited clinical research found that nonrandomized studies were subsequently contradicted or had initially stronger effects significantly more often than randomized trials (40). Accordingly, these two articles from Mangano et al. (1) and Karkouti et al. (2) should be evaluated according to the source population used, the outcome definitions used, and the propensity analyses conducted. WHY THE SOURCE POPULATIONS DETERMINE THE RESULTS? Because both studies were observational with uncontrolled decisions about which antifibrinolytic was used, there were large baseline differences between treatment groups in the original source populations. Not surprisingly, patients who received aprotinin had much higher frequencies of well-known risk factors for all adverse outcomes. These risk factors included preoperative organ dysfunction, severity, and the type of disease. Different methods were used to deal with these differences between the two studies (1,2). Mangano et al. (1) used a subset of patients from a cohort (Epi II) of more than 5000 patients enrolled in 70 institutions in 17 countries, predominantly in the late 1990s. Patients were excluded if they received an “inadequate dose of antifibrinolytic” that included an EACA dose of ≤10 gm, more than one antifibrinolytic, or did not have drug or dose information. These 691 patients had more frequent mortality (approximately 7.2%) and higher use of EACA (approximately 54%) than in the examined subset. The 226 patients who received multiple antifibrinolytics may well have received “rescue” aprotinin because of intraoperative difficulties. Exclusion of these 691 patients may have biased the results (41). In the Mangano et al. article, the propensity score was added as another variable to the multivariable logistic regression for each outcome (1). Furthermore, in the Epi II cohort, there was previously reported significant variability in the use of aprotinin among institutions participating in the cohort, ranging from 6% to 69%. (42) The use of other cardioprotective drugs such as aspirin and β-blockers also varied among institutions and countries (42). Similarly, significant inter-institutional variability in morbidity and mortality was observed in this cohort (42). Surprisingly, cardiac morbidity using a composite outcome of MI, congestive heart failure, and cardiac deaths occurred with markedly different incidences among countries, ranging from 9.2% in the United Kingdom to 18.5% in Germany, with mid-range figures for the US and Canada (P < 0.001) (42). In-hospital mortality also varied significantly among countries: 3.8% in Germany, 1.5% in the United Kingdom, 2.7% in the US, and 2.0% in Canada (P < 0.05) (42). The degree to which institutional practice differences, versus drug effect per se, impacted on outcome has not yet been published. Many other questions and criticisms about the New England Journal of Medicine article could also be obviated by the provision of supplemental information regarding the methods and data on the journal’s Web site. In contrast, the source population for the Karkouti et al. (2) study came from a single institution with a limited number of surgeons and anesthesiologists. They used the propensity score as an attempt to match individual patients receiving aprotinin to a “control” patient who received TEA, based on the propensity score. They achieved 94% success in matching patients, but were forced to exclude 24% of patients who could not be matched to TEA controls from the pool of almost 10,000 patients. This subject loss highlights the marked differences between patients who received aprotinin, or did not. Although there are no inter-hospital differences, there were still differences among individual practitioners and drug use over time, which may not have been totally accounted for (2). There were institutional guidelines for aprotinin. However, aprotinin use increased over the time period of the study, whereas TEA use decreased. In addition, no patients received aprotinin in the first year of the study. Thus, practioners were likely administering aprotinin to patients who they thought were at high risk for bleeding. No matter how they attempted to control for covariates, the selective use of aprotinin, in a time-varying fashion, in a very high risk population increases the potential for, and effect of, any unmeasured confounder to influence the outcome. THE IMPORTANCE OF OUTCOME DEFINITIONS The outcome definitions and event rates from the two studies are different. The definition of a cardiovascular event in the Mangano et al. article is significantly broader when compared with earlier publications from the same database (43–45). This produced a doubling of the cardiac event rate because of the reclassification of MI which was only by electrocardiogram changes (Q waves or new ST or T-wave changes), without a necessarily associated enzyme increase. This increased the number of reportable MIs approximately twofold, perhaps in a fashion that selectively increased the MI rate in a higher risk aprotinin group. Karkouti et al. used a definition of renal dysfunction of a 50% increase in baseline creatinine concentration provided it exceeded 1.13 mg/dL in women or 1.24 mg/dL in men, or a new requirement for dialysis (2). This definition is very liberal, resulting in a 24% rate of renal dysfunction in the aprotinin group and 17% rate in the control group, compared to the 2%–5% rates in Mangano et al. ’s groups (1) and other reports. Most clinicians would consider a small transient increase in serum creatinine as clinically insignificant. A much more restrictive outcome definition increases the clinical relevance, but resulted in lack of statistical significance. Using similar propensity matching techniques and data from the same database, Karkouti et al. have reported that recombinant factor VIIa, another hemostatic agent, was associated with an even larger (2.4-fold) increased incidence of renal dysfunction (46). In their current analysis of antifibrinolytic drugs, over 62% of rVIIa-treated patients developed renal dysfunction (2). Interestingly, they concluded that “definitive safety information … can only be achieved through adequately powered randomized controlled trials” (46). WHY ARE THESE FINDINGS DIFFERENT FROM RANDOMIZED AND The well-conducted meta-analyses by Henry et al. (6) of the and Sedrakyan et al. (5) found no deleterious effects of aprotinin and a reduction in cerebrovascular events, in direct to the of Mangano et al. and Karkouti et al. (1,2). did these meta-analyses and the trials not find these The effect and the event rates reported by Mangano et al. in primary CABG surgery can be used to the meta-analyses should have identified these Using the reported effect of aprotinin upon (OR = renal (OR = cardiac (OR = and events (OR = in the Mangano et al. study (1) to and study in randomized trials, the of the patients reported by Sedrakyan et al. (5) to examine for renal, cardiac, and outcomes would have been and at the for The of the marked between the effect and effect of the Mangano et al. study and this well-conducted are there was any adverse effect of aprotinin on these the results should have been the randomized trials, especially given the large number of trials the data were from trials per and the of the methods Mangano et al. suggested that the by Sedrakyan et al. was biased by the of a of The that the can be by the complete provision of all and trial data by to the and for independent WHY data on safety of antifibrinolytic drugs should be provided from an ongoing study, which is the largest controlled trial of antifibrinolytic drugs in high-risk cardiac The BART study is a clinical trial to aprotinin is to EACA and TEA in decreasing postoperative bleeding in high-risk cardiac surgery Patients are randomized to either aprotinin, TEA, or The primary outcome is bleeding and outcomes organ failure including renal, and cerebrovascular events. The of risks and be of by present in observational The trial is review and not by than high-risk cardiac surgical patients have been enrolled across 18 the patients included in the first combined and had independent data and safety recently the data and and of the trial without in The of patients an to a reduction in the primary bleeding outcome from 6% to but to a reduction in other clinical events, such as with a frequency of It is that the current rates of renal dysfunction, stroke, and MI in this trial are these event Thus, this and other trials provide important new information on the safety and efficacy of antifibrinolytic drugs well the of observational trials with large baseline differences in the

Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.