Response to Letter Regarding Article, “Increased Mortality, Postoperative Morbidity, and Cost After Red Blood Cell Transfusion in Patients Having Cardiac Surgery”
In our registry, treatment with aprotinin is recorded as “in theater” or subsequently. Aprotonin treatment (none, in theater, or subsequently) was included in the propensity model, although the frequencies of patients given aprotinin (9% in theater, 1% subsequently) were not reported.1
New analyses, explicitly including aprotinin treatment (in addition to nadir hematocrit stratum and propensity score deciles), show that the propensity score controlled well for confounding by aprotinin; odds ratios (ORs) for composite infection and ischemic outcomes given transfusion were 3.31 (95% confidence interval, 2.59 to 4.24, estimated by bootstrapping) and 3.29 (95% confidence interval, 2.59 to 4.26), respectively, compared with 3.38 and 3.35 without aprotinin explicitly included in the models.1 Therefore, aprotinin did not substantially confound the reported risks of the composite outcomes. Timing of aprotinin treatment did not markedly influence the risk of either the infection or ischemic composite outcome, so, in subsequent analyses, aprotinin treatment was fitted as a binary variable.
The Blood Conservation using Antifibrinolytics: A Randomized Trial in High-Risk Cardiac Surgery Patients (BART) Data Safety Monitoring Board reported that mortality was higher in patients given aprotinin who also experienced severe bleeding but “felt that this imbalance was likely due to chance.” (A. Laupacis, MD, MSc, FRCP, and the BART Data Safety Monitoring Board, written communication to Drs Paul Hébert and Dean Fergusson, October 16, 2007). However, if severe bleeding in the BART trial was a surrogate marker for high-volume blood transfusion, then it might be the combination of transfusion and aprotinin that was particularly harmful. Therefore, we also investigated whether an interaction between aprotinin and transfusion was present in our cohort. The official report of the BART has now been published.2
For the composite infection outcome, no interaction was present (P=0.57). However, for the composite ischemic outcome, the interaction was statistically significant (P=0.03). In patients who were not transfused (of whom 6.5% were given aprotinin), aprotinin was not associated with a higher risk of an ischemic event (OR =0.79, 95% confidence interval, 0.24 to 1.55). Among patients who were transfused (of whom 12.1% were given aprotinin), aprotinin was associated with a higher risk of the ischemic outcome (OR =2.19, 95% confidence interval, 1.79 to 2.70).
These latter models did not take into account the number of units of red cells transfused. We have already shown strong dose-response relationships for both outcomes.1 Since the amount of blood transfused differed substantially in transfused patients who were and were not treated with aprotinin (mean 6.3 U versus 3.4 U), we added aprotinin treatment, and the interaction of aprotinin with units of red cells transfused, to models previously reported, which included the interaction of nadir hematocrit stratum and units of red cells transfused.1 This interaction is analogous to those estimated above, except that it also takes into account the units of blood transfused.
In these analyses, no evidence was found of an interaction between units of red cells transfused and aprotinin for either outcome (infection, P=0.26; ischemic event, P=0.39). This suggests strongly that the interaction reported above between any transfusion and aprotinin treatment for the composite ischemic outcome arose from confounding by the units of red cells transfused. In models without the interaction, aprotinin was associated with increases in the risk of both outcomes (infection, OR =1.56, 95% confidence interval, 1.24 to 1.95; ischemic event, OR =1.45, 95% confidence interval, 1.14 to 1.81).
In summary, the estimates originally reported were not substantially confounded. Once the risk from increasing units of red cells are taken into account, no interaction can be found between aprotinin and transfusion.3 Aprotonin is associated with both composite outcomes, but this could be because aprotinin treatment is strongly correlated with other prognostic factors.
Source of Funding
The study on which this letter is based was supported by the British Heart Foundation.
Murphy GJ, Reeves BC, Rogers CA, Rizvi SI, Culliford L, Angelini GD. Increased mortality, postoperative morbidity, and cost after red blood cell transfusion in patients having cardiac surgery. Circulation. 2007; 116: 2544–2552.
Ferguson DA, Hébert PC, Mazer CD, Fremes S, MacAdams C, Murkin JM, Teoh K, Duke PC, Arellano R, Blajchman MA, Bussières JS, Côté D, Karski J, Martineau R, Robblee JA, Rodger M, Wells G, Clinch J, Pretorius R; BART Investigators. A comparison of aprotinin and lysine analogues in high-risk cardiac surgery. N Engl J Med. 2008; 358: 2319–2331.