To the Editor:
We read with interest the recent article “Transfusion Requirements in Surgical Oncology Patients,” by Pinheiro de Almeida et al.1 Much attention has been drawn to this study as exemplified by an accompanying editorial2 and an illustration,3 which highlight its results. We would suggest that some caution be applied when interpreting the results of this study.
Recently, there have been numerous randomized trials of restrictive versus liberal transfusion thresholds in a variety of different patient populations.4–6 With the exception of a small pilot study in patients experiencing acute cardiac events,7 all have demonstrated that patients in the restrictive groups have not experienced worse outcomes than patients in the liberal groups.5,8 So the question arises as to why the work of Pinheiro de Almeida et al.1 showed the opposite effect. Have they identified a novel patient population that might benefit from a higher hemoglobin concentration? Perhaps, but the results of this study are not straightforward.
The patients in this study were randomized to a restrictive or liberal transfusion strategy when they were admitted to the intensive care unit (ICU) following their abdominal cancer surgery. The transfusion thresholds, less than 7 g/dl or less than 9 g/dl, for the restrictive and liberal groups, respectively, only applied when the patients were in the ICU—a median of 4 days for patients in both groups. During their time in the ICU, only 21 of 101 patients (21%) in the restrictive group and 41 of 97 patients (42%) in the liberal group actually received even a single transfusion. Furthermore, among the small number of patients in both groups who were actually transfused in the ICU, the median number of erythrocyte units administered to the patients who were actually transfused was not significantly different! (Median, 1 unit in the restrictive group; median, 2 units in the liberal group; P = 0.17.) Thus, although the authors randomly allocated patients to two different transfusion threshold groups, this randomization did not actually produce two different patient populations to study as both groups received statistically the same number of erythrocyte units in the ICU. In other words, although there was a plan to administer erythrocytes based on a liberal or a restrictive transfusion threshold, thereby creating highly and not highly transfused groups, the end result was that there was not a meaningful separation of the two groups in terms of the number of units transfused per patient. Thus, any outcome differences between these two groups cannot be explained by differences in the amount of erythrocytes transfused to patients in these groups.
The absence of a difference in the number of erythrocyte units transfused to those patients who actually received one continued once the patients arrived on a regular ward. Outside the ICU, the transfusion threshold strategy, determined by randomization, no longer applied; the decision to transfuse was left to the discretion of the treating physicians who appeared to use a relatively restrictive transfusion threshold as the mean pretransfusion hemoglobin concentration in both groups of patients in the wards was identical at 7.5 g/dl. Only 33 of 101 patients (33%) in the restrictive group and 47 of 97 patients (48%)in the liberal group actually received a transfusion by the end of their hospitalization. As the total number of erythrocyte units transfused during their hospitalization to patients in the restrictive and liberal groups was 88 and 134, respectively, the average number of erythrocyte units that were administered to the patients who were actually transfused was 2.67 and 2.85 units, respectively. This difference is not clinically meaningful. So neither at the end of the ICU admission nor at the end of the hospitalization did patients in the liberal group receive more erythrocyte units per patient than those in the restrictive group. Thus, the differences in patient outcomes observed in this study cannot be ascribed to one group having received more (or fewer) erythrocyte units than the other.
It is interesting to note that statistical P values were not presented in tables 1 and 2 to accompany the presentation of the patient demographic information. Although patients in the two groups appear similar, subtle differences might also have influenced some of the outcomes. One of the adverse events included in the primary outcome measure was major cardiovascular complications, itself a composite outcome. There were no differences between the patients in the two groups in terms of peripheral arterial disease, cerebrovascular disease, congestive heart failure, and arterial coronary disease when compared individually. However, when the patients within each group with these four comorbidities are added together, and this composite metric of cardiovascular disease is compared between the two groups, there is certainly a trend (P = 0.0636) toward a higher incidence of cardiovascular disease among the patients in the restrictive group. This trend toward having a higher incidence of cardiovascular disease among patients in the restrictive group could certainly explain at least in part the overall higher incidence of cardiovascular complications observed in these patients.
Overall, the fact that there was no difference in the number of erythrocytes administered to the small number of patients who received a transfusion in both groups indicates that the lower mortality in the liberal group cannot be explained by the quantity of erythrocytes transfused to the patients who actually received them. Confounders like the higher incidence of abdominal sepsis and more patients with lower gastrointestinal tumors in the restrictive group can better explain the differences in outcomes between the groups than the quantity of transfused erythrocytes per patient. These data do not support preoperative erythrocyte transfusion for anemic patients undergoing cancer surgery; if a higher preoperative hemoglobin concentration is desired, then consider iron therapy.
The authors declare no competing interests.