Clinical Outcomes Following Institution of the Canadian Universal Leukoreduction Program for Red Blood Cell Transfusions

Context A number of countries have implemented a policy of universal leukoreduction of their blood supply, but the potential role of leukoreduction in decreasing postoperative mortality and infection is unclear.

Objective To evaluate clinical outcomes following adoption of a national universal prestorage leukoreduction program for blood transfusions.

Design, Setting, and Population Retrospective before-and-after cohort study conducted from August 1998 to August 2000 in 23 academic and community hospitals throughout Canada, enrolling 14 786 patients who received red blood cell transfusions following cardiac surgery or repair of hip fracture, or who required intensive care following a surgical intervention or multiple trauma.

Intervention Universal prestorage leukoreduction program introduced by 2 Canadian blood agencies. A total of 6982 patients were enrolled during the control period and 7804 patients were enrolled following prestorage leukoreduction.

Main Outcome Measures All-cause in-hospital mortality and serious nosocomial infections (pneumonia, bacteremia, septic shock, all surgical site infections) occurring after first transfusion and at least 2 days after index procedure or intensive care unit admission. Secondary outcomes included rates of posttransfusion fever and antibiotic use.

Results Unadjusted in-hospital mortality rates were significantly lower following the introduction of leukoreduction compared with the control period (6.19% vs 7.03%, respectively; P = .04). Compared with the control period, the adjusted odds of death following leukoreduction were reduced (odds ratio [OR], 0.87; 95% confidence interval [CI], 0.75-0.99), but serious nosocomial infections did not decrease (adjusted OR, 0.97; 95% CI, 0.87-1.09). The frequency of posttransfusion fevers decreased significantly following leukoreduction (adjusted OR, 0.86; 95% CI, 0.79-0.94), as did antibiotic use (adjusted OR, 0.90; 95% CI, 0.82-0.99).

Conclusion A national universal leukoreduction program is potentially associated with decreased mortality as well as decreased fever episodes and antibiotic use after red blood cell transfusion in high-risk patients.

Over the past decade, several studies have suggested that blood transfusions depress immune function in recipients.^1,2 Evidence of transfusion-associated immune suppression emerged following observations that blood transfusions improved renal allograft survival³ and accelerated⁴ and increased postoperative infections.⁵

A recent randomized controlled trial undertaken to examine infections in cardiovascular surgical patients found an approximately 4.2% absolute decrease in mortality but no decrease in infections among patients receiving leukoreduced blood, compared with patients receiving buffy coat–depleted blood.⁶ A second trial conducted by the same investigators designed to evaluate mortality documented a similar decrease in 30-day mortality in this same patient population.⁷ These investigators postulated that depressed immunity following blood transfusions predisposed high-risk cardiovascular surgical patients to multiple organ failure and ultimately resulted in higher mortality. However, recent meta-analyses and reviews^8,9 of the randomized trials do not provide convincing evidence for or against the potential role of leukoreduction in decreasing mortality or postoperative infections.

Given the current debate on the effectiveness of leukoreduction,¹⁰ we conducted a large national study designed to determine the association of leukoreduction on rates of in-hospital death and serious nosocomial infection in a high-risk postoperative population receiving blood transfusions.

In this before-and-after retrospective cohort study, conducted from August 1998 to August 2000, we collected information on 14 786 postoperative patients from 23 Canadian academic and community hospitals representing all regions of the country. The study was designed to detect a clinically meaningful 20% relative decrease (1% absolute difference) in rates of in-hospital death and serious nosocomial infection.

Universal prestorage leukoreduction was regionally implemented across Canada between June and September 1999. Patients in the control period were admitted to the hospital in the period commencing 373 days prior to the date of implementation of leukoreduction and ending 7 days before this date. Patients in the intervention period were admitted to the hospital in the period commencing 60 days after the implementation date of universal prestorage leukoreduction. Both intervention and control cohorts consisted of 2 complete 1-year periods interrupted by a 67-day washout period, adopted to minimize contamination. This study was approved by the research ethics committees at all participating institutions and at the coordinating center.

This study targeted surgical patient populations that consumed as much as 40% of the total blood supply¹¹ and who were considered at high risk of death or developing serious bacterial infections.¹² Based on these criteria, we enrolled all consecutive patients in 3 distinct high-risk categories: (1) patients following cardiovascular surgery requiring cardiopulmonary bypass; (2) patients requiring intraoperative repair of a hip fracture; and (3) postoperative and multiple trauma patients admitted to an intensive care unit. We excluded patients who were younger than 16 years; had a serious infection prior to receiving blood; had been previously included in this study; had no ongoing commitment to the provision of all necessary care because of a terminal illness or the patient's expressed wishes (eg, do-not-resuscitate order); were considered brain dead within the first 24 hours of hospital admission; did not survive 24 hours following the completion of the index surgical procedure or time of admission to the intensive care unit; received treatment for a hematologic malignancy in the past year or had undergone bone marrow transplantation; and who had received at least 1 blood transfusion in the year prior to the time of hospital admission.

Canadian Blood Services and Héma Québec were the only agencies in Canada that collected, processed, and provided blood products to hospitals during this study. Donated blood was collected into CP2D anticoagulant solution and stored in 100 mL of Nutricel additive during this evaluation.^13,14 Leukofiltration (Pall Medical, Blood Processing Group, East Hills, NY) using these systems reduced white blood cell content of a unit of red blood cells from an average of 3.0 × 10⁹ per unit to 2.5 × 10⁵ per unit, a decrease of 4 logs. Quality control measures were conducted by both blood services and the leukofilter manufacturer according to regulatory standards.

Both all-cause in-hospital deaths and confirmed serious nosocomial infections were considered to be primary outcomes. Serious nosocomial infections included pneumonia, bacteremia, and septic shock, as well as all surgical site infections (Box). Outcomes must have occurred after the first blood transfusion and at least 2 days after the index procedure or intensive care unit admission. The diagnosis of confirmed nosocomial pneumonia was based on stringent criteria developed by Johanson et al¹⁵ and Toews.¹⁶ Bacteremia was defined as the identification of a recognized pathogen isolated from a blood culture specimen. The definition of septic shock required evidence of a systemic inflammatory response and hypotension that was unresponsive to fluid resuscitation and acute organ hypoperfusion manifested by lactic acidosis, oliguria, and confusion.^17-19 Surgical site infections included deep incision infections and organ or surgical site infections.^20,21 For cardiac surgical procedures, we documented postsurgical major infections including mediastinitis, endocarditis, myocarditis, or pericarditis. Similarly, following intraoperative repair of hip fractures, we specifically sought to identify each episode of postoperative osteomyelitis, septic arthritis, or infected prosthesis. All organ system infections met US Centers for Disease Control and Prevention criteria.^17,20-22

Pneumonia^15,16

(1) New and progressive pulmonary infiltrate on sequential chest radiographs
(2) Temperature >38°C
(3) Total white blood cell count >12 000 cells/mm³ or >10% bands on differential cell count
(4) Purulent tracheobronchial secretions (moderate numbers of organisms and polymorphonuclear cells with a few epithelial cells on microscopic examination of tracheal aspirate)
A confirmed diagnosis of infection required all 4 criteria, while a suspected diagnosis did not require number 4.

Bacteremia/Severe Sepsis

(1) Identification of a recognized pathogen isolated from a blood culture specimen. For commensal skin organisms, at least 2 positive blood cultures collected on separate occasions or venipuncture sites were required
(2) Temperature >38°C or hypotension, defined as a systolic blood pressure either <90 mm Hg, or 40 mm Hg lower than baseline values

Septic Shock^17-19

(1) Systemic response to infection, including temperature >38°C or <36°C, heart rate >90/min, respiratory rate >20/min, CO₂ partial pressure <32 mm Hg, or an increased white blood cell count >12 000 cells/mm³ or <4000 cells/mm³
(2) Hypotension unresponsive to fluid resuscitation
(3) Acute organ hypoperfusion manifested by lactic acidosis, oliguria, and confusion

Surgical Site Infection^20,21

Deep Surgical Site Infections

Involvement of fascia or muscle layers documented by the presence of at least 1 of the following criteria: purulent drainage from a deep incision; spontaneous dehiscence of a deep incision or required surgical debridement because of a Temperature >38°C; localized pain and inflammation; abscess or other evidence of deep incision infection observed on direct examination, reoperation, or radiologic examination

Organ/Space Infections*

Purulent drainage from a sterile drain placed into the designated organ or space was documented by identification of organisms isolated from aseptic culture of fluid or tissue from the organ or space; abscess or other evidence of organ space infection observed through direct surgical examination or imaging study of the organ or space involved

*All organ/space infections met US Centers for Disease Control and Prevention criteria.^17,20-22

Secondary outcomes included an examination of item-specific criteria for all infections and each category of infection. We were also interested in the rates of fever, defined as a temperature exceeding 38.5°C and use of antibiotics for the treatment of serious infections. We also evaluated whether universal prestorage leukoreduction impacted the duration of organ support (respiratory support based on the number of days of mechanical ventilation, hemodynamic support based on the number of days requiring vasoactive drug, and renal support based on the number of days of dialysis dependence) as well as length of hospital and intensive care unit stay.

All data were abstracted from patient medical charts using standardized case report forms and detailed procedures manuals. Personnel performing data abstraction were given a dummy protocol aimed at masking the true intent of the project. All data collection was undertaken in concurrent prespecified monthly intervals in both 365-day observation periods to minimize potential information bias related to differential learning curves. To ensure data quality, all personnel completed a training session and participated in a quality assurance exercise in which data were abstracted from a standardized medical record. From this evaluation, we documented an overall accuracy rate exceeding 90% (97.5% for primary outcomes) compared with a criterion standard. In addition, experienced research personnel compared 10% of case report forms to the medical records. Once received by the coordinating center, all case report forms were manually reviewed for data quality and completeness. Each case report form was electronically scanned into a computerized TELEform database (Version 6.0, Cardiff Software Inc, Vista, Calif) that included range and logic checks. Queries were sent to centers following all manual and electronic quality checks. A minimum set of data including hospital mortality and procedure was collected on all nontransfused patients during the same time period.

We compared all major baseline variables before and after the implementation of the leukoreduction program with absolute differences and 95% confidence intervals (CIs). The effect of leukoreduction on the rates of in-hospital mortality and all confirmed serious nosocomial infections were calculated using χ² statistics and unadjusted odds ratios (ORs) with 95% CIs. Mortality rates in nontransfused patients were also compared between time periods using the same statistical techniques.

Given the possibility of differences in patient characteristics and therapeutic interventions between treatment periods, logistic regression procedures were used to calculate adjusted ORs for rates of in-hospital mortality and serious nosocomial infections. Variable selection for the multivariate models involved a predefined 2-step process. First, we examined a series of variables known to be related to mortality and bacterial infections based on clinical or biological relevance within the following categories: demographic information (age, sex, and center), major comorbid illnesses (14 major illnesses), medications used in the first 24 hours of care (13 major categories of medications), major disease categories (cardiac surgery, repair of hip fracture, or intensive care), previous transfusions (yes/no), and the number of blood transfusions (≤3 vs >3). Second, any variable with an unequal distribution between treatment groups (>1% absolute difference) at baseline was added to the model. To understand the influence of each potential confounder on mortality, we also used a Mantel-Haenszel χ² analysis including the study intervention and each variable. The final model included age, sex, center, comorbid illness (severe lung disease), medications (aspirin, β-blockers, and angiotensin-converting enzyme inhibitors), major diseases, previous transfusion(s), and the total number of transfusions as well as the treatment period. To evaluate the influence of secular trends on hospital mortality, we plotted the adjusted ORs using the first period as the reference category for twelve 2-month intervals. Multivariate models using all variables except center also were generated within quartiles of numbers of units transfused (1 unit, 2 units, 3 to 4 units, and 5 units or more).

We used an identical approach to evaluate the impact of leukoreduction on secondary outcomes including fever and antibiotic use. As a secondary analysis, we compared the effect of leukoreduction on specific categories of infections. All infections were grouped into 3 clinically meaningful categories: pneumonia, bacteremia and septic shock, and surgical site infections (ie, all deep incision infections or organ space infections). Furthermore, several prespecified definitions were used to understand how various criteria affected inferences related to leukoreduction. In order of clinical significance, the definitions used were (1) confirmed infections for which all criteria were met; (2) suspected infections for which 1 criterion was not fulfilled; and (3) a physician's diagnosis of infection documented in the medical record. Secondary analyses also included an evaluation of fever episodes and the use of antibiotics for the prespecified serious infections. All secondary analyses used the same analytic plan and choice of variables in multivariate analyses as described in the analyses of primary outcomes.

Bivariate and multivariate procedures were used to compare all outcomes in the 3 predefined major disease subgroups of cardiac surgical disease, hip fracture repair, and critical care. Odds ratios and 95% CIs were reported for all outcomes. No adjustments were made for multiple comparisons. All analyses were conducted using SAS v8.0 (SAS Institute Inc, Cary, NC); P<.05 was set as statistical significance.

During the study period a total of 14 786 patients received red blood cell transfusions and met all eligibility criteria; 7804 patients received universal prestorage leukoreduced blood products while 6982 patients were enrolled in the control period. During this same time interval, there were 26 183 patients who were not transfused: 12 927 in the leukoreduction period and 13 256 in the control period. All patients who met eligibility criteria identified through electronic search criteria at all participating institutions were included.

Overall, there were few important differences in baseline characteristics in transfused patients. Patients in the control period had a small but significant increase in the prevalence of severe lung disease. Patients who received leukoreduced blood were more likely to have been given aspirin, β-blockers, and angiotensin-converting enzyme inhibitors (Table 1). No other important differences in baseline characteristics between treatment groups were identified. The average minimum hemoglobin concentration in patients receiving leukoreduced blood was statistically but not clinically less compared with controls (7.46 [1.22] g/dL vs 7.51 [1.24] g/dL; mean difference, 0.55 g/dL; 95% CI, 0.15-0.95 g/dL; P<.001). The mean (SD) number of transfusions was similar, with an average of 3.8 (4.03) units given in the leukoreduction period compared with 3.9 (4.19) units given in the control period (mean unit difference, 0.06; 95% CI, −0.07 to 0.20 units; P = .35). In the subgroups, the mean (SD) number of units was 3.5 (3.45) vs 3.5 (3.36) units for cardiac surgical patients; 5.4 [5.58] vs 5.6 [6.02] units for critically ill/multiple trauma patients; and 2.6 [1.97] vs 2.5 [1.73] units for patients with hip fracture in leukoreduced vs control periods). The overall rate of transfusion was 50.7% vs 48.8% (−1.95% leukoreduced vs control; 95% CI, −2.80% to −1.09%).

Unadjusted hospital mortality rates were significantly lower following leukoreduction compared with the control period (6.19% vs 7.03%, respectively; OR, 0.87; 95% CI, 0.76-0.99; P = .04) (Table 2). The hospital mortality rate was 8.1% vs 8.5% (absolute difference, −0.41%; 95% CI, −1.08% to 0.25%) when nontransfused patients from the leukoreduction period were compared with controls. When adjusted, the odds of death did not change (OR, 0.87; 95% CI, 0.75-0.99; P = .04) (Figure 1). For each major disease subgroup, we observed nonsignificant decreases in the adjusted odds of death following critical care and trauma (adjusted OR, 0.94; 95% CI, 0.76-1.17; P = .57); following cardiac surgery (adjusted OR, 0.88; 95% CI, 0.72-1.07; P = .20); and following hip fracture repair (adjusted OR, 0.74; 95% CI, 0.49-1.09; P = .13).

A stratified analysis revealed that patients with severe lung disease had an unadjusted OR for mortality of 0.90 (95% CI, 0.77-1.28; P = .15) that decreased to 0.78 (95% CI, 0.51-1.19; P = .28) in patients without this comorbid condition. The use of cardiac medications including aspirin, β-blockers, and angiotensin-converting enzyme inhibitors all resulted in the unadjusted OR for mortality shifting from a significant to a nonsignificant association. After adjustment, individual medications were not independently associated with mortality (P>.05 for all). In terms of the number of transfusions, patients receiving 1 to 4 blood transfusions had an adjusted OR for mortality ranging from 0.72 to 0.79 (P>.05 for all) while patients receiving 5 units or more had an adjusted OR of 0.97 (95% CI, 0.81-1.17; P = .73). The influence of time on the adjusted OR for mortality is depicted in Figure 2. By including the additional 106 patients who died within the first 48 hours as a sensitivity analysis, the adjusted OR was 0.88 (95% CI, 0.77-1.01; P = .06). There were no important changes in adjusted ORs when all 51 explanatory variables were sequentially added to multivariate models.

There was no clinically important or statistically significant decrease in confirmed infections associated with leukoreduction (unadjusted OR, 0.93; 95% CI, 0.84-1.04; P = .20). Following multivariate adjustment, the OR for confirmed infections was 0.97 (95% CI, 0.87-1.09; P = .63) (Figure 1). For suspected infections, leukoreduction was associated with a slight decrease in events (unadjusted OR, 0.91; 95% CI, 0.83-0.99; P = .05). This association did not remain significant after multivariate adjustment (adjusted OR, 0.94; 95% CI, 0.85-1.04; P = .21). Using the physician's diagnosis of infection as a definition, the unadjusted OR was 0.91 (95% CI, 0.83-0.99; P = .05) but the adjusted OR of 0.94 (95% CI, 0.85-1.05) was not significant (P = .27).There were no detectable differences in subtypes of infections or the 3 major clinical subgroups (P>.05 for all)

The proportion of patients with fever episodes decreased from 24.7% prior to the introduction of the leukoreduction program to 22.5% following its implementation (unadjusted OR, 0.88; 95% CI, 0.82-0.95; P = .001). This clinically important and statistically significant decrease in the odds of developing a fever persisted following multivariate adjustments (adjusted OR, 0.86; 95% CI, 0.79-0.94; P<.001). The use of antibiotics also decreased following leukoreduction. The crude OR was 0.89 (95% CI, 0.81-0.97; P = .01) while the adjusted OR was 0.90 (95% CI, 0.82 to 0.99; P = .03). The decreases in the frequency of patients experiencing at least 1 episode of fever and in the use of antibiotics for serious infections were comparable in the major subgroups (Table 2 and Table 3).

In terms of other secondary outcomes, there were no major differences in average hospital length of stay (16.6 [16.7] vs 16.5 [16.8] days; P = .73) and intensive care unit (9.6 [14.2] vs 9.5 [14.1] days; P = .86) in comparing all patients in the preleukoreduction and postleukoreduction periods. There also were no important differences for the duration of patients requiring organ support. The mean (SD) duration of mechanical ventilation (9.7 [16.1] vs 8.9 [14.2] days; P = .34), hemodynamic support (3.2 [1.7] vs 3.4 [1.8] days; P = .32), and renal support (14.3 [15.3] vs 15.6 [19.8] days; P = .66) were similar from one period to the next. These trends were not altered when only survivors were considered in the analysis (P>.05 for all).

In this study, we documented a decrease in mortality associated with the implementation of universal leukoreduction without observed changes in serious infections. We also noted an important decrease in the proportion of patients experiencing at least 1 fever episode as well as an associated reduction in the use of antibiotics. Therefore, the filtration of leukocytes from donated blood appeared to be associated with important health benefits.

This before-and-after study was designed to detect an absolute decrease in both mortality and serious nosocomial infections in the range of 1%.^8,9 We reasoned that such small differences would be important, given that the results impact high-risk patients who collectively require a large portion of the blood supply and given that there are few adverse consequences other than cost. Indeed, assuming a 7% mortality rate in the control period, the decreased odds of death generated from this study would translate into 1 life saved for every 120 patients who receive leukoreduced blood, a number needed to be treated in the same order of magnitude as recent cardiovascular trials.²³ The observed improvement in mortality also was consistent among all subgroups and throughout a range of exposures to blood. These findings persisted after multivariate analysis, through a number of sensitivity analyses and demonstrated appropriate time trends. Our observations related to mortality and infections were similar to 2 studies in high-risk cardiac surgery conducted by van de Watering et al⁶ and Bilgin et al.⁷ The other studies documented an important absolute decrease in mortality between 4% and 5% in the 2 trials without consistent decreases in infections. The smaller difference in mortality rates noted in our study may be attributed to a more heterogeneous population of patients who may have received less blood.

Several recent studies concluded that leukoreduction did not affect major clinical outcomes.^24,25 In comparing 2780 patients receiving nonleukoreduced vs leukoreduced blood products, Dzik and colleagues²⁴ noted comparable in-hospital mortality rates (8.5% vs 9.0%, respectively; P = .64) but fewer fever episodes (0.77% vs 0.22%, respectively; P = .06). In contrast, we observed a decrease in mortality and in the use of antibiotics that may be explained by our selection of higher-risk patients compared with all hospitalized patients requiring a transfusion.

Our results suggest that the observed decrease in the number of deaths may not have been mediated through immune suppression and increased rates of serious infections. An alternative explanation is that transfused leukocytes result in a proinflammatory microvascular effect leading to important clinical consequences.^26-29

Given that rates of confirmed serious infections were not affected by the implementation of the leukoreduction program, the decrease in fever episodes may be predominantly related to a decrease in febrile nonhemolytic reactions. Physicians appear to have responded to lowered rates of fever by prescribing fewer antibiotics. Therefore, fever episodes in potentially unstable patients may result in lower costs of care.

Because the implementation of a universal leukoreduction program in Canada was mandated by the regulatory agency, the optimal experimental design was a before-and-after study. We took every precaution in our study to minimize the influence of information and selection biases, including objective outcomes, masking of data abstraction personnel, and standardized data collection procedures.³⁰ To ensure generalizability and to minimize the impact of secular trends, we selected patients undergoing several different high-risk procedures and enrolled patients from both community and academic centers. We noted few important differences in baseline characteristics or therapeutic interventions administered in the first 24 hours of acute care among patients treated in the control and leukoreduction periods. Differences at baseline, when detected, either had no effect on outcomes or shifted the OR toward the null. Despite careful attention to potential biases, our results may have been affected by secular trends and incomplete information.

Universal prestorage leukoreduction was associated with decreased mortality, number of fever episodes, and subsequent use of antibiotics in high-risk patients. The mechanism leading to these potential health benefits did not appear related to decreased infections. Although this study adds to the literature in support of the adoption of universal leukoreduction, additional data from clinical trials are needed to provide evidence for the efficacy of leukoreduction of red blood cell transfusions.