Erythrocyte Storage Duration Is Not Associated with Increased Mortality in Noncardiac Surgical Patients: A Retrospective Analysis of 6,994 Patients

• Transfusion of blood with a longer storage time has been associated with increased mortality in cardiac surgery patients, but studies outside this setting have been small and with mixed results

• In a study of nearly 7,000 patients receiving transfusions for noncardiac surgery, there was no relationship between median storage duration time and mortality

SINCE the discovery of circulation by Sir William Harvey, M.D. (Lumleian Lecturer, Royal College of Physicians and St. Bartholomew Hospital, London, England) (1578–1657) in the early 17th century, and the subsequent first successful intraoperative blood transfusion by George Washington Crile, M.D. (Professor of Clinical Surgery in the Western Reserve Medical College, Cleveland, OH) (1864–1943) at the Cleveland Clinic,1,2blood transfusions are important therapeutic agents. In the United States, more than 5 million patients receive erythrocytes every year.** Transfusions, specifically numbers of erythrocyte units, have been associated with severe adverse outcomes, including increased risk for infections,3–5pulmonary complications6, and cancer recurrence,7,8and prolonged intensive care unit4,5,9and hospital length of stay.5,9 

In recent years, the storage duration of allogeneic erythrocytes has been linked to the risk of severe postoperative complications, especially after cardiac surgery.10,11Time-dependent changes in the biochemical properties of stored erythrocytes collectively referred to as the “storage lesion,” have been proposed as the underlying mechanism.12,13These changes include alterations in the erythrocyte membrane and decreases in 2,3 diphosphoglycerate, nitric oxide, and intracellular potassium concentrations.14Storage lesion changes result in compromised in vitro  erythrocyte function and viability.12,15–17However, the effect of these observed changes on clinical outcomes of transfused patients remains unclear.

Some studies show a significant association between storage duration of erythrocyte units and adverse outcomes, including mortality, hospital length of stay, intubation time, infectious complications, multiorgan failure, and composites of adverse events.11,14,18,19Other studies show no association or only minor associations that were largely attributed to increased baseline risk in transfused patients rather than to the effects of older transfused cells per se .19–21Interpretation of these studies is complicated by use of different methodological and analytical approaches, not all of which adequately adjust for confounding factors and disease severity. Most studies performed to date were in the cardiac surgery, intensive care, and trauma populations. Limited data are available for noncardiac surgical patients.7,22,23We thus sought to evaluate whether or not storage duration of transfused erythrocytes is associated with postoperative all-cause mortality among noncardiac surgery patients using data derived from the Cleveland Clinic Perioperative Health Documentation System registry.

With approval from the Cleveland Clinic Institutional Review Board (Cleveland, OH), perioperative data corresponding to 63,319 adult, American Society of Anesthesiologists Physical Classification I–V, general surgery patients treated at Cleveland Clinic between January 6, 2005, and June 30, 2009, were obtained from our Perioperative Health Documentation System registry. Erythrocyte transfusion and product data were obtained from the laboratory information system. Patients receiving leukocyte-reduced allogeneic erythrocyte transfusions between 2 days before the date of surgery and 7 days afterward were included in the study. Only the most recent operation for each patient was analyzed. Postoperative mortality information in our Perioperative Health Documentation System registry was obtained by combining data from both hospital records and the U.S. Social Security Administration’s Death Master File. The U.S. Social Security Administration’s Death Master File was queried on June 15, 2011, matching patients on their social security numbers. Patients for whom long-term death information was not available from the Death Master File were censored at hospital discharge.

Erythrocyte storage duration as an exposure is complicated because patients potentially receive multiple units and thus are exposed to multiple storage durations but can experience only one mortality outcome. Koch et al .,11in a study evaluating the association between storage duration and mortality among cardiac surgery patients, created two groups by splitting the overall distribution of erythrocyte storage duration at the observed median of 14 days. This approach is useful for studying whether or not there is evidence of an overall association but does not provide an assessment of dose–response. With this latter goal in mind, we characterized the storage duration exposure using the median storage duration of all erythrocyte units transfused to a patient. But to keep the groups homogeneous, we excluded patients for whom the range of storage duration among transfused units exceeded 5 days.

First, we grouped patients into three storage duration groups based on whether their median storage duration was ⩽14 days, >14 days but ⩽28 days, or >28 days. Kaplan–Meier survival density function estimates were then obtained and plotted for each of the three groups.24Although this analysis is useful for visualizing patterns of survival, it does not adjust for potential confounding factors. Therefore, no testing for differences was undertaken in this exploratory analysis.

For the primary hypothesis, multivariable Cox proportional hazards regression was used to characterize the relationship between median erythrocyte storage duration and postoperative mortality rate over time. We adjusted for any patient and surgical characteristic plausibly influencing the storage duration of erythrocytes. Previous studies have been criticized for inadequately considering baseline parameters such as patient comorbidities. We cannot plausibly postulate a mechanism that would explain how older blood might get assigned to particular subgroups of patients. Blood banks are typically unaware of patient characteristics, and standard operating procedures require using the oldest available matching. We did adjust for blood type of the patient, Rhesus factor of the patient, number of units transfused, range of storage duration among the transfused erythrocyte units within a patient, and date of surgery.25 

First, we developed a Cox model whereby the relationship was modeled flexibly via  a smoothing procedure (specifically, penalized B -spline smoothing).26Covariate-adjusted mortality incidence estimates (at several fixed postoperative times: up to 2 yr postoperatively) were then obtained and graphed as a function of patient-median erythrocyte storage duration using this model. Nonlinearity of the relationship (between patient-median erythrocyte storage duration and the logarithm of the hazard rate) was tested for significance using a Wald chi-squared test. If no significant evidence of nonlinearity was found, we used a simpler linear model to describe the relationship.

Adequacy of the assumption of proportional hazards across postoperative time for our final model was assessed using various graphical and analytic methods (e.g. , a plot of log-cumulative hazards against log-time, Schoenfeld residual plots, and chi-squared tests for consistency of Cox model coefficients over time).27,28R statistical software version 2.1 2.1 for 64-bit Windows (The R Foundation for Statistical Computing, Vienna, Austria) was used for the statistical analysis. The Type I error rate for all hypotheses was fixed at 5%.

Of the 63,319 surgical patients included in our study, 10,090 (15.9%) were transfused with allogeneic erythrocyte products. We removed 314 patients who received one or more autologous erythrocyte units and an additional 669 who received allogeneic erythrocyte units that were not leukocyte-reduced. Of the remaining 9,107 patients, 2,113 (23.2%) received multiple units with storage duration ranging greater than 5 days; removing these patients, we analyzed perioperative data on 19,462 allogeneic erythrocyte units transfused to 6,994 patients (fig. 1).

As expected, patients included in the analysis were generally sick, with 80% being assigned an American Society of Anesthesiologists’ Physical Classification of III or above (table 1). Patients received anywhere between 1 and 41 erythrocyte units; 23, 44, 11, 9, and 13% of patients received 1, 2, 3, 4, and ≥5 erythrocyte units, respectively. The patient-specific median erythrocyte storage duration was largely representative of the age of the individual units, with 94% of the units’ storage duration being within ±2 days of the respective patient medians. Overall, there were 1,718 mortalities among the 6,994 patients analyzed (24.6%).

Estimated correlation (95% CI) between number of erythrocyte units transfused and median erythrocyte age was −0.10 (−0.13 to −0.08), indicating a slight relationship. Except for blood type and Rhesus factor—which are determinants of blood product supply—we found little evidence of relationship between patients’ baseline characteristics and median erythrocyte storage duration. A summary of the top 15 principal procedures and diagnoses, as characterized by the U.S. Agency for Healthcare Research and Quality’s single-level Clinical Classifications Software (Agency for Healthcare Research and Quality, Rockville, MD), is provided in table 2.

Two-year U.S. Social Security Administration’s Death Master File data on mortality were available for 6,717 of 6,994 patients; the remaining 4% were censored at discharge. Kaplan–Meier survival density function estimates for three median erythrocyte storage duration groups are provided in figure 2; no appreciable differences between the three groups was evident. However, as stated in the Methods section, these results were not adjusted for potential confounders and were not compared statistically.

In our Cox regression modeling, we found no significant evidence of a nonlinear relationship between patient-median erythrocyte storage duration and mortality after adjusting for potential confounders (fig 3, in red; P = 0.41, Wald chi-squared test). A simpler model that characterized the relationship linearly revealed similar conclusions: we found no evidence that increasing median storage duration was associated with a difference in the risk of postoperative mortality (P = 0.64), and based on this model, the hazard ratio (95% CI) corresponding to a relative 1-week increase in median erythrocyte storage duration was 0.99 (0.94–1.04).

No evidence against the assumption of proportionality of the erythrocyte storage duration effect over time was found in our modeling (diagnostic graphics and tests not reported).

To assess possible changes in blood product management over time, we analyzed the mean storage duration of erythrocyte units as a function of year of transfusion. Figure 4shows a relevant decrease in utilization of the oldest blood units, whereas young blood storage duration remains nearly unchanged.

Over the past 10 yr, investigators have considered whether prolonged storage duration of erythrocyte units may worsen patient outcomes in surgical and nonsurgical patients. Unfortunately, despite a large body of research, divergent results have so far not led to a conclusive answer, and no consensus has been reached. We thus used our Perioperative Health Documentation System to evaluate the relationship between erythrocyte storage duration and postoperative all-cause mortality in a cohort of 6,994 patients receiving 19,462 units of erythrocyte and undergoing general surgery. We did not find any association between erythrocyte storage duration and mortality and thus our study adds to the growing body of transfusion research, showing that erythrocyte age may not be associated with increased postoperative mortality or morbidity.

Only few studies address the general surgical population and those are restricted to colon–rectal, prostate, and liver transplant patients. For example, a study in 225 patients undergoing colon–rectal surgery showed that blood storage duration may, along with other risk factors, play a significant role in the development of postoperative infectious complications.23The same group demonstrated a decrease in the incidence of cancer recurrence in colon–rectal patients receiving younger blood.7On the other hand, our group found that storage duration did not affect the incidence of cancer recurrence in patients after prostate surgery.22In 405 patients receiving allogeneic transfusions, the biochemical recurrence–free survival rate at 5 yr was 74, 71, and 76% for those who received younger, middle, and older erythrocytes, respectively.22Recent data from 526 patients undergoing orthotopic liver transplantation do not show an association between erythrocyte age with infection, organ rejection, and death.29 

Over the past 10 yr, storage duration has received most attention in patients undergoing cardiac surgery. In one of the largest retrospective studies, also performed at the Cleveland Clinic, Koch et al.  11demonstrated an almost 60% percent increase in in-hospital mortality and significantly increased adverse events (cardiac, pulmonary, infectious, renal, and cerebrovascular) in patients receiving older blood. Andreasen et al.  10investigated 1,748 patients undergoing coronary bypass graft surgery and found a significant relationship between blood product storage time and postoperative severe infections. More recent literature in the cardiac patient population could not identify any benefit of younger blood.30,31 

Furthermore, in the largest study to date investigating storage duration in medical and surgical (cardiac and noncardiac) patients, Edgren et al.  20found no association between storage duration and mortality in more than 400,000 transfused patients for any of the studied patient populations.

One problem with currently available literature in regards to erythrocyte duration is that almost all the studies published so far are retrospective studies. It is likely that different statistical approaches contributed to the divergent results. However, it is interesting to note, that most of the earlier research showed adverse effects of increased storage duration, whereas the newer studies do not see this association.

In our study, like others previously, we adjusted for year of surgery, ABO-type, and the number of erythrocyte units transfused because these are true confounding factors (i.e. , factors that might affect the likelihood of receiving older or younger blood). Other baseline characteristics, such as comorbidities and type of surgery, probably only influence the decision to transfuse per se  rather than whether older or younger blood should be administered.

Our study, however, differs from previous ones, even from our own institution,11,32in several ways. One difference is that we calculated storage duration based on the age range of units a patient received and excluded patients with ranges greater than 5 days. This period was chosen to enforce a degree of measurement precision regarding the storage duration assigned to each patient. We thereby excluded 23% of otherwise eligible patients from the analysis but increased homogeneity of our study population. A grouping approach using predefined storage duration cutoffs would have eliminated a comparable number of patients.

A second difference is that we used storage duration as a continuous exposure rather than a dichotomous comparison for our primary analysis. This approach allowed us to address nonlinear time dependence. We found that postoperative mortality as a function of median storage duration for five specific snapshots in time (1, 3, and 6 months; 1 and 2 yr) does not change. It is thus not surprising that survival rates were similar when we divided our study population into three groups (patients receiving blood aged 3–14, 14.5–28, and 28.5–42 days) as a secondary survival analysis.

Since the first published reports on a possible association between storage duration and adverse outcomes, transfusion practice might have changed. We therefore analyzed storage duration of erythrocytes as a function of the year of transfusion. Figure 4demonstrates a clear decrease in storage duration for the oldest units, whereas ages of the youngest units seem unchanged over the study period from 2005 to 2009. Cleveland Clinic did not change policies for blood product selection and issue during this time period; therefore, the decrease in blood product age is most likely related to decreasing inventory levels combined with increasing or sustained blood product usage. Furthermore, studies showing adverse outcomes associated with transfusion might have led to decreased utilization of transfusions, thus leading blood banks to adjust the number of units ordered from blood suppliers and kept in store.

Because of the retrospective nature of our study, we were limited to investigate only the association between storage duration and mortality. It is important to note that there is a time lag between the date of death for an individual patient and when that fact is entered into the U.S. Social Security Administration’s Death Master File database. This delay at worst affected only a small portion of our sample and therefore has limited clinical relevance for our results. Currently, a number of prospective randomized trials investigating the relationship between storage duration and outcomes are underway, and results are anxiously awaited (ClinicalTrials.govIdentifier NCT00991341; NCT00458783).33 

Our results might be influenced by center-specific characteristics, because data were collected at one site only. On the other hand, we included a fairly large number of patients with a broad spread of comorbidities who had a variety of surgical procedures. As is common in clinical practice, most of our patients were given a fairly small number of transfusions. It is possible that many of our transfused patients did not have a true oxygen carrying or oxygen delivery deficiency. Because poor outcome associated with storage lesion is thought to originate from impaired erythrocyte function or its effect on perfusion/oxygenation, it might not matter in patients with adequate oxygen delivery. It is thus possible that our results would differ in even sicker patients receiving larger numbers of erythrocytes. A possible effect in these specific patients could be postulated from studies in trauma patients showing a benefit from transfusion of younger erythrocytes.34Unfortunately, our sample size was not large enough to perform a sensitivity analysis restricted to massive transfusions.

Another limitation of our study might be that we only evaluated mortality but not postoperative morbidity. Mortality is a crude outcome, and it is most certainly possible that more intermediate outcomes might be more sensitive and benefit from younger blood. Subcutaneous tissue oxygen tension and wound infection could function as intermediate and clinical outcomes to demonstrate a storage age–related difference in oxygen delivery to tissues. Natural killer cell function might be used to assess possible impairing effects of storage duration on the patient’s immune function. We chose a 2-yr mortality outcome because of emerging literature suggesting long-term effects of intraoperative interventions on postoperative outcomes. However, it is feasible that our 2-yr outcome is affected by other events occurring between transfusion and the mortality ascertainment.

In summary, in our large retrospective study in adult general surgical patients, we did not find an association between storage duration of erythrocytes and 2-yr mortality. This is in support of the majority of the most recent literature in surgical and medical patients. Our results represent yet another piece in a large puzzle and underline the importance of sufficiently powered randomized trials to finally resolve the erythrocyte storage duration debate.