[Skip to Content]
[Skip to Content Landing]
Table 1.  
Characteristics of Study Population
Characteristics of Study Population
Table 2.  
Hazard Ratios of Death in Association With Number of Red Blood Cell Transfusions From Donors of Different Age and Sex
Hazard Ratios of Death in Association With Number of Red Blood Cell Transfusions From Donors of Different Age and Sex
Table 3.  
Hazard Ratios of Death During 30 Days in Association With Number of Red Blood Cell Transfusions From Donors of Different Randomly Allocated Age and Sex
Hazard Ratios of Death During 30 Days in Association With Number of Red Blood Cell Transfusions From Donors of Different Randomly Allocated Age and Sex
1.
Koch  CG, Li  L, Sessler  DI,  et al.  Duration of red-cell storage and complications after cardiac surgery.  N Engl J Med. 2008;358(12):1229-1239.PubMedGoogle ScholarCrossref
2.
Sartipy  U, Holzmann  MJ, Hjalgrim  H, Edgren  G.  Red blood cell concentrate storage and survival after cardiac surgery.  JAMA. 2015;314(15):1641-1643.PubMedGoogle ScholarCrossref
3.
Patterson  JA, Irving  DO, Isbister  JP,  et al.  Age of blood and adverse outcomes in a maternity population.  Transfusion. 2015;55(11):2730-2737.PubMedGoogle ScholarCrossref
4.
Middelburg  RA, Borkent-Raven  BA, Janssen  MP,  et al.  Storage time of blood products and transfusion-related acute lung injury  [published correction appears in Transfusion. 2012;52(6):1386].  Transfusion. 2012;52(3):658-667.PubMedGoogle ScholarCrossref
5.
Spinella  PC, Carroll  CL, Staff  I,  et al.  Duration of red blood cell storage is associated with increased incidence of deep vein thrombosis and in hospital mortality in patients with traumatic injuries.  Crit Care. 2009;13(5):R151.PubMedGoogle ScholarCrossref
6.
Edgren  G, Kamper-Jørgensen  M, Eloranta  S,  et al.  Duration of red blood cell storage and survival of transfused patients (CME).  Transfusion. 2010;50(6):1185-1195.PubMedGoogle ScholarCrossref
7.
Leal-Noval  SR, Muñoz-Gómez  M, Arellano-Orden  V,  et al.  Impact of age of transfused blood on cerebral oxygenation in male patients with severe traumatic brain injury.  Crit Care Med. 2008;36(4):1290-1296.PubMedGoogle ScholarCrossref
8.
Marik  PE, Corwin  HL.  Efficacy of red blood cell transfusion in the critically ill: a systematic review of the literature.  Crit Care Med. 2008;36(9):2667-2674.PubMedGoogle ScholarCrossref
9.
Fergusson  DA, Hébert  P, Hogan  DL,  et al.  Effect of fresh red blood cell transfusions on clinical outcomes in premature, very low-birth-weight infants: the ARIPI randomized trial.  JAMA. 2012;308(14):1443-1451.PubMedGoogle ScholarCrossref
10.
Steiner  ME, Ness  PM, Assmann  SF,  et al.  Effects of red-cell storage duration on patients undergoing cardiac surgery.  N Engl J Med. 2015;372(15):1419-1429.PubMedGoogle ScholarCrossref
11.
Lacroix  J, Hébert  PC, Fergusson  DA,  et al; ABLE Investigators; Canadian Critical Care Trials Group.  Age of transfused blood in critically ill adults.  N Engl J Med. 2015;372(15):1410-1418.PubMedGoogle ScholarCrossref
12.
Dhabangi  A, Ainomugisha  B, Cserti-Gazdewich  C,  et al.  Effect of transfusion of red blood cells with longer vs shorter storage duration on elevated blood lactate levels in children with severe anemia: the TOTAL randomized clinical trial.  JAMA. 2015;314(23):2514-2523.PubMedGoogle ScholarCrossref
13.
Heddle  NM, Cook  RJ, Arnold  DM,  et al.  Effect of short-term vs long-term blood storage on mortality after transfusion.  N Engl J Med. 2016;375(20):1937-1945.PubMedGoogle ScholarCrossref
14.
Halmin  M, Rostgaard  K, Lee  BK,  et al.  Length of storage of red blood cells and patient survival after blood transfusion: a bi-national cohort study.  Ann Intern Med. 2017;166(4):248-256.PubMedGoogle ScholarCrossref
15.
Conboy  IM, Conboy  MJ, Wagers  AJ, Girma  ER, Weissman  IL, Rando  TA.  Rejuvenation of aged progenitor cells by exposure to a young systemic environment.  Nature. 2005;433(7027):760-764.PubMedGoogle ScholarCrossref
16.
Villeda  SA, Plambeck  KE, Middeldorp  J,  et al.  Young blood reverses age-related impairments in cognitive function and synaptic plasticity in mice.  Nat Med. 2014;20(6):659-663.PubMedGoogle ScholarCrossref
17.
Loffredo  FS, Steinhauser  ML, Jay  SM,  et al.  Growth differentiation factor 11 is a circulating factor that reverses age-related cardiac hypertrophy.  Cell. 2013;153(4):828-839.PubMedGoogle ScholarCrossref
18.
Baht  GS, Silkstone  D, Vi  L,  et al.  Exposure to a youthful circulation rejuvenates bone repair through modulation of β-catenin.  Nat Commun. 2015;6:7131.PubMedGoogle ScholarCrossref
19.
Katsimpardi  L, Litterman  NK, Schein  PA,  et al.  Vascular and neurogenic rejuvenation of the aging mouse brain by young systemic factors.  Science. 2014;344(6184):630-634.PubMedGoogle ScholarCrossref
20.
Sinha  M, Jang  YC, Oh  J,  et al.  Restoring systemic GDF11 levels reverses age-related dysfunction in mouse skeletal muscle.  Science. 2014;344(6184):649-652.PubMedGoogle ScholarCrossref
21.
Chassé  M, Tinmouth  A, English  SW,  et al.  Association of blood donor age and sex with recipient survival after red blood cell transfusion.  JAMA Intern Med. 2016;176(9):1307-1314.PubMedGoogle ScholarCrossref
22.
Vasan  SK, Chiesa  F, Rostgaard  K,  et al.  Lack of association between blood donor age and survival of transfused patients.  Blood. 2016;127(5):658-661.PubMedGoogle ScholarCrossref
23.
Edgren  G, Rostgaard  K, Vasan  SK,  et al.  The new Scandinavian Donations and Transfusions database (SCANDAT2): a blood safety resource with added versatility.  Transfusion. 2015;55(7):1600-1606.PubMedGoogle ScholarCrossref
24.
Ludvigsson  JF, Otterblad-Olausson  P, Pettersson  BU, Ekbom  A.  The Swedish personal identity number: possibilities and pitfalls in healthcare and medical research.  Eur J Epidemiol. 2009;24(11):659-667.PubMedGoogle ScholarCrossref
25.
Schmidt  M, Pedersen  L, Sørensen  HT.  The Danish civil registration system as a tool in epidemiology.  Eur J Epidemiol. 2014;29(8):541-549.PubMedGoogle ScholarCrossref
26.
Holzmann  MJ, Sartipy  U, Olsson  ML, Dickman  P, Edgren  G.  Sex-discordant blood transfusions and survival after cardiac surgery: a nationwide cohort study.  Circulation. 2016;134(21):1692-1694.PubMedGoogle ScholarCrossref
27.
Akaike  H.  A new look at the statistical model identification.  IEEE Trans Automat Contr. 1974;19(6):716-723. doi:10.1109/TAC.1974.1100705Google ScholarCrossref
28.
Andersen  PK, Gill  RD.  Cox’s regression model for counting processes: a large sample study.  Ann Stat. 1982;10(4):1100-1120. doi:10.1214/aos/1176345976Google ScholarCrossref
29.
Charlson  ME, Pompei  P, Ales  KL, MacKenzie  CR.  A new method of classifying prognostic comorbidity in longitudinal studies: development and validation.  J Chronic Dis. 1987;40(5):373-383.PubMedGoogle ScholarCrossref
30.
Sundararajan  V, Henderson  T, Perry  C, Muggivan  A, Quan  H, Ghali  WA.  New ICD-10 version of the Charlson Comorbidity Index predicted in-hospital mortality.  J Clin Epidemiol. 2004;57(12):1288-1294.PubMedGoogle ScholarCrossref
31.
Edgren  G, Rostgaard  K, Hjalgrim  H.  Methodological challenges in observational transfusion research: lessons learned from the Scandinavian Donations and Transfusions (SCANDAT) database.  ISBT Sci Ser. 2017;12(1):191-195. doi:10.1111/voxs.12342Google ScholarCrossref
32.
Rostgaard  K.  Methods for stratification of person-time and events—a prerequisite for Poisson regression and SIR estimation.  Epidemiol Perspect Innov. 2008;5:7.PubMedGoogle ScholarCrossref
33.
van de Watering  L; Biomedical Excellence for Safer Transfusion (BEST) Collaborative.  Pitfalls in the current published observational literature on the effects of red blood cell storage.  Transfusion. 2011;51(8):1847-1854.PubMedGoogle ScholarCrossref
Original Investigation
June 2017

Association of Donor Age and Sex With Survival of Patients Receiving Transfusions

Author Affiliations
  • 1Department of Medical Epidemiology and Biostatistics, Karolinska Institutet, Stockholm, Sweden
  • 2Hematology Center, Karolinska University Hospital, Stockholm, Sweden
  • 3Department of Clinical Immunology, the Blood Bank, Rigshospitalet, Copenhagen, Denmark
  • 4Department of Epidemiology Research, Statens Serum Institut, Copenhagen, Denmark
  • 5Department of Clinical Immunology, Aarhus University Hospital, Aarhus, Denmark
  • 6Section of Cardiothoracic Surgery, Karolinska University Hospital, Stockholm, Sweden
  • 7Department of Molecular Medicine and Surgery, Karolinska Institutet, Stockholm, Sweden
  • 8Department of Emergency Medicine, Karolinska University Hospital, Huddinge, Stockholm, Sweden
  • 9Department of Internal Medicine, Solna, Karolinska Institutet, Stockholm, Sweden
  • 10Department of Hematology, Copenhagen University Hospital, Copenhagen, Denmark
JAMA Intern Med. 2017;177(6):854-860. doi:10.1001/jamainternmed.2017.0890
Key Points

Question  Does the sex and age of blood donors affect survival of patients receiving transfusions?

Findings  In this binational cohort study, which included 968 264 patients who received transfusions, there was no association between age and/or sex of blood donors and survival of patients. Even among the patients who received multiple units of blood from very young or very old donors, absolute mortality differences compared with patients who received no such units of blood were consistently below 0.5%.

Meaning  The age and sex of blood donors do not influence patient survival and need not be considered in blood allocation.

Abstract

Importance  Following animal model data indicating the possible rejuvenating effects of blood from young donors, there have been at least 2 observational studies conducted with humans that have investigated whether donor age affects patient outcomes. Results, however, have been conflicting.

Objective  To study the association of donor age and sex with survival of patients receiving transfusions.

Design, Setting, and Participants  A retrospective cohort study based on the Scandinavian Donations and Transfusions database, with nationwide data, was conducted for all patients from Sweden and Denmark who received at least 1 red blood cell transfusion of autologous blood or blood from unknown donors between January 1, 2003, and December 31, 2012. Patients were followed up from the first transfusion until death, emigration, or end of follow-up. Data analysis was performed from September 15 to November 15, 2016.

Exposures  The number of transfusions from blood donors of different age and sex. Exposure was treated time dependently throughout follow-up.

Main Outcomes and Measures  Hazard ratios (HRs) for death and adjusted cumulative mortality differences, both estimated using Cox proportional hazards regression.

Results  Results of a crude analysis including 968 264 transfusion recipients (550 257 women and 418 007 men; median age at first transfusion, 73.0 years [interquartile range, 59.8-82.4 years]) showed a U-shaped association between age of the blood donor and recipient mortality, with a nadir in recipients for the most common donor age group (40-49 years) and significant and increasing HRs among recipients of blood from donors of successively more extreme age groups (<20 years: HR, 1.12; 95% CI, 1.10-1.14; ≥70 years: HR, 1.25; 95% CI, 1.08-1.44). Higher mortality was also noted among recipients of blood from female donors (HR, 1.07; 95% CI, 1.07-1.07). Adjustments for number of transfusions with a linear term attenuated the associations, but the increased mortality for recipients of blood from young, old, and female donors was not eliminated. Closer examination of the association between number of transfusions and mortality revealed a nonlinear pattern. After adjustments to accommodate nonlinearity, donor age and sex were no longer associated with patient mortality.

Conclusions and Relevance  Donor age and sex were not associated with patient survival and need not be considered in blood allocation. Any comparison between common and less common categories of transfusions will inevitably be confounded by the number of transfusions, which drives the probability of receiving the less common blood components. Previous positive findings regarding donor age and sex are most likely explained by residual confounding.

Introduction

The possible negative health effects linked to particular characteristics of transfused blood units have received much attention in recent years. Much of this research has focused on the health effects of stored blood products.1-8 Although large observational and randomized studies have not found any negative effect of stored blood,9-14 other characteristics of blood units are receiving increasing attention.

Most recently, following reports from animal model studies suggesting that transfusions from young donors may have rejuvenating effects in older recipients,15-20 there have been at least 2 observational studies investigating whether donor age influences the survival of patients receiving transfusions.21,22 The studies used dissimilar statistical methods and reported conflicting results. In the first study, which was based on the Scandinavian Donations and Transfusions (SCANDAT2) database,23 a matched cohort design was used, and no associations were found between donor age and mortality of patients who received transfusions.22 The second study, based on Canadian data, used a more complex, time-dependent survival model and conversely reported increased risks of death among recipients of blood from young and female donors.21

The difference between the findings of these 2 studies is striking. If the Canadian findings are true, they would have immediate implications for transfusion medicine practice. To understand and explain these differences and ultimately reconcile the contradictory results, we conducted a new retrospective cohort study based on the SCANDAT2 database in which we used methods similar to those used in the Canadian study by Chassé et al.21

Methods
Data Sources

All analyses were based on the SCANDAT2 database, which has been described in detail previously.23 In brief, the database contains all electronically available data from Sweden since 1968 and from Denmark since 1983 on blood donors, blood donations, blood components, and patients who received transfusions. Using national registration numbers that are available for all persons living in both countries,24,25 the SCANDAT2 database has been linked with nationwide population registers, as well as registers for hospital and outpatient care, cause of death, and cancer occurrence. The creation of the SCANDAT2 database and the conduct of this study were approved by the Danish Data Protection Agency and the ethics review board in Stockholm, Sweden. Informed consent was waived by the ethics review board in Stockholm, Sweden, and the Danish Data Protection Agency.

Study Design

Blood unit allocation typically follows a first in–first out principle: when 1 or more blood units are ordered for a patient, the blood bank selects the available compatible unit(s) that had been stored the longest. Consequently, within a given hospital or administrative region, all characteristics attributed to a blood unit—apart from blood group—should be unconnected with patient characteristics. Hence, characteristics such as donor age and sex should not be associated with any prognostic factor of the patients. However, even if the blood units of interest are effectively randomly allocated, a patient who receives many transfusions overall is more likely to receive 1 or more of these units. Because underlying disease severity is linked to both the number of transfusions and the risk of death, a comparison of the risk of death among patients who did receive particular blood units of interest and those who did not can still be confounded. As such, meticulous adjustments are warranted. In this case, adjustment for the total number of transfusions will break the association between disease severity and the exposure of interest and will thus remove the confounding. More important, because the association between the number of transfusions and the risk of death is not direct,26 this adjustment must be set up flexibly to allow nonlinear associations. In addition, because the distribution of donor age, as well as patient mortality, may differ geographically, analyses must also account for the hospital. A detailed description of the analytical background is found in the eAppendix and eFigures 1 and 2 in the Supplement.

We designed a retrospective cohort study similar to the one described by Chassé et al.21 The analyses were restricted to the period from January 1, 2003, to December 31, 2012. We included all patients who were recorded to have received at least 1 red blood cell transfusion in this period. We excluded patients who received autologous transfusions or units that could not be traced to an identified donor. Patients who received transfusions were followed up from the date of the first transfusion until death, emigration, or end of follow-up. To increase the sensitivity for short-term survival effects, we also performed analyses in which the patients were followed up for a maximum of 30 days.

Given its importance in mediating the confounding effect of disease severity (eAppendix in the Supplement), we first examined the association between total number of red blood cell transfusions and risk of death using 3 separate statistical models based on the 30-day follow-up model. We fitted 1 model that incorporated only a log-linear term for number of transfusions; another model in which this variable was categorized as 1 to 10 or 11 to 20, and so on; and a third model in which number of transfusions was fitted as a restricted cubic spline with 5 knots (eAppendix in the Supplement). For each of the 3 models, we then extracted the predicted hazard ratio (HR) for all possible values for the “number of transfusions” variable, with 15 transfusions arbitrarily set as the reference. As an objective measure of model fit to compare the 3 models, we used the Akaike information criterion.27

Statistical Analysis

Statistical analysis was performed from September 15 to November 15, 2016. All blood units were categorized by donor age (<20, 20-29, 30-39, 40-49, 50-59, 60-69, or ≥70 years) and sex (male or female). The main exposure of interest was the number of red blood cell transfusions from donors of different age and sex. The cumulative number of transfusions of each category was allowed to change in a time-dependent manner during follow-up. Because the SCANDAT2 database records only the date and not the exact time of transfusion, exposure was allowed to change only once per day. The relative risk of death, expressed as HRs, in relation to the number of blood units from donors of a particular age or sex was estimated using Cox proportional hazards regression models. The models were set up using the Andersen-Gill counting process model.28 More important, in all analyses, we only incorporated information that was known at the time to be represented in the model. Analyses were conducted separately for donor age and donor sex exposures.

We used 2 different analytical approaches. First, we fitted 1 set of models that were similar to those in the analyses by Chassé et al.21 Specifically, we added 1 variable for each of the exposures of interest to the same model. For the models investigating associations between donor and recipient mortality, we included 1 variable for each donor age category except for the reference category (40-49 years). These variables contained each recipient’s attained number of blood units of each donor age category. Similarly, in the models of the association between donor sex and recipient mortality, we included 1 variable for the number of units originating from female donors, with blood from male donors as the reference. Because we thought that adding several highly correlated variables in the same model might lead to distorted results, we also set up an alternative approach that was different from the one used by Chassé et al.21 We instead performed 1 statistical analysis per donor characteristic variable. For donor age, this meant 7 separate analyses (1 for each age group variable), and for donor sex, 2 separate analyses (1 for the number of units from male donors and 1 for the number of units from female donors). In each of these models, the reference category constituted recipients of all other types of units.

In both of the approaches outlined, all variables for number of units of each category were fitted as log-linear terms, but to examine possible dose-response effects in more detail, we also performed analyses in which the number of units from donors of a particular age or sex was categorized as 0, 1 to 2, or 3 or more units. The categorical analysis was performed only for the second approach. For the categorized analyses, we also computed covariate-adjusted cumulative mortality for each exposure group, and cumulative mortality differences at 30 days to provide a more clinically relevant outcome measure.

Both main sets of analyses were otherwise similar and were designed to be identical to the analyses in the Canadian study, wherever possible. Adjustment was thus performed for the cumulative number of transfusions, patient age, patient sex, and Charlson Comorbidity Index,29,30 as ascertained from the respective patient register (see eTable 1 in the Supplement for coding details). Because the association between the number of transfusions and the risk of death was distinctly nonlinear on the log scale (eFigure 3 in the Supplement), we fitted this variable as a restricted cubic spline with 5 knots to account for nonlinearity (eAppendix in the Supplement). To allow comparison with the results from the study by Chassé et al,21 in which this adjustment was performed using a log-linear term, we also fitted models adjusting for cumulative number of transfusions as a log-linear term. Because we included data from a large number of hospitals, which may differ both in donor demographics and patient mortality, we also adjusted for hospital by stratifying the Cox proportional hazards regression model for this variable.

To validate the overall statistical approach, we performed an analysis in which, instead of using the actual age and sex of the contributing donors, we assigned the donor age and donor sex of each blood unit randomly according to the donor age and sex distribution presented in the study by Chassé et al.21 This analysis was based on the premise that receipt of 1 or more red blood cell units with a certain randomly assigned characteristic should not be causally associated with patient outcomes.31 The analysis otherwise followed the same approach as the main analyses.

All data processing and statistical analyses were conducted using SAS statistical analysis software, version 9.4 (SAS Institute Inc). The data set for the time-dependent model was designed using the publicly available Stratify macro.32

Results

Of the 1 015 159 patients in the database who received transfusions between 2003 and 2012, a total of 981 971 received at least 1 red blood cell unit. After exclusion of 13 271 patients who received transfusions from donors with uncertain identity and 436 patients who received autologous transfusions, 968 264 patients remained for analysis. The characteristics of the study population are presented in Table 1. The median age of the patients at the first transfusion was 73.0 years (interquartile range, 59.8-82.4 years), 550 257 (56.8%) were women, and 569 119 (58.8%) were from Sweden.

eFigure 3 in the Supplement depicts the estimated association between the cumulative number of red blood cell transfusions and the HR of death when the number of transfusions was fitted as a log-linear term, a categorical term, and using a spline function. The log-linear assumption resulted in the poorest model fit (as indicated by the highest Akaike information criterion), and the model using a restricted cubic spline resulted in the best model fit.

Table 2 presents the results of models investigating the association between the numbers of transfused red blood cell units from donors with particular characteristics and the mortality of transfusion recipients. When follow-up was restricted to 30 days, there was a U-shaped association, with a nadir corresponding to recipients of the most common category of blood units (donor age, 40-49 years [reference]) and significant and increasing HRs among recipients of blood from donors of successively more extreme (and, at the same time, increasingly uncommon) age groups (<20 years: HR, 1.12; 95% CI, 1.10-1.14; ≥70 years: HR, 1.25; 95% CI, 1.08-1.44). The association was attenuated but still noticeable with statistically significantly elevated HRs among recipients of blood from the youngest and oldest donors when we adjusted for cumulative number of transfusions as a log-linear term (model 1; <20 years: HR, 1.02; 95% CI, 1.00-1.05; ≥70 years: HR, 1.16; 95% CI, 1.01-1.35). More important, when adjustment for number of transfusions was performed with a restricted cubic spline function, none of the considered factors were associated with risk of death in the recipient (model 2). The HR per transfusion from a donor younger than 20 years of age was 0.98 (95% CI, 0.96-1.00). We observed similar patterns when we considered donor sex, with an HR of 0.99 (95% CI, 0.99-1.00) per unit from a female donor in the fully adjusted model. When follow-up was unrestricted during the 10-year study period, the number of units from donors younger than 20 years of age was again associated with the risk of death in model 1 (HR, 1.04; 95% CI, 1.03-1.04) but not in model 2 (HR, 1.01; 95% CI, 1.00-1.01).

Results from the second analytical approach, in which we considered the different variables for donor age and sex in separate models, are presented in eTable 2 in the Supplement. As in Table 2, we saw statistically significant associations only in the unadjusted model and in model 1, in which we adjusted for number of red blood cell transfusions as a log-linear term. Here, associations were stronger than in Table 2, with HRs ranging from 1.05 (95% CI, 1.04-1.05) per unit from a donor who was 50 to 59 years of age to 1.32 (95% CI, 1.14-1.52) per unit from a donor who was 70 years of age or older. With more careful adjustment for number of transfusions using restrictive cubic splines, all associations disappeared. Results from analyses with unrestricted follow-up were largely similar (eTable 2 in the Supplement).

More important, when the exposure in eTable 2 in the Supplement was categorized according to number of red blood cell transfusions of each respective blood category, no clear dose-response effects emerged (eTable 3 in the Supplement). When we compared patients who received 3 or more units with a particular characteristic with patients who received no such units, adjusted 30-day cumulative mortality differences were consistently below 0.5%, with upper 95% confidence limits no higher than 1.0% (eTable 4 in the Supplement).

The sensitivity analyses considering the effects of blood units from donors with randomly assigned age and sex are presented in Table 3. Despite the fact that the analyses were conducted on data for which the exposure of interest was randomly assigned and for which there should thus be no genuine causal associations with the outcome, models that adjusted for the number of transfusions using a linear term produced statistically significant associations with the lowest HRs in the most common blood category and with increasing HRs with more extreme (and more uncommon) categories, just as in Table 2. And, as in Table 2, all associations were removed with full statistical adjustment.

Discussion

Following a recent publication from a Canadian research group reporting increased mortality among patients who received red blood cell transfusions from both young donors and female donors,21 we set out to replicate these findings in an independent patient cohort. Using the same analytical approach as in the Canadian study, we were able to replicate their findings with some variation in point estimates. However, after adjusting more carefully for the total number of red blood cell transfusions, neither donor age nor donor sex was associated with patient mortality. Assuming our approach is correct, these findings indicate that, with regard to recipient outcomes, neither of these donor characteristics need be considered when allocating red blood cell units for transfusion.

For most patients, the allocation of blood units follows very predictable rules. Typically, when a clinician orders a blood unit for a patient, the local blood bank will select the oldest available unit of a compatible blood type. Other blood unit characteristics, such as donor age and sex, can therefore be assumed to be randomly distributed among patients. When studying associations between the numbers of transfusions with a particular characteristic and the risk of death in the recipient, underlying disease severity in patients may still confound the association through total number of transfusions (eAppendix in the Supplement). However, with meticulous adjustment for total number of transfusions, it should be possible to block the confounding effect of patient disease severity entirely. This reason is why patients in a previous study of donor age and patient outcomes were matched on the number of transfusions.22

Because of the large size of the study cohort, with long-term follow-up and access to detailed data from population-based health data registers, we were able to characterize the nonlinear association between number of red blood cell transfusions33 and recipient risk of death. Based on this analysis, we created a statistical model that corrected for this factor very carefully. We then verified the adequacy of our statistical model using a simple simulation in which the age and sex of the contributing blood donors were assigned randomly. Again, this process revealed strong associations in the unadjusted model but no associations in the fully adjusted model.

Limitations

There are some differences between the study by Chassé et al21 and our study that should be mentioned. First, there may be higher-level differences in donor demographics, blood-bank logistics, and component manufacture processes that may at least theoretically have contributed to some of the differences in the results. Moreover, it is quite likely that there are differences between the patients who received transfusions in Scandinavia and the patients who received transfusions in Canada, including mean follow-up and patient mortality, which, in turn, may affect relative risk magnitudes. The key difference between the 2 studies is rather the difference in adjusting for the key confounding factor, number of transfusions, in which we show that insufficient attention to this variable will result in strong associations between number of blood components from young or very old donors and recipient risk of death, even when donor age and sex were assigned randomly in a simulation. Therefore, we believe that, rather than reflecting true biological effects, the Canadian results can be explained by residual confounding (ie, that the observations resulted from incomplete adjustment for the number of transfusions).

Conclusions

The age and sex of donors of red blood cell units are unlikely to influence the survival of patients receiving transfusions. In addition, we believe these data reinforce the importance of extreme caution in assessing epidemiologic analyses in this field given the tremendous clinical and logistical implications of false-positive findings.

Back to top
Article Information

Accepted for Publication: February 13, 2017.

Corresponding Author: Gustaf Edgren, MD, PhD, Department of Medical Epidemiology and Biostatistics, Karolinska Institutet, Box 281, SE-171 77 Stockholm, Sweden (gustaf.edgren@ki.se).

Published Online: April 24, 2017. doi:10.1001/jamainternmed.2017.0890

Author Contributions: Dr Edgren had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.

Study concept and design: Edgren, Ullum, Sartipy, Holzmann, Nyrén, Hjalgrim.

Acquisition, analysis, or interpretation of data: All authors.

Drafting of the manuscript: Edgren, Hjalgrim.

Critical revision of the manuscript for important intellectual content: All authors.

Statistical analysis: Edgren, Hjalgrim.

Obtained funding: Edgren, Hjalgrim.

Administrative, technical, or material support: Edgren, Erikstrup, Hjalgrim.

Study supervision: Edgren.

Conflict of Interest Disclosures: None reported.

Funding/Support: The assembly of the Scandinavian Donations and Transfusions (SCANDAT2) database and the conduct of this study was made possible through grants 2011-30405 and 2007-7469 from the Swedish Research Council, grant 20090710 from the Swedish Heart-Lung Foundation, support to Dr Edgren from the Swedish Society for Medical Research and the Strategic Research Program in Epidemiology at Karolinska Institutet, and grant 2009B026 from the Danish Council for Independent Research.

Role of the Funder/Sponsor: The funding sources had no role in the design of the study; collection, management, analysis, and interpretation of the data; preparation, review, or approval of the manuscript; and decision to submit the manuscript for publication.

Additional Contributions: We are grateful to Michel Chassé, MD, PhD, FRCPC, and colleagues for providing assistance in our attempts at replicating their findings. No compensation was provided for these contributors.

References
1.
Koch  CG, Li  L, Sessler  DI,  et al.  Duration of red-cell storage and complications after cardiac surgery.  N Engl J Med. 2008;358(12):1229-1239.PubMedGoogle ScholarCrossref
2.
Sartipy  U, Holzmann  MJ, Hjalgrim  H, Edgren  G.  Red blood cell concentrate storage and survival after cardiac surgery.  JAMA. 2015;314(15):1641-1643.PubMedGoogle ScholarCrossref
3.
Patterson  JA, Irving  DO, Isbister  JP,  et al.  Age of blood and adverse outcomes in a maternity population.  Transfusion. 2015;55(11):2730-2737.PubMedGoogle ScholarCrossref
4.
Middelburg  RA, Borkent-Raven  BA, Janssen  MP,  et al.  Storage time of blood products and transfusion-related acute lung injury  [published correction appears in Transfusion. 2012;52(6):1386].  Transfusion. 2012;52(3):658-667.PubMedGoogle ScholarCrossref
5.
Spinella  PC, Carroll  CL, Staff  I,  et al.  Duration of red blood cell storage is associated with increased incidence of deep vein thrombosis and in hospital mortality in patients with traumatic injuries.  Crit Care. 2009;13(5):R151.PubMedGoogle ScholarCrossref
6.
Edgren  G, Kamper-Jørgensen  M, Eloranta  S,  et al.  Duration of red blood cell storage and survival of transfused patients (CME).  Transfusion. 2010;50(6):1185-1195.PubMedGoogle ScholarCrossref
7.
Leal-Noval  SR, Muñoz-Gómez  M, Arellano-Orden  V,  et al.  Impact of age of transfused blood on cerebral oxygenation in male patients with severe traumatic brain injury.  Crit Care Med. 2008;36(4):1290-1296.PubMedGoogle ScholarCrossref
8.
Marik  PE, Corwin  HL.  Efficacy of red blood cell transfusion in the critically ill: a systematic review of the literature.  Crit Care Med. 2008;36(9):2667-2674.PubMedGoogle ScholarCrossref
9.
Fergusson  DA, Hébert  P, Hogan  DL,  et al.  Effect of fresh red blood cell transfusions on clinical outcomes in premature, very low-birth-weight infants: the ARIPI randomized trial.  JAMA. 2012;308(14):1443-1451.PubMedGoogle ScholarCrossref
10.
Steiner  ME, Ness  PM, Assmann  SF,  et al.  Effects of red-cell storage duration on patients undergoing cardiac surgery.  N Engl J Med. 2015;372(15):1419-1429.PubMedGoogle ScholarCrossref
11.
Lacroix  J, Hébert  PC, Fergusson  DA,  et al; ABLE Investigators; Canadian Critical Care Trials Group.  Age of transfused blood in critically ill adults.  N Engl J Med. 2015;372(15):1410-1418.PubMedGoogle ScholarCrossref
12.
Dhabangi  A, Ainomugisha  B, Cserti-Gazdewich  C,  et al.  Effect of transfusion of red blood cells with longer vs shorter storage duration on elevated blood lactate levels in children with severe anemia: the TOTAL randomized clinical trial.  JAMA. 2015;314(23):2514-2523.PubMedGoogle ScholarCrossref
13.
Heddle  NM, Cook  RJ, Arnold  DM,  et al.  Effect of short-term vs long-term blood storage on mortality after transfusion.  N Engl J Med. 2016;375(20):1937-1945.PubMedGoogle ScholarCrossref
14.
Halmin  M, Rostgaard  K, Lee  BK,  et al.  Length of storage of red blood cells and patient survival after blood transfusion: a bi-national cohort study.  Ann Intern Med. 2017;166(4):248-256.PubMedGoogle ScholarCrossref
15.
Conboy  IM, Conboy  MJ, Wagers  AJ, Girma  ER, Weissman  IL, Rando  TA.  Rejuvenation of aged progenitor cells by exposure to a young systemic environment.  Nature. 2005;433(7027):760-764.PubMedGoogle ScholarCrossref
16.
Villeda  SA, Plambeck  KE, Middeldorp  J,  et al.  Young blood reverses age-related impairments in cognitive function and synaptic plasticity in mice.  Nat Med. 2014;20(6):659-663.PubMedGoogle ScholarCrossref
17.
Loffredo  FS, Steinhauser  ML, Jay  SM,  et al.  Growth differentiation factor 11 is a circulating factor that reverses age-related cardiac hypertrophy.  Cell. 2013;153(4):828-839.PubMedGoogle ScholarCrossref
18.
Baht  GS, Silkstone  D, Vi  L,  et al.  Exposure to a youthful circulation rejuvenates bone repair through modulation of β-catenin.  Nat Commun. 2015;6:7131.PubMedGoogle ScholarCrossref
19.
Katsimpardi  L, Litterman  NK, Schein  PA,  et al.  Vascular and neurogenic rejuvenation of the aging mouse brain by young systemic factors.  Science. 2014;344(6184):630-634.PubMedGoogle ScholarCrossref
20.
Sinha  M, Jang  YC, Oh  J,  et al.  Restoring systemic GDF11 levels reverses age-related dysfunction in mouse skeletal muscle.  Science. 2014;344(6184):649-652.PubMedGoogle ScholarCrossref
21.
Chassé  M, Tinmouth  A, English  SW,  et al.  Association of blood donor age and sex with recipient survival after red blood cell transfusion.  JAMA Intern Med. 2016;176(9):1307-1314.PubMedGoogle ScholarCrossref
22.
Vasan  SK, Chiesa  F, Rostgaard  K,  et al.  Lack of association between blood donor age and survival of transfused patients.  Blood. 2016;127(5):658-661.PubMedGoogle ScholarCrossref
23.
Edgren  G, Rostgaard  K, Vasan  SK,  et al.  The new Scandinavian Donations and Transfusions database (SCANDAT2): a blood safety resource with added versatility.  Transfusion. 2015;55(7):1600-1606.PubMedGoogle ScholarCrossref
24.
Ludvigsson  JF, Otterblad-Olausson  P, Pettersson  BU, Ekbom  A.  The Swedish personal identity number: possibilities and pitfalls in healthcare and medical research.  Eur J Epidemiol. 2009;24(11):659-667.PubMedGoogle ScholarCrossref
25.
Schmidt  M, Pedersen  L, Sørensen  HT.  The Danish civil registration system as a tool in epidemiology.  Eur J Epidemiol. 2014;29(8):541-549.PubMedGoogle ScholarCrossref
26.
Holzmann  MJ, Sartipy  U, Olsson  ML, Dickman  P, Edgren  G.  Sex-discordant blood transfusions and survival after cardiac surgery: a nationwide cohort study.  Circulation. 2016;134(21):1692-1694.PubMedGoogle ScholarCrossref
27.
Akaike  H.  A new look at the statistical model identification.  IEEE Trans Automat Contr. 1974;19(6):716-723. doi:10.1109/TAC.1974.1100705Google ScholarCrossref
28.
Andersen  PK, Gill  RD.  Cox’s regression model for counting processes: a large sample study.  Ann Stat. 1982;10(4):1100-1120. doi:10.1214/aos/1176345976Google ScholarCrossref
29.
Charlson  ME, Pompei  P, Ales  KL, MacKenzie  CR.  A new method of classifying prognostic comorbidity in longitudinal studies: development and validation.  J Chronic Dis. 1987;40(5):373-383.PubMedGoogle ScholarCrossref
30.
Sundararajan  V, Henderson  T, Perry  C, Muggivan  A, Quan  H, Ghali  WA.  New ICD-10 version of the Charlson Comorbidity Index predicted in-hospital mortality.  J Clin Epidemiol. 2004;57(12):1288-1294.PubMedGoogle ScholarCrossref
31.
Edgren  G, Rostgaard  K, Hjalgrim  H.  Methodological challenges in observational transfusion research: lessons learned from the Scandinavian Donations and Transfusions (SCANDAT) database.  ISBT Sci Ser. 2017;12(1):191-195. doi:10.1111/voxs.12342Google ScholarCrossref
32.
Rostgaard  K.  Methods for stratification of person-time and events—a prerequisite for Poisson regression and SIR estimation.  Epidemiol Perspect Innov. 2008;5:7.PubMedGoogle ScholarCrossref
33.
van de Watering  L; Biomedical Excellence for Safer Transfusion (BEST) Collaborative.  Pitfalls in the current published observational literature on the effects of red blood cell storage.  Transfusion. 2011;51(8):1847-1854.PubMedGoogle ScholarCrossref
×