A total of 1,011 units of pRBCs were administered during the study period. Of these, 594 (56.9%) units were given to 342 patients as single-unit transfusions. Complete data were available for 575 (96.8%) units, which served as the final sample size. Sample demographics are shown in . The vast majority of pRBCs transfusions were given for ICU anemia (91.4%), well into the recipient's ICU course (mean time from ICU admission to transfusion 13.7 days), and without additional blood products (95.1%). The mean Hctpre was 21.5 (range 8.2–32). The mean pRBCs storage time was 29.5 days, and 94.3% of pRBCs units were ≥14 days old. Donor gender was distributed equally among male (49.9%) and female (50.1%) donors.
Correlations between independent variables and ΔHct are listed in . Recipient female gender (r2 = 0.23, P < 0.01) and Hctpre (r2 = −0.27, P < 0.01) were correlated most strongly with ΔHct. Additional significant correlations [in order of decreasing magnitude (|r2|)] were as follows: increasing recipient admission weight was associated with a decreased ΔHct (r2 = −0.15, P < 0.001), transfusion for ICU anemia as opposed to hemorrhage was associated with a decreased ΔHct (r2 = 0.13, P = 0.003), increasing age was associated with an increased ΔHct (r2 = 0.10, P = 0.02), increasing 24 hour fluid balance was associated with a decreased ΔHct (r2 = 0.10, P = 0.02), and increasing Tpre was associated with increased ΔHct (r2 = 0.09, P < 0.04). Admission diagnosis, recipient admission BMI, time from ICU admission to transfusion, Tpost, the age of blood, donor gender, and transfusion of additional blood products with the pRBCs unit were not correlated significantly with ΔHct.
Univariate correlations with ΔHct.
lists sample characteristics stratified by recipient gender. As compared to female recipients, male recipients were significantly younger (P < 0.01), more likely to have an admission diagnosis of trauma (P < 0.01), more likely to be transfused for an indication of ICU anemia (P < 0.01), had a higher BMI (P < 0.01), and received older blood (P < 0.01).
Sample characteristics stratified by recipient gender.
By univariate analysis, ΔHct was significantly greater for female recipients, as compared to male recipients (3.81% versus 2.82%, resp., P
< 0.01). However, donor gender did not impact ΔHct, which was 3.02% following receipt of female blood versus 3.23% following receipt of male blood (P
= 0.21). In order to assess for any possible effect of fluid equilibration over time, these analyses were repeated using ΔHct2
instead of ΔHct. The mean ΔHct2
was 2.97% (range −12.3, 15.8) and the mean Tpost2
was 14.1 hours (range 1–92). ΔHct2
remained significantly higher for female as compared to male recipients (3.71% versus 2.63%, resp., P
< 0.01). Furthermore, donor gender was not associated with ΔHct2
(2.91% following receipt of female blood versus 3.03% following receipt of male blood, P
The relationship between both donor and recipient gender and ΔHct was explored further by stratifying the sample into the 4 donor-recipient gender pairs listed in : Male donor male recipient (MM), female donor male recipient (FM), male donor female recipient (MF), and female donor female recipient (FF). ΔHct for each of these pairs is listed in and depicted in . Although significant differences in ΔHct existed between these 4 groups (ANOVA DF = 3, F = 10.78, P < 0.01), analysis of differences between individual pairs revealed only those with disparate recipient genders to be significant. By contrast, ΔHct was not significantly different when comparing pairs with disparate donor genders and identical recipient genders. Differences in ΔHct as well as adjusted P-values for individual pair comparisons are listed in .
Change in hematocrit following transfusion stratified by donor-recipient gender pairs.
Figure 1 Change in Hematocrit as a function of donor-recipient gender pair. Donor-recipient pairs: FM, female donor male recipient; MM, male donor male recipient; FF, female donor female recipient; MF male donor female recipient. Pair 1 − Pair 2 = ΔHct (more ...)
Results from multivariable logistic regression analysis are shown in . The following variables were added to the model in addition to recipient gender based on their association with either recipient gender or ΔHct (or both) by univariate analysis: recipient age, admission diagnosis, indication for transfusion, Hctpre, recipient admission weight, Tpre, age of blood, and recipient 24 hour fluid balance. Of note, recipient BMI was not included in the model based on a high degree of colinearity with recipient weight (r2 = 0.90, P < 0.01). After controlling for these covariates, recipient gender remained significantly associated with ΔHct (t = 4.32, P < 0.01) (model DF-9, F-12.57, P < 0.01). In additional to recipient gender, patient age, Hctpre, recipient admission weight, and Tpre were also associated independently with ΔHct.
Multivariable linear regression.