The ABO Histo-Blood Group and AKI in Critically Ill Patients with Trauma or Sepsis

Abstract

Background and objective

ABO blood types are determined by antigen modifications on glycoproteins and glycolipids and associated with altered plasma levels of inflammatory and endothelial injury markers implicated in AKI pathogenesis. We sought to determine the association of ABO blood types with AKI risk in critically ill patients with trauma or sepsis.

Design, setting, participants, & measurements

We conducted two prospective cohort studies at an urban, academic, level I trauma center and tertiary referral center; 497 patients with trauma admitted to the surgical intensive care unit between 2005 and 2010 with an injury severity score >15 and 759 patients with severe sepsis admitted to the medical intensive care unit between 2008 and 2013 were followed for 6 days for the development of incident AKI. AKI was defined by Acute Kidney Injury Network creatinine and dialysis criteria.

Results

Of 497 patients with trauma, 134 developed AKI (27%). In multivariable analysis, blood type A was associated with higher AKI risk relative to type O among patients of European descent (n=229; adjusted risk, 0.28 versus 0.14; risk difference, 0.14; 95% confidence interval, 0.03 to 0.24; P=0.02). Of 759 patients with sepsis, AKI developed in 326 (43%). Blood type A again conferred higher AKI risk relative to type O among patients of European descent (n=437; adjusted risk, 0.53 versus 0.40; risk difference, 0.14; 95% confidence interval, 0.04 to 0.23; P=0.01). Findings were similar when analysis was restricted to those patients who did not develop acute respiratory distress syndrome or were not transfused. We did not detect a significant association between blood type and AKI risk among individuals of African descent in either cohort.

Conclusions

Blood type A is independently associated with AKI risk in critically ill patients with trauma or severe sepsis of European descent, suggesting a role for ABO glycans in AKI susceptibility.

Introduction

AKI is common among the critically ill and associated with significant mortality and morbidity (1–3). The mechanisms underlying the development of AKI remain incompletely understood, leading to a dearth of specific therapies beyond supportive care and renal replacement. Although evidence from human studies and animal models suggests some differences in AKI pathogenesis in the critically ill population, certain mechanisms are likely shared across at-risk groups (4,5). Factors that may influence the development of AKI, including renal ischemia, host immune response, and endothelial activation, are common in severe sepsis, major trauma, and other critical illness syndromes (5–10).

Recent studies have implicated the ABO histo-blood group system as a potential mediator of endothelial dysfunction, microcirculatory coagulation, and inflammation linked to risk of acute myocardial infarction (MI), venous thromboembolism (VTE), and acute respiratory distress syndrome (ARDS) (11–16). The ABO histo-blood group system, originally described in red blood cells, consists of terminal carbohydrate modifications on glycoproteins and glycolipids (17). The ABO antigens are major components of the glycocalyx coating the endothelium and epithelium of several tissues (11,18). In the kidneys, ABO antigens are located on the renal vascular endothelium of arteries, veins, and peritubular and glomerular capillaries in addition to the epithelial cells of the convoluted tubules and collecting ducts (19). Although the precise biologic significance of the ABO antigens is not fully known, they are hypothesized to be important mediators of cell-cell interactions, including the leukocyte-endothelial cell interactions that are important in AKI pathogenesis (5–7,11). Genetic polymorphisms in the ABO gene have been associated with circulating levels of glycoproteins important in endothelial function and inflammation, including soluble intercellular adhesion molecule-1 (ICAM-1), thrombomodulin, vWF, and the selectins (12,20–23). These same proteins have been implicated in the pathogenesis of AKI (24–28). Although non–O blood types are reported risk factors for MI, VTE, and ARDS, it is unknown if ABO blood type is associated with AKI risk (13–16).

The objective of this study was to determine if specific ABO blood types represent a common risk factor for AKI across two separate populations of critically ill patients: major trauma and severe sepsis. Because blood type A has shown the strongest association with higher ARDS, MI, and VTE risk, we hypothesized that blood type A would be associated with a higher risk of AKI relative to type O. Although many aspects of AKI pathogenesis likely differ between patients with trauma and patients with sepsis, we also hypothesized that shared mechanisms of AKI development would result in similar associations of ABO blood type and AKI risk in both populations. Some of the results described here have been presented in abstract form (Reilly J, et al., unpublished data).

Materials and Methods

Study Design

We conducted two independent cohort studies of critically ill patients admitted to the Hospital of the University of Pennsylvania, an urban tertiary referral center and level I trauma center. The first cohort consisted of severely injured patients with trauma, and the second cohort consisted of patients with severe sepsis.

Trauma Population

Severely injured patients with trauma admitted through the emergency department (ED) to the surgical intensive care unit (ICU) were screened prospectively for enrollment between 2005 and 2010. Detailed inclusion and exclusion criteria have been previously described (29,30). Patients were included if they had an injury severity score >15 (31). Patients were excluded for isolated severe head injury, death, or discharge from the ICU within 24 hours of presentation, age <14 years old, or need for chronic dialysis before presentation. The Institutional Review Board of the University of Pennsylvania approved the study with a waiver of informed consent.

Severe Sepsis Population

Patients admitted to the medical ICU from the ED or transferred from hospital wards or other institutions were screened for enrollment in the Molecular Epidemiology of Severe Sepsis in the ICU (MESSI) Cohort Study between 2008 and 2013. Patients were enrolled if the primary reason for ICU admission was severe sepsis as defined by the American College of Chest Physicians (ACCP) consensus definition (32). Patients were excluded for a lack of commitment to life-sustaining measures at the time of screening, admission from a long–term acute care facility, a primary reason for ICU admission unrelated to sepsis, or need for chronic dialysis before presentation. The Institutional Review Board of the University of Pennsylvania approved the study with a waiver of timely informed consent. Consent was obtained from patients or their surrogates as soon as feasible, and patients could withdraw consent at any time.

Data Collection

Trained research personnel collected clinical data for both cohorts, including demographics and chronic health information, and for patients with trauma, mechanism, severity, and location of injury. Blood transfusions, physiologic and laboratory variables, and microbiologic data were tracked over the first 6 days in the ICU. ABO blood type was determined using standard methods by the hospital blood bank for clinical purposes and extracted from the medical record. Patients missing ABO blood type were excluded from final analyses.

Outcome Definition

The primary outcome was the development of AKI within 6 days of ICU admission as defined by the Acute Kidney Injury Network (AKIN) creatinine and RRT consensus criteria (33). We chose to follow patients for 6 days to capture AKI reflective of the pathogenic mechanisms involved in acute trauma or severe sepsis and limit the confounding effects of later ICU interventions and complications. AKIN criteria define AKI as an increase in serum creatinine ≥0.3 mg/dl or ≥50% over a 48-hour period or the need for acute RRT. We determined whether such an increase occurred over any 2-day interval through the first 6 days of ICU admission starting at ED presentation or if transferred from a hospital ward, 8 hours before ICU admission. Staging was determined for each patient who met criteria for AKI using creatinine and RRT data collected for 7 additional days after the diagnosis of AKI. We collected creatinine values within the 6 months before enrollment when available. Because we would have had to estimate baseline creatinine for many patients, disproportionately so in the trauma cohort, we did not use prehospitalization values as part of the AKI definition (34). We did not include AKIN urine output criteria, because hourly urine output data were not available, and patients did not universally have urinary catheters during the study period.

Statistical Analyses

Analyses for the trauma and MESSI cohorts were conducted separately but with identical approaches. Clinical characteristics were compared between those patients with and without AKI as well as across ABO blood types using the Pearson chi–squared, Fisher exact, Wilcoxon rank sum, or Kruskal–Wallis test as appropriate. We used multivariable logistic regression to adjust the association of ABO blood type with AKI for potential confounders. Our primary analysis compared the two most common ABO blood types, A and O, and excluded patients with the less common blood types. Secondary analyses were conducted comparing each of four main blood types with each other. We considered all baseline variables listed in Table 1 that had an unadjusted association with AKI or ABO blood type at P<0.20 for inclusion in the multivariable models. Using backward elimination, we then removed variables that did not alter the odds ratio for the association of ABO blood type and AKI by ≥10% to reach our final models (35). Certain variables, including severity of illness, age, sex, and red blood cell transfusion, were included in all final models. To adjust for severity of illness, we used Acute Physiology, Age, Chronic Health Evaluation III (APACHE III) scores without renal components to avoid colinearity with the outcome of AKI. All analyses were a priori stratified by race given the substantial differences in blood type frequencies and genetic background between racial groups. Postestimation marginal analysis was used for each final logistic regression model to compute standardized, adjusted AKI risks as well as convert odds ratios to risk differences (RDs) by ABO blood type, including 95% confidence intervals (95% CIs) (36). On the basis of our fixed sample sizes and the relative frequencies of blood types A and O in each race, we estimated that we would have 80% power to detect 15% and 18% differences in AKI risk in patients with trauma of European and African descent, respectively, and 13% and 20% AKI RDs in patients with sepsis of European and African descent, respectively.

Table 1.

Patient characteristics by presence or absence of AKI

Characteristics	Trauma Cohort			Severe Sepsis Cohort
	AKI (n=134; 27%)	No AKI (n=363; 73%)	P Valuei	AKI (n=335; 45%)	No AKI (n=417; 55%)	P Valuei
Demographics
Race			<0.01			0.68
African descent	85 (63%)	168 (46%)		106 (32%)	136 (33%)
European descent	48 (36%)	181 (50%)		201 (60%)	236 (57%)
Asian descent	1 (1%)	10 (3%)		7 (2%)	12 (3%)
Other	0 (0%)	4 (1%)		14 (4%)	18 (4%)
Unknown	0 (0%)	0 (0%)		7 (2%)	15 (4%)
Age (yr)	35 (23–69)	36 (24–50)	0.70	60 (52–68)	63 (52–68)	0.04
Men	102 (76%)	268 (74%)	0.60	207 (62%)	221 (53%)	0.02
Body mass index (kg/m2)	27 (24–31)	26 (22–29)	<0.001	27 (23–32)	26 (22–31)	<0.01
Medical history
Hypertension	24 (18%)	56 (15%)	0.50	161 (48%)	207 (51%)	0.52
Diabetes mellitus	13 (10%)	13 (4%)	<0.01	97 (29%)	119 (29%)	0.92
CKD	0 (0%)	2 (1%)	>0.99	41 (13%)	39 (10%)	0.22
Congestive heart failurea	0 (0%)	0 (0%)	>0.99	56 (17%)	47 (11%)	0.03
Malignancyb	0 (0%)	0 (0%)	>0.99	116 (35%)	139 (34%)	0.75
Smoking historyc	51 (40%)	152 (43%)	0.56	175 (60%)	224 (60%)	0.93
Initial creatinine (mg/dl)	1.2 (0.9–1.4)	1.1 (0.9–1.3)	<0.01	1.6 (1.0–2.5)	1.3 (0.9–2.0)	<0.001
Severity and mechanism
APACHE III score	46 (37–61)	34 (26–41)	<0.001	82 (67–98)	70 (56–85)	<0.001
APACHE III without renald	42 (30–51)	31 (25–37)	<0.001	74 (61–89)	64 (49–77)	<0.001
Injury severity scoree	25 (19–29)	22 (19–29)	0.17	N/A	N/A
Mechanism of traumae			<0.001	N/A	N/A
Blunt	76 (57%)	272 (75%)
Penetrating	58 (43%)	91 (25%)
Surgery performede	96 (72%)	180 (50%)	<0.001	N/A	N/A
Blood transfusions (%)f
Red blood cells	104 (78%)	197 (55%)	<0.001	32 (10%)	15 (4%)	<0.01
Platelets	51 (38%)	68 (19%)	<0.001	34 (10%)	23 (6%)	0.02
Fresh frozen plasma	88 (66%)	130 (36%)	<0.001	26 (8%)	8 (2%)	<0.01
Nephrotoxinsg
Intravenous contrast	74 (55%)	281 (77%)	<0.01	61 (18%)	108 (26%)	0.01
Amioglycosides	20 (15%)	39 (11%)	0.21	151 (45%)	110 (26%)	<0.001
NSAIDs	4 (3%)	8 (2%)	0.62	17 (5%)	23 (6%)	0.79
Calcineurin inhibitor	0 (0%)	0 (0%)	>0.99	48 (14%)	43 (10%)	0.09
Source of infection						0.07
Pulmonary	N/A	N/A		148 (44%)	172 (41%)
Genitourinary				29 (9%)	57 (14%)
Abdominal/gastrointestinal				50 (15%)	52 (12%)
Head/neck				9 (3%)	6 (1%)
Blood				25 (7%)	38 (9%)
Skin/soft tissue/bone				12 (4%)	27 (6%)
Gynecologic				0 (0%)	2 (1%)
Unknown				62 (19%)	63 (15%)
ABO blood type			0.01			0.08
A	49 (37%)	116 (32%)		143 (43%)	140 (34%)
B	30 (22%)	49 (14%)		44 (13%)	61 (15%)
AB	3 (2%)	24 (7%)		14 (4%)	22 (5%)
O	52 (39%)	174 (48%)		134 (40%)	194 (47%)
Outcome
30-d Mortality	27 (20%)	15 (4%)	<0.001	198 (59%)	131 (31%)	<0.001
ICU length of stay (d)h	20 (9–38)	7 (4–17)	<0.001	∞ (7–∞)	6 (2–∞)	<0.001

Data are shown as n (%) for categorical variables and median (interquartile range) for continuous variables. APACHE III, Acute Physiology, Age, Chronic Health Evaluation III; NSAID, nonsteroidal anti–inflammatory drug; ICU, intensive care unit; N/A, not applicable.

^aPatients with congestive heart failure were excluded from enrollment in the trauma cohort.

^bMetastatic solid malignancy or hematologic malignancy.

^cSmoking history includes current and former cigarette smokers.

^dAPACHE III without renal refers to the APACHE III score with the renal components removed.

^eInjury severity score, mechanism of trauma, and surgery performed are only relevant to the trauma cohort. Source of infection is only relevant to the severe sepsis cohort. Surgery performed refers to patients undergoing emergent surgical intervention in an operating room between emergency department presentation and ICU admission.

^fBlood transfusion refers to the percentage of patients transfused during the first 24 hours of resuscitation (trauma cohort) or the first 24 hours of ICU admission (severe sepsis cohort).

^gNephrotoxins refer to the exposure of intravenous contrast or aminoglycosides during the first 6 days of admission and a medical history of NSAID or calcineurin inhibitor use before admission. In our cohort, all patients with a medical history of calcineurin inhibitors had undergone a solid organ or bone marrow transplant.

^hIntensive care unit length of stay was assumed to be infinite in patients who died. Given that >50% of patients with AKI died in the severe sepsis cohort, the median ICU length of stay is infinity.

ⁱContinuous variables were compared using a Wilcoxon rank sum test, and categorical variables were compared using a Pearson chi–squared or Fisher’s exact test.

In secondary analyses, we determined the association of ABO blood type and AKI stage. Given the limited number of patients with stage 3 AKI, those with stages 2 and 3 AKI were combined. Patients with stage 2/3 AKI and those with stage 1 AKI were separately compared with patients without AKI using multivariable logistic regression. To determine independence from our previously reported associations between ABO blood types and ARDS (14), we also conducted analyses stratified by the development of ARDS. ARDS was phenotyped using the Berlin definition and our previously published methods over the first 6 days of ICU admission (37,38). We also conducted analyses within the subgroup of patients who did not receive blood product transfusions during the first 24 hours of enrollment. In the severe sepsis cohort, acute renal dysfunction at ICU admission was one measure of organ dysfunction that could qualify a patient’s sepsis as severe (32). To ensure that using renal dysfunction at enrollment as an inclusion criterion did not bias our results, we conducted additional analyses in this cohort restricted to patients who did not have acute renal dysfunction at ICU admission. All statistical analyses were performed using Stata/IC 12.0 (StataCorp LP, College Station, TX).

Results

Trauma Population

We enrolled 497 patients with trauma (Figure 1) with a median age of 35 years old and injury severity score of 22. The majority of patients were of African (51%) or European descent (45%). AKI developed in 134 (27%) patients. Comparisons of patient characteristics by AKI status are provided in Table 1. As we have reported previously, risk factors for AKI included African ancestry, increased body mass index, and transfusion of blood products (29). ABO blood type was available in all 497 (100%) patients with trauma. Frequencies of the ABO blood types by race were similar to reported distributions in the United States population (39). Distributions and unadjusted associations between clinical variables and ABO blood type are provided in Table 2.

Figure 1.

Screening and enrollment. (A) Trauma cohort. (B) Sepsis cohort. ^aSome patients met more than one exclusion criterion. BMI, body mass index; BSA, body surface area; CHF, congestive heart failure; ICU, intensive care unit; ISS, injury severity score; LTACH, long–term acute care hospital; MI, myocardial infarction.

Table 2.

Characteristics of patients with trauma by ABO blood type

Characteristics	ABO Blood Type				P Valueg
	A (n=165; 33%)	B (n=79; 16%)	AB (n=27; 6%)	O (n=226; 45%)
Demographics
Race					<0.001
African descent	64 (39%)	50 (63%)	12 (44%)	127 (56%)
European descent	94 (57%)	28 (35%)	12 (44%)	95 (42%)
Asian descent	6 (3%)	1 (1%)	2 (7%)	2 (1%)
Other	1 (1%)	0 (0%)	1 (4%)	2 (1%)
Age (yr)	39 (26–52)	39 (22–54)	47 (26–59)	32 (23–46)	0.08
Men	119 (72%)	54 (68%)	21 (78%)	176 (78%)	0.31
Body mass index (kg/m2)	26 (23–30)	27 (24–31)	27 (24–31)	26 (23–30)	0.90
Medical history
Hypertension	27 (16%)	18 (23%)	9 (33%)	26 (12%)	<0.01
Diabetes mellitus	8 (5%)	6 (8%)	3 (11%)	9 (4%)	0.25
CKD	1 (1%)	1 (1%)	0 (0%)	0 (0%)	0.24
Smoking historya	55 (34%)	29 (39%)	10 (37%)	109 (49%)	0.02
Initial creatinine (mg/dl)	1.1 (0.9–1.2)	1.2 (0.9–1.4)	1.1 (0.8–1.4)	1.1 (0.9–1.3)	0.10
Severity and mechanism
APACHE III score	36 (28–46)	39 (31–49)	36 (25–41)	36 (29–45)	0.39
APACHE III without renalb	31 (26–42)	36 (26–43)	33 (25–37)	32 (26–41)	0.53
Injury severity score	22 (19–29)	22 (19–29)	22 (20–29)	22 (18–29)	0.89
Mechanism of trauma					0.17
Blunt	125 (76%)	50 (63%)	20 (74%)	153 (68%)
Penetrating	40 (24%)	29 (37%)	7 (26%)	73 (32%)
Surgery performedc	87 (53%)	45 (57%)	13 (48%)	131 (58%)	0.63
Blood transfusions (%)d
Red blood cells	99 (60%)	53 (67%)	12 (44%)	137 (61%)	0.22
Platelets	41 (25%)	24 (30%)	4 (15%)	50 (22%)	0.32
Fresh frozen plasma	73 (44%)	39 (49%)	4 (15%)	50 (22%)	0.71
Nephrotoxinse
Intravenous contrast	119 (72%)	54 (68%)	18 (67%)	164 (73%)	0.84
Aminoglycosides	19 (12%)	10 (13%)	1 (4%)	29 (13%)	0.60
NSAIDs	4 (2%)	1 (1%)	1 (4%)	6 (3%)	0.88
Outcomes
30-d Mortality	21 (13%)	5 (6%)	1 (4%)	15 (7%)	0.11
ICU length of stay (d)f	10 (5–23)	9 (4–26)	10 (4–19)	10 (4–19)	0.46
AKI (any stage)	49 (30%)	30 (38%)	3 (11%)	52 (23%)	0.01
AKI (stage 2 or 3)	13 (8%)	4 (5%)	2 (7%)	16 (7%)	0.88
Need for dialysis	11 (7%)	1 (1%)	1 (4%)	9 (4%)	0.26

Data are shown as n (%) for categorical variables and median (interquartile range) for continuous variables. APACHE III, Acute Physiology, Age, Chronic Health Evaluation III; NSAID, nonsteroidal anti–inflammatory drug; ICU, intensive care unit.

^aSmoking history includes current and former cigarette smokers.

^bAPACHE III without renal refers to the APACHE III score with the renal components removed.

^cSurgery performed refers to patient undergoing emergent surgical intervention in an operating room between emergency department presentation and ICU admission.

^dBlood transfusions are presented as the number (%) of patients who received a transfusion of the specified blood product in the first 24 hours after emergency department presentation.

^eNephrotoxins refer to the exposure of intravenous contrast or aminoglycosides during the first 6 days of admission and a medical history of NSAID use just before admission.

^fICU length of stay was assumed to be infinite in patients who died.

^gContinuous variables were compared using a Kruskal–Wallis test, and categorical variables were compared using a Pearson chi–squared or Fisher’s exact test.

In multivariable analysis, blood type A was associated with higher AKI risk relative to blood type O among patients of European descent (adjusted RD, 0.14; 95% CI, 0.03 to 0.24; P=0.02) but not among patients of African descent (adjusted RD, 0.03; 95% CI, −0.09 to 0.15; P=0.61) (Figure 2A, Table 3). Comparisons of blood types B and AB with types O and A were limited by the smaller sample size of the B and AB types and did not result in any statistically significant associations (Supplemental Table 1). The association of blood type A relative to O with AKI was not substantially changed by excluding from the analysis patients who developed ARDS or were transfused during resuscitation (Table 3). Among patients of European descent, the RD between those who developed stage 2/3 AKI and those without AKI was similar to the RD observed between stage 1 AKI and no AKI (Supplemental Table 2), with no incremental increase in stage 2/3 AKI risk over stage 1 AKI risk. Initial creatinine at presentation and creatinine at discharge or death did not differ by ABO blood type in either racial group (Supplemental Table 3). A creatinine within the 6 months before trauma was only available in 86 (17%) patients and also, did not differ by ABO blood type (Supplemental Table 3); 22 (4%) patients required dialysis during their hospital stay (8 patients of European descent and 14 patients of African descent). Of eight patients of European descent, all eight were blood type A.

Figure 2.

Adjusted risk of AKI by ABO blood type among patients of European descent. (A) Patients with trauma. (B) Patients with severe sepsis. Points represent the adjusted AKI risk for each ABO blood type. Vertical error bars represent 95% confidence intervals. P values are for the comparison of blood type A with blood type O in adjusted logistic regression models. Tables 3 and 5 list the covariates included in the final multivariable models.

Table 3.

Associations of ABO blood types and AKI in the trauma cohort

Population	No.	Adjusted AKI Risk (%)c				Adjusted RDd (95% Confidence Interval) for A Versus O	P Value
		A	B	AB	O
Overall trauma cohort
European descent	229	28	24	13	14	0.14 (0.03 to 0.24)	0.02
African descent	253	35	39	17	32	0.03 (−0.09 to 0.15)	0.61
Not transfuseda
European descent	96	19	10	0	6	0.13 (0.00 to 0.27)	0.05
African descent	71	17	18	0	14	0.03 (−0.18 to 0.25)	0.74
Not ARDSb
European descent	156	26	18	14	10	0.16 (0.03 to 0.29)	0.02
African descent	190	32	29	9	23	0.09 (−0.05 to 0.22)	0.21

ARDS, acute respiratory distress syndrome; RD, risk difference.

^aAnalyses limited to patients who were not transfused any blood products during the first 24 hours after emergency department presentation.

^bAnalyses limited to patients who did not develop ARDS during the first 6 days of hospital admission.

^cAKI risk by ABO blood type on the basis of multivariable logistic regression models adjusted for age; Acute Physiology, Age, Chronic Health Evaluation III score without the renal components; body mass index; mechanism of injury; injury severity score; history of diabetes; and units of red blood cells transfused in the first 24 hours.

^dRDs are the adjusted differences in AKI risk between blood types A and O. All other comparisons between blood types A, B, AB, and O were not statistically significant and are displayed in Supplemental Table 1.

Severe Sepsis Population

We enrolled 759 critically ill patients with severe sepsis (Figure 1), with a median age of 61 years old and a median APACHE III score of 75. The source of sepsis was pulmonary in 43%. The majority of patients were of European (58%) or African descent (32%). Supplemental Table 4 provides the frequency of each organ dysfunction defining severe sepsis at ICU admission. Acute renal dysfunction according to ACCP sepsis criteria was present in 279 (37%) of patients at ICU admission. AKIN-defined AKI developed in 337 of 759 (44%) patients. Characteristics of those patients who did and did not develop AKI are provided in Table 1. ABO blood type was available in 752 of 759 (99%) enrolled patients. Seven patients without ABO blood type available, two of whom developed AKI, were excluded from additional analyses. Frequencies of the ABO blood types by race were similar to those in the trauma cohort and the reported distributions in the United States (39). Distributions and unadjusted associations between clinical variables and ABO blood type are shown in Table 4.

Table 4.

Characteristics of patients with severe sepsis by ABO blood type

Characteristics	ABO Blood Typeg				P Valueh
	A (n=283; 38%)	B (n=105; 14%)	AB (n=36; 5%)	O (n=328; 44%)
Demographics
Race					<0.001
African descent	76 (27%)	50 (48%)	13 (36%)	103 (31%)
European descent	188 (66%)	40 (38%)	19 (53%)	190 (58%)
Asian descent	3 (1%)	5 (5%)	2 (6%)	9 (3%)
Other	7 (3%)	6 (6%)	1 (3%)	18 (5%)
Unknown	9 (3%)	4 (4%)	1 (3%)	8 (2%)
Age (yr)	62 (52–70)	61 (52–72)	62 (58–70)	61 (51–70)	0.57
Men	170 (60%)	53 (51%)	22 (61%)	183 (56%)	0.40
Body mass index (kg/m2)	26 (22–30)	27 (23–32)	26 (24–30)	26 (22–31)	0.11
Medical history
Hypertension	134 (47%)	59 (56%)	23 (64%)	152 (47%)	0.07
Diabetes mellitus	77 (27%)	35 (33%)	12 (33%)	92 (28%)	0.48
Congestive heart failure	46 (16%)	16 (15%)	3 (8%)	38 (12%)	0.26
CKD	32 (11%)	12 (11%)	4 (11%)	32 (10%)	0.91
Malignancya	102 (36%)	29 (28%)	16 (44%)	108 (33%)	0.25
Smoking historyb	145 (51%)	55 (52%)	19 (53%)	180 (55%)	0.84
Initial creatinine (mg/dl)	1.3 (0.9–2.2)	1.4 (1.0–2.1)	1.2 (0.9–2.7)	1.4 (0.9–2.3)	0.99
Severity of illness
APACHE III score	76 (59–93)	73 (58–90)	77 (64–96)	75 (60–91)	0.70
APACHE III without renalc	69 (54–84)	66 (53–82)	69 (59–90)	69 (56–83)	0.65
Source of infection
Pulmonary	115 (41%)	49 (47%)	18 (50%)	138 (42%)	0.77
Genitourinary	27 (10%)	14 (13%)	4 (11%)	41 (13%)
Abdominal/gastrointestinal	37 (13%)	9 (9%)	4 (11%)	52 (16%)
Head/neck	6 (2%)	4 (4%)	0 (0%)	5 (2%)
Blood	25 (9%)	10 (10%)	2 (6%)	26 (8%)
Skin/soft tissue/bone	16 (6%)	6 (6%)	2 (6%)	15 (5%)
Gynecologic	2 (1%)	0 (0%)	0 (0%)	0 (0%)
Unknown	55 (19%)	13 (12%)	6 (17%)	51 (16%)
Blood transfusions (%)d
Red blood cells	23 (8%)	6 (6%)	1 (3%)	17 (5%)	0.37
Platelets	24 (8%)	9 (9%)	3 (8%)	21 (6%)	0.76
Fresh frozen plasma	20 (7%)	4 (4%)	1 (3%)	9 (3%)	0.07
Nephrotoxinse
Intravenous contrast	61 (22%)	20 (19%)	6 (17%)	82 (25%)	0.43
Aminoglycosides	113 (40%)	29 (28%)	11 (31%)	108 (33%)	0.09
NSAIDs	16 (6%)	6 (6%)	1 (3%)	17 (5%)	0.90
Calcineurin inhibitor	39 (14%)	8 (8%)	3 (8%)	41 (13%)	0.35
Outcomes
30-d Mortality	135 (48%)	35 (33%)	19 (53%)	140 (43%)	0.05
ICU length of stay (d)f	28 (4–∞)	6 (3–∞)	∞ (3–∞)	12 (3–∞)	0.05
AKI (any stage)	143 (51%)	44 (42%)	14 (40%)	134 (41%)	0.08
AKI (stage 2 or 3)	73 (26%)	17 (16%)	8 (22%)	74 (23%)	0.26
Need for dialysis	22 (8%)	4 (4%)	3 (8%)	22 (7%)	0.53

Data are shown as n (%) for categorical variables and median (interquartile range) for continuous variables. APACHE III, Acute Physiology, Age, Chronic Health Evaluation III; NSAID, nonsteroidal anti–inflammatory drug; ICU, intensive care unit.

^aMetastatic solid malignancy or hematologic malignancy.

^bSmoking history includes current and former cigarette smokers.

^cAPACHE III without renal refers to the APACHE III score with the renal components removed.

^dBlood transfusions are presented as the number (%) of patients who received a transfusion of a specified blood product in the first 24 hours of ICU admission.

^eNephrotoxins refer to the exposure of intravenous contrast or aminoglycosides during the first 6 days of admission and a medical history of NSAID use just before admission.

^fICU length of stay was assumed to be infinite in patients who died. Given that >50% of patients with blood type AB died, the median is infinity.

^gValues do not add up to 100% because of rounding.

^hContinuous variables were compared using a Wilcoxon rank sum test, and categorical variables were compared using a Pearson chi–squared or Fisher’s exact test.

Similar to the trauma cohort, blood type A again conferred a higher AKI risk relative to type O among patients of European descent (adjusted RD, 0.14; 95% CI, 0.04 to 0.23; P=0.01) but not among patients of African descent (adjusted RD, 0.04; 95% CI, −0.10 to 0.18; P=0.55) (Figure 2B, Table 5). Comparisons of blood types B and AB with types O and A, again limited by the low frequency of the B and AB types, did not show any statistically significant differences in AKI risk (Supplemental Table 1). Compared with the overall sepsis cohort, the point estimates for the RD between blood types A and O were similar in patients who did not develop ARDS, were not transfused, and did not have acute renal dysfunction at ICU admission (Table 5). The AKI RD between blood types A and O in patients of European descent were similar comparing AKI stage 2/3 or 1 with no AKI (Supplemental Table 2), with no incremental increase in risk for stage 2/3 AKI over stage 1 AKI. Initial creatinine at enrollment and creatinine at death or discharge did not differ by ABO blood type (Supplemental Table 3). A lowest creatinine from the 6 months before ICU enrollment was available in 80% of patients, and it also did not differ by ABO blood type (Supplemental Table 3). We did not detect differences in the percentage requiring dialysis in patients of European descent (9% versus 5% for types A and O, respectively) and patients of African descent (7% versus 9% for types A and O, respectively).

Table 5.

Associations of ABO blood types and AKI in the severe sepsis cohort

Population	No.	Adjusted AKI Risk (%)d				Adjusted RDe (95% Confidence Interval) for A Versus O	P Value
		A	B	AB	O
Overall sepsis cohort
European descent	437	53	42	45	40	0.14 (0.04 to 0.23)	<0.01
African descent	242	47	43	39	43	0.04 (−0.10 to 0.18)	0.55
Not transfuseda
European descent	374	50	40	48	37	0.13 (0.02 to 0.23)	0.02
African descent	213	41	37	32	40	0.01 (−0.14 to 0.16)	0.89
Not ARDSb
European descent	245	38	31	36	29	0.09 (−0.03 to 0.22)	0.14
African descent	153	29	38	25	35	−0.06 (−0.23 to 0.11)	0.48
No renal dysfunction at enrolmentc
European descent	277	53	34	27	33	0.20 (0.08 to 0.32)	0.001
African descent	131	42	28	39	43	−0.01 (−0.20 to 0.18)	0.93

ARDS, acute respiratory distress syndrome; RD, risk difference.

^aAnalyses limited to patients who were not transfused any blood products during the first 24 hours after intensive care unit admission.

^bAnalyses limited to patients who did not develop ARDS during the first 5 days of hospital admission.

^cAnalyses limited to patients who did not have evidence of acute renal dysfunction by American College of Chest Physicians severe sepsis criteria at intensive care unit enrollment.

^dAKI risk by ABO blood type on the basis of multivariable logistic regression models adjusted for Acute Physiology, Age, Chronic Health Evaluation III score without renal components, age, sex, history of diabetes, history of congestive heart failure, malignancy, pulmonary source of infection, and units of red blood cells transfused in the first 24 hours of intensive care unit admission.

^eRDs are the adjusted differences in AKI risk between blood types A and O. All other comparisons between blood types A, B, AB, and O were not statistically significant and are displayed in Supplemental Table 1.

Discussion

In this study, we showed that ABO blood type A was associated with an estimated 14% higher risk of developing AKI in the ICU compared with type O among individuals of European descent in two critically ill populations: major trauma and severe sepsis. The ABO and AKI risk associations were also present within patients who were not transfused, showing the independence of the relationship from characteristics of blood transfusion (e.g., storage duration, volume, and compatible versus identical blood type). Additionally, our findings were similar in patients who did not develop ARDS, indicating an association between ABO and AKI independent of ARDS. Our findings are notable for several reasons. First, they point toward the possibility that ABO glycans may be important mediators of AKI. Second, the presence of a similar ABO-AKI association in both patients with sepsis and patients with trauma, despite differences in study population, comorbidities, and years of enrollment, indicates that the underlying mechanisms may be shared in these distinct populations. Placed in the context of prior demonstrations by us and others that blood type A is a risk factor for vascular diseases and ARDS, our findings also suggest that ABO blood types may have common effects on multiple organ dysfunctions in critical illness (13–16).

Prior literature on the ABO histo-blood group and AKI had focused on the now rare renal injury from ABO-incompatible transfusions and renal allograft dysfunction after ABO-incompatible transplantation (40,41). Our findings of an association between ABO blood type and AKI likely reflect a different underlying pathophysiology. We hypothesize that the ABO glycans increase AKI risk through their effects on inflammation, endothelial function, and microvascular coagulation (6,7). ABO blood types are major genetic determinants of blood concentrations of soluble ICAM-1, selectins, vWF, and thrombomodulin (12,20–23), and several of these proteins are known to be glycated with the ABO antigens in renal vascular endothelium (19). Animal models of renal ischemia-reperfusion injury show endothelial upregulation of integrins, selectins, and ICAM-1 (5,42), downregulation of thrombomodulin (43), and secretion of vWF (44). Although the exact biologic role of the ABO glycans is unknown, their effects on these endothelial mediators may underlie the relationship between ABO blood type and AKI. As treatments are developed targeting these mechanisms, blood type A may represent an easily identifiable population to test novel, tailored therapeutics aimed at glycoproteins possessing the ABO antigens.

We identified a significant association between blood type A and higher AKI risk only in patients of European descent, similar to the association between ABO blood type and ARDS. There are several potential explanations for this finding. African descent is an independent risk factor for AKI (29,45). It is possible that, in individuals of African descent, any protective effects of blood type O are overwhelmed by other mechanisms that increase their AKI risk. Alternatively, an unidentified genetic or environmental factor, with distinct frequencies between individuals of European and African descent, may modify the ABO-AKI association depending on race. For example, polymorphisms in the vWF gene are associated with expression of vWF in healthy Americans of European but not African descent, suggesting racial divergence in the genetic regulation of vWF expression (46,47) and possibly, contributing to the racial difference in ABO-AKI associations. The baseline characteristics of our cohorts also exhibit racial differences. Among patients with trauma, individuals of African descent were significantly more likely to suffer penetrating trauma; in sepsis, they were significantly less likely to have certain comorbidities, particularly hematologic malignancies. However, adjusting for these clinical characteristics did not explain the observed racial differences in our study. Lastly, because our study only had power to detect an 18%–20% difference in AKI risk by blood type among patients of African descent, it is possible that a smaller but clinically relevant association of ABO blood type and AKI went undetectable.

Our study has several important limitations. The two patient cohorts were enrolled at a single center, and generalizability to other centers may be limited. Additionally, enrollment procedures, including inclusion and exclusion criteria and era of enrollment, were different in the cohorts. However, we found a similar risk pattern by blood type in two distinct critically ill patient populations. Our study warrants replication at other centers and in additional at–risk populations. The AKI outcome definition did not include urine output criteria. These criteria are noted to be the least specific for AKI and therefore, may be suboptimal for studies of potential AKI pathophysiologic mechanisms (48). Additionally, we did not use premorbid baseline creatinine values given the high frequency of missingness that was disproportionately in patients with trauma. This approach sacrificed some sensitivity for admission creatinine elevations that did not subsequently worsen but avoided the lack of specificity that comes with using premorbid creatinine estimates. Sample size and cohort composition limited our ability to reach conclusions among the less common blood types (B and AB) and racial groups. Similarly, we did not have statistical power to determine the effects of genetic interactions on the ABO-AKI relationship. Severe sepsis is defined by sepsis in the setting of acute organ dysfunction, including renal dysfunction (32). Given that our inclusion criteria included renal dysfunction, we may have introduced selection bias. However, renal dysfunction at ICU admission was not associated with ABO blood type, but rather, ABO blood type was associated with the subsequent development of AKI in the ICU. Additionally, our analyses, which were restricted to patients without acute renal dysfunction at enrollment, show an association of ABO blood type and AKI. Finally, despite consideration of multiple potential confounders for adjusted analyses, unmeasured confounders or linked genetic traits may have affected the association between ABO blood type and AKI.

In conclusion, our study identified ABO blood type A as a risk factor for the development of AKI in critically ill patients of European descent with trauma or severe sepsis. This association was independent of blood transfusion and other clinical confounders, implicating the ABO glycans in AKI risk among the critically ill. Future research is warranted to identify and detail mechanisms underlying this relationship.

Disclosures

None.