Abstract
Background
Preoperative anaemia is an important risk factor for adverse outcomes in cardiac surgery, however data on postoperative anaemia is sparse. The aim of this study is to characterise the association of postoperative haemoglobin with 30-day mortality and morbidity after cardiac surgery.
Methods
We performed a retrospective cohort study of adults (age ≥18 yr) undergoing coronary revascularisation, valve surgery, or a combination at Toronto General Hospital from 2016 to 2020. We analysed the association between nadir postoperative day 1 (POD1) haemoglobin as a continuous and binary variable (haemoglobin ≤80 g L−1), with a primary composite outcome of 30-day mortality, stroke, myocardial infarction, acute kidney injury, sternal wound infection, or a combination. The secondary outcome was the incidence of adverse events. The primary outcome was analysed using logistic regression, secondary using Poisson regression; adjusted models accounted for clustering and confounders.
Results
We included 5960 patients. On POD1, mean haemoglobin was 90.1g L−1 (standard deviation 15.2) and 1794 patients (30%) had haemoglobin ≤80 g L−1. Red blood cells were transfused to 49% of the cohort, and to 90% of patients with POD1 haemoglobin ≤80 g L−1. Each 10 g L−1 decrease in POD1 haemoglobin increased the odds of the primary outcome (adjusted odds ratio [OR] 1.15 [1.05–1.25], P<0.001), as did haemoglobin ≤80 g L−1 (adjusted OR 1.44 [1.19–1.75], P<0.001). For adverse events, each 10 g L−1 decrease in haemoglobin was associated with an increased incidence rate ratio (IRR) (adjusted IRR 1.14 [1.07–1.20], P<0.001), as was haemoglobin <80 g L−1 (adjusted IRR 1.33 [1.16–1.54], P<0.001).
Conclusions
In postoperative cardiac surgical patients, progressive decreases in postoperative haemoglobin are associated with increased risk of mortality and major morbidity at 30 days.
Keywords: anaemia, cardiac surgical procedures, humans, mortality, organs at risk, risk factors
Perioperative anaemia and subsequent red blood cell (RBC) transfusion are a common occurrence for patients undergoing cardiac surgery.1, 2, 3 Postoperative anaemia is ubiquitous after cardiac surgery and can exist independently of preoperative anaemia. While 20% of cardiac surgical patients have preoperative anaemia,4 40–90% will meet World Health Organization (WHO) criteria for moderate to severe anaemia after surgery.5 Given the increased mortality and morbidity associated with RBC transfusion in patients having cardiac surgery,3,6 many blood conservation strategies have been implemented with hopes of reducing transfusion rates and improving patient-centred outcomes.7,8 However, the independent prognostic importance and risk associated with postoperative anaemia is poorly characterised. Despite its high prevalence after cardiac surgery, few strategies have been studied or clinically adopted to facilitate prevention and resolution of postoperative anaemia. In this context, it is important to define the relationship between postoperative haemoglobin and outcomes after cardiac surgery, as existing evidence suggests that postoperative anaemia is sustained (>50 days after operation) even in patients undergoing routine procedures such as coronary artery bypass grafting (CABG).9
In cardiac surgical patients, preoperative anaemia is associated with a more than two-fold increase in postoperative mortality, infection, and acute kidney injury, and a 50% increased odds of stroke.10 Through the perioperative period, the patient is exposed to blood loss, haemodilution, diagnostic phlebotomy, and elevated hepcidin concentrations,11 resulting in 30–40% lower haemoglobin concentrations than before surgery and a mean postoperative haemoglobin of <100 g L−1.10 Although sparse, existing literature suggests an association between decreased postoperative haemoglobin and increased risk of adverse events, with preoperative anaemia and postoperative anaemia also potentially interacting to modify the risk of outcomes.5,12 Clarifying this relationship may facilitate identification of a subgroup of patients in whom adjunct therapies such as intravenous iron, low volume sample collection tubes, and strict avoidance of haemodilution may lead to improved outcomes.
To address this knowledge gap, we performed a single-centre cohort study utilising comprehensive health administrative data with the primary aim to estimate the association of postoperative haemoglobin on a composite of adverse surgical outcomes including stroke, myocardial infarction, acute kidney injury, sternal wound infection, and 30-day mortality. The secondary aim was to report the incidence of adverse events in this patient population.
Methods
Design and data sources
We performed a single-centre, retrospective cohort study of patients undergoing cardiac surgery at Toronto General Hospital (TGH). All data were obtained from existing clinical databases at TGH and The Institute for Clinical Evaluative Sciences (ICES). ICES is an independent research institute that houses anonymised linked administrative healthcare data for residents of Ontario, Canada. Institutional data containing patient characteristics, procedural characteristics, and transfusion data were linked deterministically using unique patient identifiers to administrative data in ICES. TGH databases used to create our dataset included the Cardiovascular (CV) Surgery Database, capturing all cardiac surgeries at TGH; Anesthesia Information System (CAIS) and DREAM (Developing Research Excellence in Anesthesia Management) anaesthesia databases, capturing all preoperative and intraoperative data; Enterprise Data Warehouse (EDW), housing electronic medical record and laboratory values; and the Blood Bank Database (Wellsky, Overland Park, Kansas), capturing transfusion data. ICES databases included the Canadian Institute of Health Information Discharge Abstract Database (CIHI-DAD), capturing all hospitalisations; Ontario Health Insurance Plan (OHIP) database, capturing physician service claims; the Registered Persons Database (RPDB), capturing patient characteristics and deaths; the National Ambulatory Care Reporting System (NACRS), capturing details of emergency and outpatient care; and the Ontario Laboratories Information System (OLIS), which records laboratory data.
Research ethics board approval from University Health Network (UHN) was obtained before data collection and analysis. Under section 45 of Ontario’s Personal Health Information Protection Act, we did not require informed patient consent to link TGH institutional databases with ICES databases. The study and findings are reported according to the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) and REporting of studies Conducted using Observational Routinely-collected health DATA (RECORD) statements.13,14
Cohort
We identified adult patients, aged ≥18 yr, who underwent an index cardiac surgical procedure between 1 January 2016 and 31 December 2020 at TGH. We included patients undergoing CABG, valvular procedures (repair or replacement), or a combination (CABG + valve). We excluded patients undergoing heart transplantation, mechanical circulatory support, isolated aortic surgery, or adult congenital cardiac surgery.
Exposure
Our primary exposure was the lowest plasma haemoglobin measured on postoperative day 1 (POD1), identified from EDW and OLIS data. Postoperative haemoglobin was retained as a continuous variable (g L−1). Given the high incidence of severe anaemia after cardiac surgery,5 postoperative anaemia was also dichotomously defined as a POD1 haemoglobin of ≤80 g L−1, consistent with the WHO definition for severe anaemia in non-pregnant adults.15 Only the index operation was utilised for determination of postoperative values for each patient.
Outcomes
Our primary outcome was a composite of adverse surgical events occurring within 30 days including stroke, myocardial infarction, acute kidney injury, sternal wound infections, and mortality. Secondary outcomes included a count of adverse events and each of the individual adverse surgical events. Stroke (binary), myocardial infarction (binary), and sternal wound infections requiring operative intervention (binary) were identified through the CV Surgery Database and ICES (CIHI-DAD and NACRS) via International Classification of Diseases (ICD) codes and are in keeping with STS outcome definitions (Appendix A). Acute kidney injury was classified, according to the Kidney Disease: Improving Global Outcomes (KDIGO) classification system, as an increase in creatinine to 1.5 times the baseline or greater.16 Postoperative creatinine values were obtained from the EDW and OLIS. Mortality was identified using the RPDB.
Covariates
Patients were characterised based on age, sex, height, weight, body mass index (BMI), neighbourhood income quintile, and comorbidities including: diabetes, alcohol use, chronic lung disease, liver disease, cancer, peripheral arterial disease, cerebrovascular disease (prior stroke or transient ischaemic attack), preoperative renal dysfunction, and preoperative dialysis, extracted from the CV Surgery Database and the CAIS/DREAM anaesthesia databases. Procedural characteristics including anonymised surgeon identifier, urgency, procedure type, cardiopulmonary bypass and aortic cross-clamp time, and length of procedure were collected through the CV Surgery Database and DREAM Anesthesia record. Laboratory values, including preoperative haemoglobin, creatinine, and the lowest recorded postoperative haemoglobin were collected through the CV Surgery Database, EDW, and OLIS. RBC transfusion was captured as the number of allogenic RBC units transfused intraoperatively and any time after operation; this was collected through the CV Surgery and Blood Bank Database.
Missing data
The variables used in this study were routinely collected in all patients before and after cardiac surgery, with a low rate of missing values. Postoperative haemoglobin, outcome data, income quintile, rurality, individual comorbidities, surgical urgency, BMI, preoperative haemoglobin, preoperative creatinine, cardiopulmonary bypass time, and cross-clamp time had a missing data rate of <1%. Given the low rate of missing data, and the assumption that data were missing at random, we performed mean imputation (for continuous variables) or the most common value (for categorical variables).
Sample size
An a priori power calculation was performed assuming a risk of 30-day mortality of 1% in the non-anaemic study population, a risk of 30-day mortality of 2% in an anaemic population after operation,17 and a conservative baseline prevalence of postoperative anaemia of 50%,5 a sample size of ∼4700 patients would be required to detect a difference in mortality, assuming a two-sided alpha=0.05 and a power=0.80.
Statistical analysis
Version 9.4 of the SAS System for Windows (SAS Institute Inc., Cary, NC, USA) was used to perform all analyses. Patient characteristics were reported using mean and standard deviations (sd) for normally distributed data, otherwise median and inter-quartile ranges were reported. Categorical variables were presented as counts and percentages. Baseline characteristics between those with and without postoperative anaemia (haemoglobin on POD1 ≤80 g L−1) were compared using absolute standardised differences (where values >0.1 are suggested to represent an imbalance between groups).18
Our primary analysis examined the association of POD1 haemoglobin concentration with a composite of adverse surgical outcomes including in-hospital stroke, myocardial infarction, acute kidney injury, sternal wound infection, and mortality. Postoperative haemoglobin concentration (g L−1) was retained as a continuous variable to avoid information loss and assumptions inherent in categorisation of a continuous variable, with adjusted outcomes reported per 10 g L−1 decrease in haemoglobin. We also examined postoperative anaemia, defined as POD1 haemoglobin concentration ≤80 g L−1, as a dichotomous exposure for our outcomes.
Unadjusted and adjusted analyses were conducted utilising logistic regression. Given that patients are nested by surgeon within TGH, generalised estimating equations were used to appropriately adjust standard errors for within-surgeon clustering. We report multiplicative effect of an incremental decrease in our exposure (each 10 g L−1 decrease in haemoglobin) on our outcome. Multivariable models were adjusted for age, sex, procedure, surgical urgency, preoperative haemoglobin concentration, preoperative creatinine, BMI, cardiopulmonary bypass time, aortic cross-clamp time, and preoperative comorbidities listed above (Appendix B). Model fit and convergence were examined. Findings are presented as odds ratio (OR), 95% confidence intervals (CI), with corresponding P-values. Statistical significance was defined as P<0.05. The same procedures detailed above were followed for our individual secondary outcomes, treating postoperative haemoglobin as both a continuous and dichotomous exposure. Our secondary outcome involving a count of adverse events was analysed using Poisson regression. Our count outcome is reported as an incidence rate ratio (IRR), with the corresponding 95% CI and P-values. The interaction of sex and haemoglobin was examined using interaction terms in the multivariable models.
In a prespecified sensitivity analysis, we performed a propensity score approach for our primary outcome. We estimated a propensity score for POD1 haemoglobin ≤80 g L−1 using a logistic regression model including aforementioned covariates. We then used overlap weighting, where the assigned weight to each patient is the probability of being assigned to the opposite group.19 This results in smaller weights for extreme propensity score values, while patients with more common characteristics contribute more to estimation.20 Secondly, using restricted cubic splines, we set to explore if an inflection point might exist with POD1 haemoglobin concentrations and our primary outcome, through visualisation of the association of the continuous, non-linear haemoglobin value vs model predicted probability of primary outcome. Where multiple inflection points existed, we used the most prominent inflection point of increased risk acceleration, where the second derivative of the risk function changed its sign. We also conducted a sensitivity analysis testing linear splines with a knot at 120 g L−1, as risk appeared to level off at this threshold.
A post hoc sensitivity analysis was also performed to estimate whether varying inflection points existed based on patient sex. Specifically, we wanted to examine if males and females have differing thresholds of postoperative haemoglobin where their risk for an adverse event is increased. This was decided as current literature demonstrates that females have lower preoperative and postoperative haemoglobin concentrations, and higher rates of transfusion after major surgery.21,22
Lastly, our primary model did not include transfusion as a covariate as we felt it would be on the causal pathway; however, a sensitivity analysis was performed including the total intraoperative and POD0 RBC transfusions (continuous predictor). We did not include transfusions occurring beyond POD0 as these transfusions would have occurred after the exposure (POD1 haemoglobin).
Results
We identified 5960 consecutive adults aged ≥18 yr who underwent cardiac surgery during the study period (Fig. 1). The mean POD1 haemoglobin was 90.1 g L−1 (sd 15.2); 1794 (30.1%) patients had a POD1 haemoglobin ≤80g L−1. RBC transfusions were administered to 2923 (49.0%) patients. Mean preoperative haemoglobin of all patients was 136.4 g L−1 (sd 20.2). Baseline characteristics (Table 1) demonstrate patients with lower postoperative haemoglobin were more likely to be older, female, have a lower preoperative haemoglobin, have multiple comorbidities, undergo valve or combined procedures, and have non-elective surgery. For the total cohort at 30 days, 756 (12.3%) patients experienced the primary composite outcome, 158 (2.7%) had a stroke, 81 (1.4%) had a myocardial infarction, 420 (7.1%) had an acute kidney injury, 141 (2.4%) had a sternal wound infection, and 150 (2.5%) died.
Fig 1.
Study flow chart for patients included in final analysis. TGH, Toronto General Hospital
Table 1.
Baseline cohort characteristics by postoperative haemoglobin concentration. AIDS, acquired immunodeficiency syndrome; BMI, body mass index; CABG, coronary artery bypass grafting; CPB, cardiopulmonary bypass; Hb, haemoglobin; HIV, human immunodeficiency virus; sd, standard deviation; IQR, inter-quartile range.∗Indicates suppression as cell count <6.
| Postoperative Hb ≤80 g L−1 (n=1794) | Postoperative Hb >80 g L−1 (n=4166) | Absolute standardised difference | ||
|---|---|---|---|---|
| Female, n (%) | 609 (34.0) | 1011 (24.3) | 0.21 | |
| Age group (yr), n (%) | 18–39 | 134 (7.5) | 311 (7.5) | 0 |
| 40–64 | 662 (36.9) | 1835 (44.0) | 0.15 | |
| 65–69 | 998 (55.6) | 2020 (48.5) | 0.14 | |
| BMI, mean (sd) | 27.7 (5.7) | 28.4 (5.4) | 0.11 | |
| Preoperative haemoglobin (g L−1), mean (sd) | 127.3 (19.6) | 140.3 (19.2) | 0.66 | |
| Preoperative creatinine (μmol L−1), median (IQR) | 88.0 (73.0–109.0) | 83.9 (73.0–98.0) | 0.24 | |
| Rural, yes (%) | 212 (11.9) | 474 (11.4) | 0.02 | |
| Income quintile (%) | Lower two quintiles | 799 (44.5) | 1617 (38.8) | 0.11 |
| Middle quintile | 355 (19.9) | 895 (21.5) | 0.04 | |
| Comorbidities, n (%) | Myocardial infarction | 465 (26.0) | 813 (19.5) | 0.16 |
| Congestive heart failure | 486 (27.1) | 638 (15.3) | 0.29 | |
| Peripheral vascular disease | 260 (14.5) | 724 (17.4) | 0.08 | |
| Cerebrovascular disease | 162 (9.1) | 255 (6.1) | 0.11 | |
| Dementia | 1–5∗ | 1–5∗ | 0.03 | |
| Chronic obstructive pulmonary disease | 152 (8.5) | 342 (8.2) | 0.01 | |
| Rheumatic disease | 16 (0.9) | 34 (0.8) | 0.01 | |
| Peptic ulcer disease | 39 (2.2) | 51 (1.2) | 0.08 | |
| Liver disease | 40 (2.1) | 61 (1.5) | 0.05 | |
| Diabetes mellitus | 621 (34.7) | 1118 (26.9) | 0.17 | |
| Hemiparesis | 14 (0.8) | 20 (0.5) | 0.04 | |
| Renal disease | 156 (8.7) | 165 (4.0) | 0.19 | |
| Primary cancer | 68 (3.8) | 129 (3.1) | 0.04 | |
| Metastatic cancer | 8 (0.5) | 18 (0.4) | 0.01 | |
| HIV/AIDS | 0 (0) | 8 (0.2) | 0.06 | |
| Charlson Index, n (%) | 0 | 436 (24.3) | 1364 (32.8) | 0.19 |
| 1–3 | 1042 (58.1) | 2436 (58.5) | 0.01 | |
| 4+ | 313 (17.4) | 360 (8.6) | 0.26 | |
| Procedure, n (%) | CABG | 878 (48.9) | 1786 (42.9) | 0.12 |
| Valve | 622 (34.7) | 1992 (47.8) | 0.27 | |
| CABG + valve | 294 (16.4) | 388 (9.3) | 0.21 | |
| Urgency, n (%) | Elective | 1030 (57.5) | 2999 (72.0) | 0.31 |
| Urgent | 665 (37.1) | 1052 (25.3) | 0.26 | |
| Emergent | 89 (5.0) | 108 (2.6) | 0.13 | |
| Cross-clamp time (min), median (IQR) | 82.0 (62.0–108.0) | 73.0 (54.0–96.0) | 0.25 | |
| CPB time (min), median (IQR) | 103.0 (80.0–136.0) | 93.0 (71.0–121.0) | 0.27 | |
| Procedure time (min), median (IQR) | 238.0 (193.0–302.0) | 208.0 (166.0–259.0) | 0.36 | |
Postoperative haemoglobin and adverse outcomes
Table 2 provides the unadjusted and adjusted OR and 95% CI for the primary and secondary outcomes. Each 10 g L−1 decrease in postoperative haemoglobin increased odds of the primary composite outcome at 30 days in both unadjusted (OR 1.44, 95% CI 1.36–1.53, P<0.001) and adjusted (OR 1.15, 95% CI 1.05–1.25; P<0.001) models. This was consistent when examining a dichotomous haemoglobin cut-off of ≤80 g L−1 in unadjusted (OR 2.44, 95% CI 2.10–2.84, P<0.001) and adjusted (OR 1.44, 95% CI 1.19–1.75, P<0.001) models.
Table 2.
Unadjusted and adjusted odds ratio, with 95% confidence intervals, for postoperative haemoglobin and adverse outcomes. ∗Adjusted for age, sex, surgical urgency, preoperative haemoglobin concentration, preoperative creatinine, procedure type, body mass index, cardiopulmonary bypass time, aortic cross-clamp time, and preoperative comorbidities.
| Per 10 g L−1 decrease in haemoglobin |
Postoperative anaemia (haemoglobin ≤80 g L−1) |
|||||||
|---|---|---|---|---|---|---|---|---|
| Unadjusted | P-value | Adjusted∗ | P-value | Unadjusted | P-value | Adjusted∗ | P-value | |
| Composite outcome | 1.44 (1.36–1.53) | <0.001 | 1.15 (1.05–1.25) | 0.002 | 2.44 (2.10–2.84) | <0.001 | 1.44 (1.19–1.75) | <0.001 |
| Mortality | 1.75 (1.51–2.07) | <0.001 | 1.29 (1.09–1.53) | 0.006 | 3.83 (2.74–5.34) | <0.001 | 2.00 (1.68–2.39) | <0.001 |
| Myocardial infarction | 1.19 (1.02–1.40) | 0.031 | 1.09 (0.91–1.31) | 0.358 | 1.61 (1.03–2.51) | 0.037 | 1.27 (0.77–2.09) | 0.359 |
| Stroke | 1.31 (1.16–1.47) | <0.001 | 1.17 (1.02–1.34) | 0.027 | 1.65 (1.19–2.27) | 0.002 | 1.29 (0.91–1.84) | 0.162 |
| Acute kidney injury | 1.62 (1.49–1.76) | <0.001 | 1.22 (1.11–1.35) | <0.001 | 2.66 (2.18–3.25) | <0.001 | 1.37 (1.15–1.63) | 0.011 |
| Wound infection | 1.26 (1.11–1.42) | <0.001 | 1.03 (0.90–1.18) | 0.301 | 2.08 (1.49–2.91) | <0.001 | 1.32 (0.92–1.91) | 0.143 |
For our secondary outcomes, unadjusted analysis showed a significant association between each 10 g L−1 decrease in haemoglobin and increased risk of mortality (OR 1.75, 95% CI 1.51–2.07, P<0.001), myocardial infarction (OR 1.19, 95% CI 1.02–1.40, P=0.031), stroke (OR 1.31, 95% CI 1.16–1.47, P<0.001), acute kidney injury (OR 1.62, 95% CI 1.49–1.76, P<0.001), and sternal wound infection (OR 1.26, 95% CI 1.11–1.42, P<0.001). Multivariable analysis showed for each 10 g L−1 decrease in haemoglobin an increased risk of mortality (OR 1.29, 95% CI 1.09–1.53, P<0.001), stroke (OR 1.17, 95% CI 1.02–1.34, P=0.027), and acute kidney injury (OR 1.22, 95% CI 1.11–1.35, P<0.001).
When examining a haemoglobin ≤80 g L−1, unadjusted analysis showed an increased risk of mortality, myocardial infarction, stroke, acute kidney injury, and sternal wound infection (Table 2). After adjustment, there was a significant association between a haemoglobin ≤80 g L−1 and mortality (OR 2.00, 95% CI 1.68–2.39, P<0.001) and acute kidney injury (OR 1.37, 95% CI 1.15–1.63, P=0.011).
In examining a count of complications, each 10 g L−1 decrease in postoperative haemoglobin resulted in an increased IRR in unadjusted (IRR 1.42, 95% CI 1.34–1.51, P<0.001) and adjusted (IRR 1.14, 95% CI 1.07–1.20, P<0.001) models. This relationship was consistent for a haemoglobin of ≤80 g L−1 in unadjusted (IRR 2.29, 95% CI 2.01–2.60, P<0.001) and adjusted (IRR 1.33, 95% CI 1.16–1.54, P<0.001) models.
Sensitivity analysis
Our results were consistent when using a propensity score approach to adjust for confounders, finding that individuals with a haemoglobin of ≤80 g L−1 had 25% increased odds of the primary outcome (OR 1.25, 95% CI 1.07–1.46, P<0.001). Including the total number of RBC transfused as a covariate did not impact the effect estimate (Appendix Table 1).
The predicted probability of the primary outcome for any given concentration of haemoglobin is shown in Figure 2. The greatest increase in risk of the primary outcome is seen at a haemoglobin of <100 g L−1. Additional sensitivity analysis testing linear splines with a knot at 120 g L−1 is included in Appendix Table 2. There was no interaction effect for sex and postoperative haemoglobin concentrations as demonstrated in Figure 3.
Fig 2.
Adjusted predicted probability, and 95% confidence intervals, of primary outcome as a function of postoperative haemoglobin.
Fig 3.
Sex-specific analysis of adjusted predicted probability, and 95% confidence intervals, of primary outcome as a function of postoperative haemoglobin.
Discussion
In this retrospective cohort study, after adjusting for prespecified confounders, lower concentrations of haemoglobin after cardiac surgery are associated with a significant increase in the risk of a composite outcome of death, myocardial infarction, stroke, acute kidney injury, and sternal wound infection at 30 days. This finding was directionally consistent across each component of the composite outcome, when morbidity events were analysed as a count, and when using propensity score adjustment. Additionally, non-linear analyses suggest that risk associated with postoperative anaemia may occur at postoperative haemoglobin concentrations as high as 100 g L−1. These data provide an important overview of the negative associations of low haemoglobin concentrations after cardiac surgery and suggest a need to evaluate strategies to address postoperative anaemia in supporting optimised recovery.
Evaluating the association of postoperative haemoglobin concentrations with outcomes is a complex undertaking, as preoperative and intraoperative confounders, along with care processes such as transfusion, will influence both exposure and outcome. Importantly, our adjusted models were prespecified, and accounted for key preoperative factors, including preoperative haemoglobin, and intraoperative duration and complexity, while we avoided including mediators of outcome, such as transfusion, which could lead to overadjustment bias.23 In a post hoc sensitivity analysis we adjusted for the number of RBC transfusions intraoperatively and on POD0, and this had no change on the effect estimate (Appendix Table 1). This suggests an effect of postoperative anaemia, independent of RBC transfusion, on patient outcomes. These findings are consistent with previous work, including mediation analysis, which has suggested that both preoperative anaemia and transfusion have independent contributions to adverse outcomes in cardiac surgery, and that preoperative anaemia, regardless of transfusion, is associated with length of stay and acute kidney injury.6 Furthermore, we found that there was a significant increase in the risk of major adverse outcomes with a postoperative haemoglobin concentration of <100 g L−1. This threshold is slightly lower than other studies that have found increasing risk at <110 g L−1,23,24 however these studies did not limit their populations to cardiac surgical patients.
Haemoglobin is critical in delivery of oxygen to end-organ tissue and an anaemic state can contribute to tissue hypoxia.25 Given the postoperative inflammatory and hypermetabolic state after cardiac surgery,26 this increased oxygen demand likely amplifies the effects of postoperative anaemia and decreased oxygen delivery—increasing the risk of adverse outcomes. Our findings support this hypothesis and are in keeping with previous work done at our institution using local data,27 and previously published literature demonstrating that the majority of patients tested for postoperative anaemia have iron deficiency after cardiac surgery.12 Our findings are also consistent with data from the heart failure population, a group with significant overlap with cardiac surgery patients, where the 2022 AHA/ACC/HFSA Guideline for the Management of Heart Failure provides a Class IIa (moderate strength of recommendation) and Level of Evidence B-R (moderate quality of evidence from at least one randomised study) recommendation that in patients with heart failure and reduced ejection fraction, treatment of iron deficiency and anaemia with i.v. iron is reasonable to improve functional status and quality of life.28 Clinical trials in the heart failure population addressing treatment of iron deficiency and anaemia with i.v. iron (FAIR-HF, CONFIRM-HF, AFFIRM-AHF, and IRON-HF)29, 30, 31 have demonstrated improvements in symptoms as assessed by New York Heart Association (NYHA) class, exercise capacity, and quality of life. Future interventional trials in cardiac surgery patients assessing the impact of therapies for preoperative and postoperative anaemia such as i.v. iron may help determine the optimal treatment regimens, which can additionally be targeted in the postoperative period.32 In major abdominal surgery, preoperative i.v. iron, compared with a placebo infusion, reduced postoperative anaemia rates and re-admission rates over the 6 months after surgery.33 Our work highlights that targeting postoperative anaemia is an opportunity for improvement in patient outcomes, however does not preclude optimisation before surgery. The routine practice at TGH is to refer all anaemic patients for preoperative patient blood management and to perform retrograde autologous priming in all cardiac surgery cases where a patient is eligible (able to tolerate haemodynamically and not severely anaemic pre procedure).
Our study has several strengths and limitations. In using ICES administrative data, we were able to utilise a consecutive cohort of patients, have a complete episode of care, accurate exposure identification (through linking local and provincial administrative data), and outcomes derived from gold standard reference sources (i.e. vital statistics). Our rate of missing data for exposures, covariates, and outcomes was <5%, improving the internal validity of our work. While our observational effect estimates were robust using both hierarchical regression and propensity score approaches, results represent association, not causation. Additionally, despite confounder control, other unmeasured confounders could exist, biasing our results. Of interest, there was a U-curve of risk for females which is difficult to explain given the rarity of high haemoglobin concentrations in this specific subpopulation. The low sample size at the extremes of haemoglobin makes the data difficult to interpret. In analysis, including unadjusted predicted probabilities, these higher haemoglobin concentrations and increased risk could potentially be attributable to over-transfusion.34 Moreover, our definition of postoperative myocardial infarction was based on the current STS definition (Appendix A), and not in keeping with recent recommendations surrounding the use of postoperative troponin. This limitation was largely because of the timeline, as most patients included did not have routinely measured postoperative troponin. Lastly, the nature of our data did not allow us to investigate the cause of anaemia, whether that be related to surgical bleeding or haemodilution. Further work should incorporate longer-term follow-up and patient-centred outcomes, and explore the aetiology of postoperative anaemia to help guide treatment and prevention, including adjunct therapies beyond blood transfusion.32
Conclusions
In postoperative cardiac surgery patients, each additional 10 g L−1 decrease in haemoglobin is associated with an increased composite risk of stroke, myocardial infarction, acute kidney injury, sternal wound infection, and mortality at 30 days. This identifies a high-risk group of patients in whom adjunct therapies for prevention and mitigation of the severity of postoperative anaemia may lead to improved outcomes. However, high-quality data from interventional randomised trials are needed before clinical recommendations for interventions can be made.
Authors’ contributions
Conception: JA, JB, KK
Study design: JA, JB, KK
Data acquisition: JA, JB, KK, SAM, RS
Data analysis: JA, JB
Data interpretation: JA, JB, KK, MO, SAM, DIM, MC, AG, JC, YL
Drafting of the manuscript: JA, JB
Revision of the manuscript: JA, JB, KK, MO, SM, RS, DIM, MC, AG, JC, YL, SAM
Approval of the final manuscript version: JA, JB, KK, MO, SM, RS, DIM, MC, AG, JC, YL, SAM
Guarantor: JA, JB
Declarations of interest
JB has reported receiving research support from Canadian Blood Services, the Society for the Advancement of Blood Management, and personal fees from Grifols and Octapharma outside the submitted work. KK reported receiving personal fees from Octapharma and Werfen outside the submitted work. JC reported receiving research grants from Canadian Blood Services and Octapharma, and personal fees from Octapharma and Pfizer outside of the submitted work. YL reported receiving research grants from Canadian Blood Services and Octapharma and personal fees from Pfizer outside the submitted work.
Acknowledgements
This study was supported by ICES, which is funded by an annual grant from the Ontario Ministry of Health (MOH) and the Ministry of Long-Term Care (MLTC). This study contracted ICES Data & Analytic Services (DAS) and used de-identified data from the ICES Data Repository, which is managed by ICES with support from its funders and partners: Canada’s Strategy for Patient-Oriented Research (SPOR), the Ontario SPOR Support Unit, the Canadian Institutes of Health Research, and the Government of Ontario. The opinions, results, and conclusions reported are those of the authors. No endorsement by ICES or any of its funders or partners is intended or should be inferred. Parts of this material are based on data provided by CIHI, information compiled and provided by CIHI, or both. However, the analyses, conclusions, opinions, and statements expressed in the material are those of the author(s), and not necessarily those of CIHI. Parts of this material are based on data and information compiled and provided by the Ontario Ministry of Health. The analyses, conclusions, opinions, and statements expressed herein are solely those of the authors and do not reflect those of the funding or data sources; no endorsement is intended or should be inferred. This document used data adapted from the Statistics Canada Postal CodeOM Conversion File, which is based on data licensed from Canada Post Corporation, data adapted from the Ontario Ministry of Health Postal Code Conversion File, or both, which contains data copied under licence from ©Canada Post Corporation and Statistics Canada.
Handling Editor: Susan M. Goobie
Footnotes
Supplementary data to this article can be found online at https://doi.org/10.1016/j.bjao.2025.100508.
Appendix A. Supplementary data
The following is the Supplementary data to this article:
References
- 1.Bennett-Guerrero E., Zhao Y., O’Brien S.M., et al. Variation in use of blood transfusion in coronary artery bypass graft surgery. JAMA. 2010;304:1568–1575. doi: 10.1001/jama.2010.1406. [DOI] [PubMed] [Google Scholar]
- 2.Murphy G.J., Pike K., Rogers C.A., et al. Liberal or restrictive transfusion after cardiac surgery. N Engl J Med. 2015;372:997–1008. doi: 10.1056/NEJMoa1403612. [DOI] [PubMed] [Google Scholar]
- 3.Murphy G.J., Reeves B.C., Rogers C.A., Rizvi S.I.A., Culliford L., Angelini G.D. Increased mortality, postoperative morbidity, and cost after red blood cell transfusion in patients having cardiac surgery. Circulation. 2007;116:2544–2552. doi: 10.1161/CIRCULATIONAHA.107.698977. [DOI] [PubMed] [Google Scholar]
- 4.Hazen Y.J.J.M., Noordzij P.G., Gerritse B.M., et al. Preoperative anaemia and outcome after elective cardiac surgery: a Dutch national registry analysis. Br J Anaesth. 2022;128:636–643. doi: 10.1016/j.bja.2021.12.016. [DOI] [PubMed] [Google Scholar]
- 5.Li M.M., Miles S., Callum J., Lin Y., Karkouti K., Bartoszko J. Postoperative anemia in cardiac surgery patients: a narrative review. Can J Anaesth. 2024;71:408–421. doi: 10.1007/s12630-023-02650-9. [DOI] [PubMed] [Google Scholar]
- 6.Warner M.A., Hanson A.C., Schulte P.J., et al. Preoperative anemia and postoperative outcomes in cardiac surgery: a mediation analysis evaluating intraoperative transfusion exposures. Anesth Analg. 2024;138:728–737. doi: 10.1213/ANE.0000000000006765. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 7.Guinn N.R., Schwartz J., Arora R.C., et al. Perioperative Quality Initiative and Enhanced Recovery After Surgery-Cardiac Society consensus statement on the management of preoperative anemia and iron deficiency in adult cardiac surgery patients. Anesth Analg. 2022;135:532–544. doi: 10.1213/ANE.0000000000006148. [DOI] [PubMed] [Google Scholar]
- 8.Padmanabhan H., Siau K., Curtis J., et al. Preoperative anemia and outcomes in cardiovascular surgery: systematic review and meta-analysis. Ann Thorac Surg. 2019;108:1840–1848. doi: 10.1016/j.athoracsur.2019.04.108. [DOI] [PubMed] [Google Scholar]
- 9.Westenbrink B.D., Kleijn L., de Boer R.A., et al. Sustained postoperative anaemia is associated with an impaired outcome after coronary artery bypass graft surgery: insights from the IMAGINE trial. Heart. 2011;97:1590–1596. doi: 10.1136/heartjnl-2011-300118. [DOI] [PubMed] [Google Scholar]
- 10.Karkouti K., Wijeysundera D.N., Beattie W.S. Reducing Bleeding in Cardiac Surgery (RBC) Investigators. Risk associated with preoperative anemia in cardiac surgery: a multicenter cohort study. Circulation. 2008;117:478–484. doi: 10.1161/CIRCULATIONAHA.107.718353. [DOI] [PubMed] [Google Scholar]
- 11.Bisbe E. Postoperative anemia: is there a role for iron replacement therapy? Blood Transfus. 2025;23:42–46. doi: 10.2450/BloodTransfus.906. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 12.Zhou L., Liu X., Yan M., et al. Postoperative nadir hemoglobin and adverse outcomes in patients undergoing on-pump cardiac operation. Ann Thorac Surg. 2021;112:708–716. doi: 10.1016/j.athoracsur.2021.01.016. [DOI] [PubMed] [Google Scholar]
- 13.von Elm E., Altman D.G., Egger M., et al. The Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) statement: guidelines for reporting observational studies. Lancet. 2007;370:1453–1457. doi: 10.1016/S0140-6736(07)61602-X. [DOI] [PubMed] [Google Scholar]
- 14.Benchimol E.I., Smeeth L., Guttmann A., et al. The REporting of studies Conducted using Observational Routinely-collected health Data (RECORD) statement. PLoS Med. 2015;12 doi: 10.1371/journal.pmed.1001885. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 15.WHO . 2011. Haemoglobin Concentrations for the Diagnosis of Anaemia and Assessment of Severity. [Google Scholar]
- 16.Khwaja A. KDIGO clinical practice guidelines for acute kidney injury. Nephron Clin Pract. 2012;120:c179–c184. doi: 10.1159/000339789. [DOI] [PubMed] [Google Scholar]
- 17.Manlhiot C., Rao V., Rubin B., Lee D.S. Comparison of cardiac surgery mortality reports using administrative and clinical data sources: a prospective cohort study. CMAJ Open. 2018;6:E316–E321. doi: 10.9778/cmajo.20180072. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 18.Austin P.C. Using the standardized difference to compare the prevalence of a binary variable between two groups in observational research. Commun Stat Simul Comput. 2009;38:1228–1234. [Google Scholar]
- 19.Li F., Thomas L.E., Li F. Addressing extreme propensity scores via the overlap weights. Am J Epidemiol. 2019;188:250–257. doi: 10.1093/aje/kwy201. [DOI] [PubMed] [Google Scholar]
- 20.Thomas L.E., Li F., Pencina M.J. Overlap weighting. JAMA. 2020;323:2417. doi: 10.1001/jama.2020.7819. [DOI] [PubMed] [Google Scholar]
- 21.Skorupski C.P., Cheung M.C., Lin Y. Preoperative anemia in major elective surgery. Can Med Assoc J. 2023;195:E551. doi: 10.1503/cmaj.221635. E551. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 22.Arya S., Howell A., Vernich L., Lin Y., Pavenski K., Freedman J. Re-evaluating treatment thresholds in patient blood management: female patients experience more perioperative anaemia and higher transfusion rates in major elective surgery. Vox Sang. 2024;119:1090–1095. doi: 10.1111/vox.13717. [DOI] [PubMed] [Google Scholar]
- 23.Myles P.S., Richards T., Klein A., et al. Postoperative anaemia and patient-centred outcomes after major abdominal surgery: a retrospective cohort study. Br J Anaesth. 2022;129:346–354. doi: 10.1016/j.bja.2022.06.014. [DOI] [PubMed] [Google Scholar]
- 24.Warner M.A., Hanson A.C., Plimier C., et al. Association between anaemia and hospital readmissions in patients undergoing major surgery requiring postoperative intensive care. Anaesthesia. 2023;78:45–54. doi: 10.1111/anae.15859. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 25.Hare GMT, Han K, Leshchyshyn Y, et al. Potential biomarkers of tissue hypoxia during acute hemodilutional anemia in cardiac surgery: A prospective study to assess tissue hypoxia as a mechanism of organ injury. Can J Anaesth. [DOI] [PubMed]
- 26.Parolari A., Alamanni F., Juliano G., et al. Oxygen metabolism during and after cardiac surgery: role of CPB. Ann Thorac Surg. 2003;76(3):737–743. doi: 10.1016/S0003-4975(03)00683-0. [DOI] [PubMed] [Google Scholar]
- 27.Chen M, Li M, Ouzounian M, McCluskey S, Karkouti K, Bartoszko J. Predictors and prognostic value for in-hospital outcomes associated with postoperative anemia in cardiac surgery: A retrospective cohort study. [DOI] [PubMed]
- 28.Heidenreich P.A., Bozkurt B., Aguilar D., et al. AHA/ACC/HFSA Guideline for the Management of Heart Failure: A Report of the American College of Cardiology/American Heart Association Joint Committee on Clinical Practice Guidelines. Circulation. 2022;145:e895–e1032. doi: 10.1161/CIR.0000000000001063. 2022. [DOI] [PubMed] [Google Scholar]
- 29.Ponikowski P., Kirwan B.A., Anker S.D., et al. Ferric carboxymaltose for iron deficiency at discharge after acute heart failure: a multicentre, double-blind, randomised, controlled trial. Lancet. 2020;396:1895–1904. doi: 10.1016/S0140-6736(20)32339-4. [DOI] [PubMed] [Google Scholar]
- 30.Ponikowski P., van Veldhuisen D.J., Comin-Colet J., et al. Beneficial effects of long-term intravenous iron therapy with ferric carboxymaltose in patients with symptomatic heart failure and iron deficiency. Eur Heart J. 2015;36:657–668. doi: 10.1093/eurheartj/ehu385. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 31.Beck-da-Silva L., Piardi D., Soder S., et al. IRON-HF study: a randomized trial to assess the effects of iron in heart failure patients with anemia. Int J Cardiol. 2013;168:3439–3442. doi: 10.1016/j.ijcard.2013.04.181. [DOI] [PubMed] [Google Scholar]
- 32.Bartoszko J., Miles S., Ansari S., et al. Postoperative intravenous iron to treat iron-deficiency anaemia in patients undergoing cardiac surgery: a protocol for a pilot, multicentre, placebo-controlled randomized trial (the POAM trial) BJA Open. 2024;11 doi: 10.1016/j.bjao.2024.100303. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 33.Richards T., Baikady R.R., Clevenger B., et al. Preoperative intravenous iron to treat anaemia before major abdominal surgery (PREVENTT): a randomised, double-blind, controlled trial. Lancet. 2020;396:1353–1361. doi: 10.1016/S0140-6736(20)31539-7. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 34.Gombotz H., Schreier G., Neubauer S., Kastner P., Hofmann A. Gender disparities in red blood cell transfusion in elective surgery: a post hoc multicentre cohort study. BMJ Open. 2016;6 doi: 10.1136/bmjopen-2016-012210. [DOI] [PMC free article] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.