Association of perioperative blood transfusion with increased 1-year mortality after surgery: a research letter

Perioperative blood transfusion remains a cornerstone intervention for managing acute hemorrhage and critical anemia in the perioperative setting[1,2]. Accumulating evidence suggests that perioperative blood transfusion has a potential negative impact on outcomes in patients undergoing single-procedure surgeries, such as hip fracture repair and cancer-related operations[3,4]. However, the association between perioperative blood transfusion and 1-year mortality in patients undergoing various types of surgery remains unclear. Therefore, our study aimed to investigate the association. This retrospective cohort analysis was conducted using the electronic medical records of 97 443 adult patients who underwent surgery at a tertiary hospital in Singapore between 1 January 2012 and 31 October 2016[5]. The data were obtained from the Dryad Digital Library (doi: https://doi.org/10.5061/dryad.5772v). The original authors have licensed the ownership of the original data to the Dryad. The work has been reported in line with the STROCSS criteria[6]. Institutional Review Board approval was obtained from the hospital in Singapore, which waived the requirement for individual informed consent. Patients undergoing first-time cardiothoracic surgery, orthopedics, obstetrics, and gynecology, general surgery, otolaryngology, hand surgery, neurosurgery, colorectal surgery, urology, plastic surgery, and oral and maxillofacial surgery were included. Patients with missing values for the American Society of Anesthesiologists (ASA) physical status, revised cardiac risk index (RCRI), degree of preoperative anemia according to revised WHO Guidelines[7], kidney disease, or race were excluded. Extracted variables included sex, age, ASA physical status, type of anesthesia, race, priority of surgery, degree of preoperative anemia, and medical history of cerebrovascular accidents (CVA), ischemic heart disease (IHD), chronic heart failure (CHF), insulin use for diabetes mellitus (DM), and kidney disease, as well as surgical risk scores. The RCRI was calculated based on a history of CVA, IHD, CHF, insulin use for DM, creatinine >2 mg/dL, and high-risk surgery based on the 2022 European Society of Cardiology/European Society of Anesthesiology guidelines on non-cardiac surgery[8]. The primary exposure was perioperative blood transfusion, defined as the administration of red blood cell concentrate units during surgery and up to 1 month postoperatively. The primary outcome was 1-year mortality after surgery. Kaplan–Meier curves were generated to compare survival probabilities between patients who received perioperative blood transfusions and those who did not. The association between perioperative blood transfusion and 1-year mortality was assessed using a multivariate Cox proportional hazards model, with adjusting for sex, age, ASA physical status, type of anesthesia, race, priority of surgery, degree of preoperative anemia, severity of kidney disease, and RCRI score. In addition, propensity score methods, including propensity score matching, adjusted propensity score, inverse probability of treatment weighting (IPTW), standardized mortality ratio weighting (SMRW), pairwise algorithmic (PA), and overlap weight (OW) models were conducted to validate the association. The propensity score-matching included sex, age, ASA physical status, type of anesthesia, race, priority of surgery, degree of preoperative anemia, severity of kidney disease, and RCRI score, using a multivariable logistic regression model with 1:1 nearest-neighbor matching algorithm and a caliper width of 0.2. A standardized mean difference <0.1 was considered indicative of variable balance between the groups before and after matching. The IPTW, SMRW, PA, and OW models were applied to generate a weighted cohort, adjusting for baseline confounders to provide a more independent assessment of the association between perioperative blood transfusion and mortality, which was also conducted after multiple imputation of missing data. Stratified and interaction analyses were performed based on age (<70 or ≥70 years), sex (male or female), ASA physical status (<III or ≥III), type of anesthesia (general or regional), priority of surgery (elective or emergency), preoperative anemia (no anemia, mild anemia, moderate anemia, or severe anemia), and RCRI (<III or ≥III) after propensity score matching. Subgroup analysis adjusted for sex, age, ASA physical status, type of anesthesia, race, priority of surgery, degree of preoperative anemia, severity of kidney disease, and RCRI score, excluding stratification factor itself. Statistical analyses were performed using R software (http://www.R-project.org, The R Foundation) and Free Statistics software version 1.7.1. Statistical significance was set at P < 0.05. After excluding cases with missing variables, 54 933 patients were included in the entire cohort (Fig. 1). Among them, 3347 patients received perioperative blood transfusion, and 1934 died during the 1-year postoperative follow-up period. The matched cohort included 6478 patients (3239 in each group; Table 1). The 1-year mortality rate was higher in the perioperative blood transfusion group than that in the no perioperative blood transfusion group both in the entire cohort (14.8% vs. 2.8%) and in the matched cohort (14.0% vs. 9.7%), which mean approximately one death would be prevented for every 23 surgical patients with perioperative blood transfusion. Perioperative blood transfusion was associated with increased 1-year mortality in multiple Cox regression analysis [hazard ratio (HR) = 1.69; 95% confidence interval = 1.51–1.90] and in various propensity score analyses (HRs ranging from 1.45 to 2.32; Table 2), which was also validated after imputation of missing data (Supplemental Digital Content Table S1, available at: https://links.lww.com/JS9/G317). In addition, the association was consistent across different subgroups (Supplemental Digital Content Figure S1, available at: https://links.lww.com/JS9/G317). Meanwhile, Kaplan–Meier curves showed that patients in the perioperative blood transfusion group had significantly lower 1-year survival than those in the no perioperative blood transfusion group in the matched cohort (Supplemental Digital Content Figure S2, available at: https://links.lww.com/JS9/G317) and in excluded patients (Supplemental Digital Content Figure S3, available at: https://links.lww.com/JS9/G317). Moreover, the 1-year mortality trends were similar between exclusion and inclusion groups (Supplemental Digital Content Figure S4, available at: https://links.lww.com/JS9/G317). Figure 1.: Flowchart detailing the selection process for participants. Table 1 - Demographics and clinical characteristics before and after propensity score matching. Before matching After matching Characteristics No PBT (N = 51 586) PBT (N = 3347) SMD No PBT (N = 3239) PBT (N = 3239) SMD Sex 0.182 0.025 Female 26 011 (50.4) 1989 (59.4) 1963 (60.6) 1923 (59.4) Male 25 575 (49.6) 1358 (40.6) 1276 (39.4) 1316 (40.6) Age (years) 0.378 0.083 18–29 4848 (9.4) 186 (5.6) 220 (6.8) 179 (5.5) 30–49 14 298 (27.7) 640 (19.1) 550 (17.0) 638 (19.7) 50–69 23 176 (44.9) 1428 (42.7) 1406 (43.4) 1382 (42.7) ≥70 9264 (18) 1093 (32.7) 1063 (32.8) 1040 (32.1) ASA physical status 0.638 0.019 I 12 046 (23.4) 360 (10.8) 366 (11.3) 357 (11.0) II 30 954 (60) 1554 (46.4) 1509 (46.6) 1531 (47.3) III 8013 (15.5) 1221 (36.5) 1193 (36.8) 1173 (36.2) IV–VI 573 (1.1) 212 (6.3) 171 (5.3) 178 (5.5) Type of anesthesia 0.165 0.013 General anesthesia 43 815 (84.9) 3024 (90.3) 2907 (89.7) 2920 (90.2) Regional anesthesia 7771 (15.1) 323 (9.7) 332 (10.3) 319 (9.8) Race 0.138 0.015 Chinese 37 544 (72.8) 2527 (75.5) 2452 (75.7) 2449 (75.6) Malay 4900 (9.5) 324 (9.7) 316 (9.8) 307 (9.5) Indian 4497 (8.7) 175 (5.2) 172 (5.3) 172 (5.3) Others 4645 (9) 321 (9.6) 299 (9.2) 311 (9.6) Priority of surgery 0.140 0.008 Elective 42 186 (81.8) 2546 (76.1) 2521 (77.8) 2510 (77.5) Emergence 9400 (18.2) 801 (23.9) 718 (22.2) 729 (22.5) Preoperative anemia 1.281 0.068 No anemia 39 056 (75.7) 841 (25.1) 825 (25.5) 841 (26.0) Mild anemia 7490 (14.5) 665 (19.9) 694 (21.4) 665 (20.5) Moderate anemia 4981 (9.7) 1658 (49.5) 1663 (51.3) 1645 (50.8) Server anemia 59 (0.1) 183 (5.5) 57 (1.8) 88 (2.7) Grade of kidney disease 0.335 0.017 Stage 1 31 228 (60.5) 1859 (55.5) 1801 (55.6) 1815 (56.0) Stage 2 15 422 (29.9) 781 (23.3) 781 (24.1) 765 (23.6) Stage 3 3062 (5.9) 400 (12) 385 (11.9) 377 (11.6) Stage 4–5 1874 (3.6) 307 (9.2) 272 (8.4) 282 (8.7) RCRI 0.592 0.027 I 43 317 (84) 1957 (58.5) 1871 (57.8) 1912 (59.0) II 6263 (12.1) 955 (28.5) 963 (29.7) 927 (28.6) III 1471 (2.9) 300 (9) 278 (8.6) 276 (8.5) IV 535 (1) 135 (4) 127 (3.9) 124 (3.8) Data are presented as frequency (%). ASA, American Society of Anesthesiologists; PBT, perioperative blood transfusion; RCRI, revised cardiac risk index; SMD, standardized mean difference. Table 2 - Propensity score analyses to assess the association between perioperative blood transfusion and 1-year mortality after surgery. Analysis One-year mortality (%) P No. of events/no. of patients at risk (%) in entire cohort <0.001 No PBT 1440/51 586 (2.8) PBT 494/3347 (14.8) <0.001 No. of events/no. of patients at risk (%) in matched cohort No PBT 313/3239 (9.7) PBT 454/3239 (14.0) Crude analysis-HR (95% CI) 5.71 (5.15–6.32) <0.001 Multivariable analysis-HR (95% CI) 1.69 (1.51–1.90) <0.001 Adjusted for propensity score-HR (95% CI) 2.10 (1.84–2.39) <0.001 Propensity score matched-HR (95% CI) 1.50 (1.30–1.73) <0.001 With IPTW-HR (95% CI) 2.32 (2.02–2.67) <0.001 With SMRW-HR (95% CI) 1.45 (1.33–1.59) <0.001 With PA-HR (95% CI) 1.58 (1.36–1.83) <0.001 With OW-HR (95% CI) 1.65 (1.39–1.97) <0.001 CI, confidence interval; IPTW, inverse probability of treatment weighting; OW, overlap weight; HR, hazard ratio; PA, pairwise algorithmic; PBT, perioperative blood transfusion; SMRW, standardized mortality ratio weighting. This study had several strengths. It is the first large-scale retrospective cohort study to specifically examine the association between perioperative blood transfusion and 1-year mortality in patients undergoing various types of surgery. In addition, the robustness of our findings was enhanced by various multivariable analyses and subgroup analyses. However, this study also had limitations. The exclusion of 42 510 patients due to missing data, particularly among individuals aged >70 years, may have introduced selection bias, limiting the applicability of our findings to older populations. However, sensitive analysis after imputation of missing data was consistent with the results. Besides, despite rigorous statistical adjustments, unmeasured confounders such as perioperative medication regimens, intraoperative blood loss, procedure duration, variations in clinical care, and socioeconomic factors could have influenced the observed association between transfusion and mortality. Finally, the predominance of Chinese individuals in Singapore raises concerns about the applicability of our findings to other ethnic groups and health care settings. In conclusion, this finding primarily applies to hemodynamically stable patients without active massive bleeding. In elective or non-emergent surgical patients, perioperative blood transfusion correlated with increased 1-year mortality, potentially mediated by underlying risk factors. Perioperative blood transfusion is a clinical marker for high-risk surgical patients, and its correlation with mortality may mainly stem from the burden of underlying diseases. However, further research is needed to validate these findings.