Study Population
A total of 911 ECMO treatment episodes were recorded between 1 January, 2010 and 31 December, 2019 in mixed disciplinary ICUs in Hong Kong. After excluding 126 episodes without documented transfusion, 11 episodes with no available pre-transfusion haemoglobin, and 11 episodes with unknown APACHE IV score, a total of 763 episodes were included in the final analysis. By defining the transfusion threshold at 8.5 g/dL, 138 (18.1%) patients were included in the restrictive transfusion strategy group and 625 (81.9%) were included in the liberal transfusion strategy group. Fig.1 shows the study flow.
The baseline characteristics of the two groups are shown in Table 1. The median age was 55 (43-63) years. The median APACHE IV score was 101 (73-133). The median duration of ECMO was 5.9 (3.1-10.8) days. In terms of ECMO configuration, 253 (33.2%) were V-A ECMO, 382 (50.1%) were veno-venous (V-V) ECMO, and 128 (16.8%) were ECPR. A total of 161 (21.1%) and 137 (18.0%) patients were on antiplatelet or anticoagulation medication before admission, respectively. The major diagnoses necessitating V-A ECMO were MI in 58 (23.0%) patients and acute myocarditis in 41 (16.2%) patients. For patients on veno-venous (V-V) ECMO, 162 (42.4%) had bacterial pneumonia and 82 (21.5%) had viral pneumonia. For patients undergoing ECPR, 38 (29.7%) were due to MI and 9 (7.0%) were due to acute myocarditis.
Compared with the liberal strategy group, those who were in the restrictive strategy group had a higher APACHE IV score [109 (80-143) vs 99 (72-130), P=0.003], were less likely male [71 (51%) vs 198 (64%), P=0.008)], and had a shorter duration of ECMO [5.2 (2.1-9.7) vs 6.1 (3.2-10.9) days, P=0.020]. There were no significant differences in age [54 (42-63) vs 55 (44-63) years, P=0.54] or type of ECMO used [43 (31.2%) vs 210 (33.6%) for V-A ECMO, P=0.58; 65 (47.1%) vs 317 (50.7%) for V-V ECMO, P=0.44; 30 (21.7%) vs 98 (15.7%) for ECPR, P=0.09)]. Both groups had similar comorbidities, except the restrictive group which had more patients with malignancy [(18 (13.0%) vs 37 (5.9%), P=0.003]. The primary diagnoses were similar in both groups.
The median transfusion threshold was lower in the restrictive strategy group [7.2 (6.8-7.7) vs 9.4 (8.7-10.2) g/dL, P<0.001]. Patients in the restrictive strategy group had a lower haemoglobin level [8.5 (7.3-11.2) vs 10.4 (8.7-12.3) g/dL, P<0.001], longer prothrombin time (PT) [18.5 (14.9-27.3) vs 16.6 (13.7-22.6) seconds, P=0.006] and INR [1.6 (1.3-2.4) vs 1.5 (1.2-1.9), P=0.004] upon ECMO initiation. Median values of haemoglobin during ECMO were lower in the restrictive strategy group [8.3 (7.7-9.0) vs 9.9 (9.2-10.8) g/dL, P<0.001], while PT [15.7 (13.7-24.6) vs 15.1 (13.3-18.0) seconds, P=0.008], INR [1.4 (1.2-2.2) vs 1.3 (1.2-1.6), P=0.006], and creatinine level [165 (104-247) vs 137 (86-201) μmol/L, P=0.003] were higher, compared with the liberal strategy group. There were no significant differences in activated partial thromboplastin time (aPTT) and platelet count between the two groups. Laboratory results are shown in Table 2.
Blood product usage
The total units of red cell transfusion were lower in the restrictive group than in the liberal group [4 (2-7) vs 8 (4-16) units, P<0.001]. The average pRBC transfused per day was lower in the restrictive group (0.7 vs 1.2 packs per day, P<0.001). Patients in the restrictive group had fewer episodes of red cell transfusion (2 vs 5 times, P<0.001).
Clinical Outcomes
The ICU mortality was 63 (46%) and 256 (41%) in the restrictive group and liberal group, respectively (P=0.31) (Table 3). Secondary outcomes such as hospital mortality were 68 (49%) in the restrictive group and 284 (45%) in the liberal group (P=0.41), and death at 90 days after ECMO initiation were 71 (51%) and 294 (47%), respectively (P=0.35). The unadjusted outcomes were not significantly different between the two groups.
After adjustment for confounders including APACHE IV score, age, type of ECMO performed, baseline haemoglobin, and baseline INR, there was no significant association between a restrictive transfusion strategy and ICU mortality (adjusted odds ratio [OR], 0.86; 95% CI 0.56-1.30; P=0.47). The model had good discriminatory performance and was well-calibrated (area under receiver operating characteristic curve, 0.73; 95% CI 0.69-0.76; Hosmer-Lemeshow test, P=0.97; Additional file: Fig. S1). There were no significant association between restrictive transfusion and hospital mortality (adjusted OR, 0.79; 95% CI 0.52-1.21; P=0.28), and 90-day mortality (adjusted OR, 0.84; 95% CI 0.55-1.28; P=0.42). Unadjusted and adjusted outcomes are shown in Table 3.
Complications
The restrictive strategy group had fewer patients with new-onset ESRF [8 (6%) vs 83 (13%), P=0.014]. There were no significant differences in the number of patients with bacteraemia [7 (5%) vs 52 (8%), P=0.20) and thrombotic complications [10 (7%) vs 53 (8%), P=0.63] from ECMO initiation to hospital discharge. The percentage of patients with secondary MI after discharge was similar [2 (1%) vs 30 (5%), P=0.01]. The results of adjusted analyses were similar (Table 3).
Subgroup analyses
There was significant interaction between the transfusion strategy and the type of ECMO configuration. Restrictive strategy was associated with lower ICU mortality rates for the subgroup of patients who had V-V ECMO (adjusted OR, 0.36; 95% CI 0.17-0.73; P=0.005) (P for interaction “restrictive strategy * V-V ECMO” = 0.007). There were no significant differences in ICU mortality between the restrictive and liberal strategies in the V-A ECMO (adjusted OR, 1.53; 95% CI 0.74-3.15; P=0.25) and ECPR (adjusted OR, 1.59; 95% CI 0.55-4.59; P=0.39) subgroups.
Subgroup analyses were performed for patients with APACHE IV score (≤ 120 vs > 120), and age (≤ 55 years vs > 55 years). There were 57 (41%) patients with APACHE IV score > 120 in the restrictive group and 199 (32%) in the liberal group. There were no significant differences in ICU mortality between the restrictive and liberal groups in the subgroup of patients with an APACHE IV score of ≤120 (adjusted OR, 0.70; 95% CI 0.40-1.23; P=0.21) or APACHE IV score of >120 (adjusted OR, 1.27; 95% CI 0.65-2.48; P=0.48). There were 66 (48%) patients with age > 55 years in the restrictive group and 317 (51%) in the liberal group. There were similarly no effects of transfusion strategy on ICU mortality in the subgroup of age ≤ 55 years (adjusted OR, 0.80; 95% CI 0.42-1.50; P=0.48) and age >55 years (adjusted OR, 0.85; 95% CI 0.47-1.51; P=0.57)
A total of 13 (9%) and 200 (32%) patients had massive blood transfusion in the restrictive group and liberal group, respectively. There were no significant differences in ICU mortality between the two groups for patients with a massive blood transfusion (adjusted OR, 0.49; 95% CI 0.14-1.71; P=0.26) or without a massive blood transfusion (adjusted OR, 1.05; 95% CI 0.65-1.70; P=0.84).
For a subgroup of 58 (23%) patients who required V-A ECMO or ECPR for a diagnosis of MI, no differences in ICU mortality were found between the transfusion strategies (adjusted OR, 0.51; 95% CI 0.14-1.77; P=0.29).
The results of subgroup analyses are shown in Table 4.
Sensitivity analyses
Results remained similar for ICU mortality after separate adjustment of the definition of transfusion threshold to 7.5 g/dL (adjusted OR, 0.75; 95% CI 0.45-1.25; P=0.27) and 8 g/dL (adjusted OR, 0.98; 95% CI 0.51-1.88; P=0.95), respectively.
A further sensitivity analysis was performed for patients on V-V ECMO to explore the possible transfusion threshold. The restrictive group had a lower ICU mortality when the transfusion threshold was 8.0 g/dL (adjusted OR, 0.38; 95% CI 0.16-0.91; P=0.030) and there was no difference in ICU mortality at a transfusion threshold of 7.5 g/dL (adjusted OR, 0.39; 95% CI 0.11-1.39; P=0.15).
After adjusting for confounders specific to renal outcomes including APACHE IV score, age, type of ECMO configuration, pre-ECMO hypertension and diabetes mellitus, restrictive strategy was still associated with lower risks of new-onset ESRF (adjusted OR, 0.39; 95% CI 0.18-0.84; P=0.015).
A total of 264 patients were successfully matched at a 1:1 ratio in propensity score matching. There was no significant association between a restrictive strategy and ICU mortality (adjusted OR, 0.86; 95% CI 0.53-1.39; P=0.54) in the propensity score–matched cohort.
The results of sensitivity analyses are shown in Table 4.