Results

By the end of the study period, 293 patients had been included. Of these, 97 (33.1%) had an autologous blood transfusion after ICS (95% confidence interval [CI] 27.7–38.8). ICS failed in 10 patients (3.4%; 95% CI 1.65–6.2).
The characteristics of the patients are summarized in Table 1.
The 2 groups did not differ significantly on maternal age, body mass index, parity, birth date, history of postpartum haemorrhage, type of anesthesia, or urgent cesarean section. Cesarean sections for placenta abnormality were more frequent in the autotransfusion group (58.8% vs. 45.9%, P = .04), but the opposite was observed for cesarean section for uterine scar(s) (13.4% vs. 27.6%, respectively; P = .007).
The overall rate of PPH was 53.6% (95% CI 47.7–59.4). A higher percentage of patients had severe bleeding (>1,000 mL) in the autotransfusion group (86.6% vs. 10.7%, P < .0001). The hemostasis hysterectomy rate in our series was 10.2% (95% CI 6.7–13.7).
Of the 97 women who received autotransfusion, 47 (48%) did not receive a homologous transfusion (Annex Table).
Of the patients with postoperative bleeding, 76.9% were autotransfused. For patients with uterine fibroids and placenta accreta, autotransfusion was performed in 56.5% and 52.9% of patients, respectively. None of the 8 patients with hemostasis disorder were reinfused.
The results of the univariate analysis of autotransfusion according to the indications for ICS are detailed in Table 2. The univariate analysis showed a significant association between autotransfusion and the indications of placenta abnormality (P = .039) and transfusion difficulty (P = .042). Significantly more patients received ICS and autotransfusion for placenta accreta (P < .001), uterine fibroids (P = .013), and postoperative bleeding (P= .001). By contrast, patients with a rare blood group were significantly less likely to receive an autotransfusion (P< .001).
The results of the multivariate regression analysis are shown in Table 3. This analysis showed a significant independent association between autotransfusion and placenta accreta (P < .001; odds ratio [OR] 3.42, 95% CI 1.8–6.54), uterine fibroids (P = .005; OR 3.74, 95% CI 1.49–9.38), and postoperative bleeding (P< .001; OR 10.15, 95% CI 2.67–38.53). There was an inverse relationship between transfusion difficulties and autotransfusion; that is, the group with transfusion difficulties had a lower risk of autotransfusion (P = .034; OR 0.11, 95% CI 0.015–0.85).
The median volume of blood used in the autotransfusion, expressed as quartile (Q1–Q3), was 270 mL (170–490). In this study, 170 autologous RBC units were transfused with an average volume of 284 mL per RBC unit.
The ROC curve indicated a predictive bleeding threshold for autotransfusion of 936mL, a positive predictive value (PPV) of 77.4% and a negative predictive value (NPV) of 95.5% (Figure 1).
No cases of amniotic fluid embolism and no immediate transfusion complications were observed in our series. Five patients with sickle cell disease received autotransfusion with a median volume of 131 mL (84–148, 95% CI 70–341). No transfusion-related incidents occurred in these patients.