Abstract
Objective
To identify preoperative risk factors and develop predictive models for severe postpartum hemorrhage (PPH) (≥2000 mL) and hysterectomy in patients with placenta accreta spectrum (PAS).
Methods
This retrospective multicenter cohort study included 508 patients diagnosed with PAS. Medical history, timing of ultrasound diagnosis (TUD), specific sonographic signs, and pre-delivery laboratory parameters were analyzed. Multivariate logistic regression was employed to identify independent risk factors for severe PPH and hysterectomy. Model discrimination was evaluated using the area under the receiver operating characteristic curve (AUC).
Results
Severe PPH occurred in 138 (27.2%) patients, and 90 (17.7%) underwent hysterectomy. Independent predictors for severe PPH included TUD in the first or second trimester (odds ratio [OR] 9.45 and 5.97, respectively), presence of placental lacunae (OR 1.75), and prolonged preoperative prothrombin time (PT) (OR 1.35). For hysterectomy, independent risk factors included TUD in the first trimester (OR 8.34), placental lacunae feeder vessels (OR 3.07), bladder wall interruption (OR 2.88), and prolonged preoperative activated partial thromboplastin time (APTT) (OR 1.19). The predictive models showed good performance, with an AUC of 0.744 (95% CI 0.696–0.792) for severe PPH and 0.831 (95% CI 0.788–0.875) for hysterectomy.
Conclusion
Preoperative risk stratification for PAS could be improved by integrating routinely available coagulation profiles and specific sonographic findings. Early ultrasound identification of PAS-related signs and prolonged preoperative PT and APTT may indicate an increased risk of severe complications. Placental lacunae feeder vessels, bladder wall interruption may provide additional prognostic value for hysterectomy and severe PPH.
Keywords: Placenta accreta spectrum, Postpartum hemorrhage, Hysterectomy, Prenatal ultrasound, Coagulation profile
Highlights
-
•
First-trimester PAS detection strongly predicts severe PPH (OR 8.24) and hysterectomy (OR 7.88).
-
•
Placental lacunae feeder vessels increase hysterectomy risk 3-fold; bladder wall interruption also independently predictive.
-
•
Prolonged preoperative PT (severe PPH) and APTT (hysterectomy) are independent coagulopathy markers, but Fib is not.
-
•
Predictive models show AUC 0.752 (severe PPH) and 0.837 (hysterectomy) using routine ultrasound and lab data.
1. Introduction
Placenta accreta spectrum (PAS) remains one of the most formidable challenges in modern obstetrics, characterized by the abnormal adherence or invasion of the placenta into the uterine wall. Driven by the global rise in cesarean section (CS) rates, the increasing incidence of PAS may contribute to elevated rates of massive postpartum hemorrhage (PPH) and hysterectomy [1].
In clinical practice, while the diagnosis of PAS has been largely standardized through advanced imaging techniques [2], accurately identifying patients at greater risk of severe hemorrhage or hysterectomy remains challenging. Currently, the severity of PAS is often classified by pathological subtypes; however, these are predominantly based on intraoperative findings and postoperative pathology [3]. Consequently, they offer limited utility for preoperative risk stratification and fail to improve surgical planning or multidisciplinary resource allocation. Therefore, there is a critical need to utilize accessible preoperative information to predict massive hemorrhage and the necessity for hysterectomy.
Ultrasound is the primary diagnostic modality for PAS [2]. Aryananda et al. [4] reported that intracervical hypervascularity involving more than 50% of the cervix, urinary bladder wall distortion, and parametrial hypervascularity were the most influential sonographic predictors of hysterectomy. Nevertheless, the specific correlation between individual sonographic signs and the quantified risk of blood loss remains poorly defined. Moreover, existing models for predicting hysterectomy have generally been developed in relatively small cohorts and require further multicenter validation. While our team’s previous research established an-ultrasound based scoring system correlated with blood loss exceeding 1500 mL [5], that study did not delineate the predictive weight of specific, individual signs. Furthermore, although prior studies have suggested that pre-delivery markers such as hemoglobin and D-dimer levels are associated with intraoperative blood loss [6,7], such findings are often limited by small sample sizes and single-center designs, which lack the generalizability required for clinical decision-making.
To address these limitations, we conducted a larger multicenter cohort study of patients with PAS in China. By integrating readily available preoperative data, including medical history, specific ultrasound descriptors, and laboratory parameters. This study aims to identify independent risk factors and develop predictive models for severe PPH (≥ 2000 mL) and hysterectomy.
2. Materials and methods
2.1. Study population
This retrospective cohort study included patients diagnosed with PAS aligning with FIGO classification system [3] according to clinical and/or pathological feathers. For patients who underwent hysterectomy and pathological examination, the final diagnosis was determined by that; for those without available pathological specimens, the diagnosis was based on intraoperative findings. The representative intraoperative images (Figure S1) illustrate the uterine serosal appearance in patients with PAS. The exclusion criteria were: (1) gestational age at delivery < 28 weeks; (2) vaginal delivery; (3) multiple gestations; (4) missing pregnancy outcome data. Clinical records in 11 tertiary hospitals between January 2018 and June 2023 were evaluated. The study was approved by the Peking University Third Hospital Medical Science Research Ethics Committee (IRB00006761-M2022740), and the requirement for informed consent was waived due to the retrospective nature of the study.
2.2. Data collection and variables
Clinical, laboratory, and ultrasound data were collected from electronic medical records. Demographic characteristics included age and pre-pregnancy body mass index (BMI). Obstetric history variables comprised gravidity, parity, abortions, history of PAS, hypertensive disorders of pregnancy, and gestational diabetes mellitus. Diagnostic ultrasound signs for PAS were adapted from previous expert consensus [2,8] and our previous publication [9]. The specific sonographic markers were defined as follows: Placental lacunae: large, irregular, intraplacental hypoechoic spaces (lacking a hyperechogenic halo) situated above large feeder vessels, creating a “moth-eaten” appearance and containing turbulent flow (Fig. 1a). Loss of clear zone: the disappearance or irregularity of the normal hypoechoic interface between the uterine wall and the placental basal plate (Fig. 1b). Bladder wall interruption: partial or complete loss, interruption, or irregularity of the hyperechoic line between the uterine serosa and the bladder lumen (Fig. 1c). Uterovesical hypervascularity: a striking density of color Doppler signals within the placental bed of a low-lying placenta or placenta previa, characterized by multidirectional flow and aliasing artifacts reaching the bladder wall (Fig. 1d). Subplacental hypervascularity: extensive color Doppler signals within the placental bed demonstrating multidirectional flow and aliasing artifacts (Fig. 1b). Placental lacunae feeder vessels: high-velocity blood jets from myometrium strike the placental lacunae, generating turbulence upon entry and producing a swirling, rich blood flow signal detectable by ultrasound (Video S1). The timing of ultrasound diagnosis (TUD) was defined as the gestational age at which PAS was first suspected based on the presence of at least two of the aforementioned sonographic signs.
Fig. 1.
Ultrasound features for PAS. (a) Placental lacunae. The yellow arrows indicate placental lacunae within the placenta. (b) Loss of clear zone and subplacental hypervascularity. The green area on the left demonstrates loss of the retroplacental clear zone, while the red area on the right shows prominent subplacental blood flow signals. (c) Bladder wall interruption. The yellow area indicates interruption of the bladder line. (d) Uterovesical hypervascularity. The red area demonstrates abundant bridging vessels between the uterus and bladder.
Supplementary material related to this article can be found online at doi:10.1016/j.eurox.2026.100495.
The following is the Supplementary material related to this article Video S1.
The placental lacunae feeder vessels. The visualization of rapid, turbulent blood flow within placental lacunae on dynamic, real-time ultrasound imaging; characterized by high-velocity swirling that mimics the appearance of boiling water or shifting sands, this sign represents a severe manifestation of placental lacunae
Preoperative laboratory parameters included red blood cell count (RBC), white blood cell count, neutrophils, hemoglobin (HGB), platelets, hematocrit (HCT), prothrombin time (PT), activated partial thromboplastin time (APTT), thrombin time (TT), fibrinogen (Fib), alanine aminotransferase (ALT), aspartate aminotransferase (AST), and total protein (TP). Pregnancy outcomes included placenta previa, hysterectomy, shock, disseminated intravenous coagulation, estimated blood loss (EBL) volume.
The primary outcome was severe PPH, defined as EBL ≥ 2000 mL, and hysterectomy. EBL was assessed using volumetric and gravimetric methods. Two suction canisters were used: one collected blood throughout the procedure, whereas the other was used exclusively for amniotic fluid during fetal delivery. Blood absorbed by gauze pads was estimated from the difference between wet and dry weights, assuming 1 g was equivalent to 1 mL of blood. Final EBL was calculated as the sum of blood collected in the suction canister and the gravimetrically estimated blood volume. Uterine-preserving surgery was the preferred initial approach across all centers in patients without extensive placental invasion or severe bladder or parametrial involvement. Hysterectomy was reserved for cases in which uterine preservation could not be safely achieved because of extensive invasion, inability to reconstruct the uterus, or uncontrolled hemorrhage. Conservative surgery was adapted from the one-step uterine-preserving procedure described by Nieto-Calvache AJ et al. [10] and the Nine-Step Uterine-Preservation Protocol described by Zeng et al. [11], generally involving bladder dissection, uterine incision away from the invaded area, resection of the placenta together with the invaded myometrium, hemostatic control, and uterine reconstruction. The technique was individualized according to the location and extent of placental invasion.
2.3. Statistical analysis
Missing data were handled using multiple imputation by chained equations with the “mice” package in R. Normally distributed continuous variables were presented as mean ± standard deviation (SD) and compared between groups using Student’s t-test. Non-normally distributed variables were presented as median with interquartile range and compared using the Mann-Whitney U test. Categorical variables were presented as frequencies and percentages and compared using the chi-square test or Fisher's exact test. Univariate logistic regression analysis was first performed to identify potential risk factors for EBL ≥ 2000 mL and hysterectomy. Variables with a P-value < 0.1 in the univariate analysis were considered candidates for the multivariate model. Multivariate logistic regression analysis with backward stepwise selection was then conducted to identify independent risk factors for EBL ≥ 2000 mL and hysterectomy. Results were presented as odds ratios (OR) with 95% confidence intervals (CI). For the multivariate model, predicted probabilities were generated and used to construct the receiver operating characteristic (ROC) curve. The area under the ROC curve (AUC) with 95% CI was calculated. All statistical analyses were performed using R software (version 4.4.3; R Foundation for Statistical Computing, Vienna, Austria). A two-tailed P-value < 0.05 was considered statistically significant.
3. Results
3.1. Baseline characteristics and univariate analysis
A total of 508 patients were included in this multicenter cohort study. Table 1 summarizes the baseline characteristics according to blood loss volume. Patients with severe PPH (EBL ≥ 2000 mL) had significantly higher gravidity, parity, more times of CS and earlier TUD compared to those with EBL < 2000 mL blood loss (P < 0.05). Specifically, the presence of placental lacunae was more frequent in the severe PPH group (35.5% vs. 24.3%, P = 0.016). Preoperative laboratory findings showed that patients with severe PPH had significantly lower RBC, HGB, HCT, Fib, and AST while PT, APTT, TT were significantly prolonged (P < 0.05). Hysterectomy was significantly associated with earlier TUD, and a higher prevalence of “placental lacunae feeder vessels” and “loss of clear zone” on ultrasound (P < 0.05) (Table 2). Patients underwent hysterectomy had significantly lower HGB, HCT, TT, Fib, ALT and AST while APTT were significantly prolonged (P < 0.05). Median EBL was significantly higher in the hysterectomy group (1950 mL vs. 1036 mL, P < 0.001). Patients with EBL ≥ 2000 mL and those who underwent hysterectomy had more severe PAS according to the FIGO classification, with a higher proportion of placenta percreta and a lower proportion of placenta accreta (P < 0.001).
Table 1.
Demographic and clinical characteristics of patients with placenta accreta spectrum grouped by blood loss volume.
| Variable | < 2000 mL (n = 370) | ≥ 2000 mL (n = 138) | P value |
|---|---|---|---|
| Age | 33.35 ± 4.71 | 34.03 ± 4.14 | 0.112 |
| Pre-pregnancy BMI | 66.00 (61.00–73.00) | 67.00 (61.00–72.00) | 0.868 |
| Gravidity | 0.012 | ||
| 0 | 0 (0.0%) | 0 (0.0%) | |
| 1 | 26 (7.0%) | 2 (1.4%) | |
| 2 | 66 (17.8%) | 19 (13.8%) | |
| ≥ 3 | 278 (75.1%) | 117 (84.8%) | |
| Parity | < 0.001 | ||
| 0 | 0 (0.0%) | 0 (0.0%) | |
| 1 | 71 (19.2%) | 12 (8.7%) | |
| 2 | 223 (60.3%) | 69 (50.0%) | |
| ≥ 3 | 76 (20.5%) | 57 (41.3%) | |
| History of CS | < 0.001 | ||
| 0 | 100 (27.0%) | 12 (8.7%) | |
| 1 | 205 (55.4%) | 76 (55.1%) | |
| 2 | 59 (15.9%) | 46 (33.3%) | |
| 3 | 6 (1.6%) | 4 (2.9%) | |
| History of abortions | 0.391 | ||
| 0 | 96 (25.9%) | 33 (23.9%) | |
| 1 | 106 (28.6%) | 46 (33.3%) | |
| 2 | 92 (24.9%) | 26 (18.8%) | |
| ≥ 3 | 76 (20.5%) | 33 (23.9%) | |
| TUD | < 0.001 | ||
| Not suspected | 39 (10.5%) | 3 (2.2%) | |
| First trimester | 11 (3.0%) | 11 (8.0%) | |
| Second trimester | 127 (34.3%) | 58 (42.0%) | |
| Third trimester | 193 (52.2%) | 66 (47.8%) | |
| Placental lacunae | 90 (24.3%) | 49 (35.5%) | 0.016 |
| Loss of clear zone | 295 (79.7%) | 112 (81.2%) | 0.815 |
| Bladder wall interruption | 49 (13.2%) | 21 (15.2%) | 0.668 |
| Uterovesical hypervascularity | 25 (6.8%) | 19 (13.8%) | 0.020 |
| Subplacental hypervascularity | 64 (17.3%) | 17 (12.3%) | 0.220 |
| Placental lacunae feeder vessels | 29 (7.8%) | 7 (5.1%) | 0.376 |
| RBC | 3.70 (3.40–4.00) | 3.60 (3.30–3.80) | 0.030 |
| WBC | 8.75 (7.10–10.47) | 8.50 (7.23–10.28) | 0.982 |
| Neu | 75.00 (71.00–79.00) | 74.00 (72.00–79.00) | 0.454 |
| HGB | 113.00 (103.00–121.00) | 108.00 (98.00–118.00) | 0.015 |
| PLT | 193.00 (157.00–235.00) | 188.50 (150.00–225.75) | 0.236 |
| HCT | 34.00 (31.25–36.00) | 32.00 (30.00–35.00) | 0.012 |
| PT | 10.90 (10.40–11.60) | 11.25 (10.80–12.10) | < 0.001 |
| APTT | 26.90 (25.00–28.80) | 27.50 (24.90–30.78) | 0.040 |
| TT | 15.05 (13.70–15.80) | 15.15 (14.20–16.08) | 0.038 |
| Fib | 4.20 (3.60–4.80) | 4.00 (3.50–4.50) | 0.004 |
| ALT | 13.00 (8.00–19.00) | 11.50 (7.00–17.00) | 0.116 |
| AST | 17.00 (14.00–21.00) | 15.00 (12.00–18.00) | < 0.001 |
| TP | 63.00 (59.00–66.00) | 62.00 (58.00–65.00) | 0.094 |
| PAS | < 0.001 | ||
| Accreta | 85 (23.0%) | 5 (3.6%) | |
| Increta | 234 (63.2%) | 88 (63.8%) | |
| Percreta | 51 (13.8%) | 45 (32.6%) | |
| Hysterectomy | 45 (12.2%) | 45 (32.6%) | < 0.001 |
| Shock | 2 (0.5%) | 12 (8.7%) | < 0.001 |
| DIC | 2 (0.5%) | 5 (3.6%) | 0.018 |
| PP | 0.051 | ||
| Normal | 40 (10.8%) | 14 (10.1%) | |
| Low | 11 (3.0%) | 0 (0.0%) | |
| Partial | 31 (8.4%) | 6 (4.3%) | |
| Complete | 288 (77.8%) | 118 (85.5%) |
BMI, body mass index; CS, cesarean section; TUD, timing of ultrasound diagnosis; RBC, red blood cell count; WBC, white blood cell count; Neu, neutrophils; HGB, hemoglobin; PLT, platelets; HCT, hematocrit; PT, prothrombin time; APTT, activated partial thromboplastin time; TT, thrombin time; Fib, fibrinogen; ALT, alanine aminotransferase; AST, aspartate aminotransferase; TP, total protein; PAS, placenta accreta spectrum; DIC, disseminated intravascular coagulation; PP, placenta previa.
Table 2.
Demographic and clinical characteristics of patients with placenta accreta spectrum grouped by hysterectomy.
| Variable | No hysterectomy (n = 418) | Hysterectomy (n = 90) | P value |
|---|---|---|---|
| Age | 33.61 ± 4.48 | 33.14 ± 4.95 | 0.408 |
| Pre-pregnancy BMI | 66.25 (61.00–73.00) | 66.00 (62.00–72.88) | 0.808 |
| Gravidity | 0.013 | ||
| 0 | 0 (0.0%) | 0 (0.0%) | |
| 1 | 28 (6.7%) | 0 (0.0%) | |
| 2 | 71 (17.0%) | 14 (15.6%) | |
| ≥ 3 | 319 (76.3%) | 76 (84.4%) | |
| Parity | < 0.001 | ||
| 0 | 0 (0.0%) | 0 (0.0%) | |
| 1 | 81 (19.4%) | 2 (2.2%) | |
| 2 | 245 (58.6%) | 47 (52.2%) | |
| ≥ 3 | 92 (22.0%) | 41 (45.6%) | |
| History of CS | < 0.001 | ||
| 0 | 107 (25.6%) | 5 (5.6%) | |
| 1 | 233 (55.7%) | 48 (53.3%) | |
| 2 | 71 (17.0%) | 34 (37.8%) | |
| 3 | 7 (1.7%) | 3 (3.3%) | |
| History of abortions | 0.849 | ||
| 0 | 103 (24.6%) | 26 (28.9%) | |
| 1 | 127 (30.4%) | 25 (27.8%) | |
| 2 | 97 (23.2%) | 21 (23.3%) | |
| ≥ 3 | 91 (21.8%) | 18 (20.0%) | |
| TUD | < 0.001 | ||
| Not suspected | 37 (8.9%) | 5 (5.6%) | |
| First trimester | 11 (2.6%) | 11 (12.2%) | |
| Second trimester | 136 (32.5%) | 49 (54.4%) | |
| Third trimester | 234 (56.0%) | 25 (27.8%) | |
| Placental lacunae | 122 (29.2%) | 17 (18.9%) | 0.063 |
| Loss of clear zone | 343 (82.1%) | 64 (71.1%) | 0.027 |
| Bladder wall interruption | 52 (12.4%) | 18 (20.0%) | 0.086 |
| Uterovesical hypervascularity | 32 (7.7%) | 12 (13.3%) | 0.126 |
| Subplacental hypervascularity | 62 (14.8%) | 19 (21.1%) | 0.188 |
| Placental lacunae feeder vessels | 22 (5.3%) | 14 (15.6%) | 0.001 |
| RBC | 3.70 (3.30–3.98) | 3.60 (3.30–3.80) | 0.369 |
| WBC | 8.70 (7.12–10.30) | 8.60 (7.15–11.20) | 0.516 |
| Neu | 75.00 (71.00–79.00) | 75.00 (71.00–80.00) | 0.279 |
| HGB | 112.00 (102.25–122.00) | 108.50 (99.00–117.00) | 0.012 |
| PLT | 193.00 (157.00–235.00) | 185.50 (154.25–218.75) | 0.226 |
| HCT | 34.00 (31.00–36.00) | 32.00 (30.00–35.00) | 0.012 |
| PT | 11.00 (10.50–11.80) | 11.00 (10.70–11.40) | 0.338 |
| APTT | 26.80 (24.80–28.80) | 28.10 (26.13–30.80) | < 0.001 |
| TT | 15.20 (14.00–15.80) | 14.60 (13.20–15.60) | 0.014 |
| Fib | 4.25 (3.60–4.80) | 3.90 (3.32–4.47) | < 0.001 |
| ALT | 13.00 (8.00–19.00) | 9.50 (6.00–14.00) | < 0.001 |
| AST | 17.00 (14.00–21.00) | 15.00 (12.25–18.00) | 0.005 |
| TP | 63.00 (59.00–66.00) | 62.50 (58.00–66.00) | 0.339 |
| PAS | < 0.001 | ||
| Accreta | 87 (20.8%) | 3 (3.3%) | |
| Increta | 271 (64.8%) | 51 (56.7%) | |
| Percreta | 60 (14.4%) | 36 (40.0%) | |
| EBL | 1036.00 (650.00–1800.00) | 1950.00 (1203.25–3024.00) | < 0.001 |
| Shock | 5 (1.2%) | 9 (10.0%) | < 0.001 |
| DIC | 3 (0.7%) | 4 (4.4%) | 0.021 |
| PP | 0.089 | ||
| Normal | 49 (11.7%) | 5 (5.6%) | |
| Low | 11 (2.6%) | 0 (0.0%) | |
| Partial | 32 (7.7%) | 5 (5.6%) | |
| Complete | 326 (78.0%) | 80 (88.9%) |
BMI, body mass index; CS, cesarean section; TUD, timing of ultrasound diagnosis; RBC, red blood cell count; WBC, white blood cell count; Neu, neutrophils; HGB, hemoglobin; PLT, platelets; HCT, hematocrit; PT, prothrombin time; APTT, activated partial thromboplastin time; TT, thrombin time; Fib, fibrinogen; ALT, alanine aminotransferase; AST, aspartate aminotransferase; TP, total protein; PAS, placenta accreta spectrum; EBL, estimated blood loss; DIC, disseminated intravascular coagulation; PP, placenta previa.
3.2. Risk factors for severe PPH and hysterectomy
Table 3 demonstrated that earlier TUD was strongly associated with severe PPH, both in the first trimester (OR 8.24, 95% CI: 1.89–45.43, P = 0.008) and second trimester (OR 5.44, 95% CI: 1.73–24.19, P = 0.009). In addition, the presence of placental lacunae (OR 1.73, 95% CI: 1.07–2.78, P = 0.024) and prolonged PT (OR 1.37, 95% CI: 1.06–1.78, P = 0.016) were identified as independent predictors for severe PPH.
Table 3.
Univariate and multivariate logistic regression analysis for risk factors of blood loss ≥ 2000 mL.
|
Univariate analysis |
Multivariate analysis |
|||||
|---|---|---|---|---|---|---|
| Variable | OR | 95% CI | P value | OR | 95% CI | P value |
| Gravidity | ||||||
| 2 | 3.74 | 0.99–24.53 | 0.090 | 1.76 | 0.36–13.17 | 0.519 |
| ≥ 3 | 5.47 | 1.60–34.31 | 0.022 | 1.85 | 0.42–13.14 | 0.464 |
| Parity | ||||||
| 2 | 1.83 | 0.97–3.73 | 0.076 | 0.73 | 0.28–1.98 | 0.525 |
| ≥ 3 | 4.44 | 2.26–9.30 | < 0.001 | 1.90 | 0.57–6.41 | 0.295 |
| Previous CS | ||||||
| 1 | 3.09 | 1.66–6.21 | < 0.001 | 2.93 | 1.27–7.52 | 0.017 |
| 2 | 6.50 | 3.28–13.75 | < 0.001 | 2.82 | 0.94–9.20 | 0.073 |
| 3 | 5.56 | 1.27–22.44 | 0.016 | 1.80 | 0.31–10.18 | 0.507 |
| TUD | ||||||
| First trimester | 13.00 | 3.40–65.50 | < 0.001 | 8.24 | 1.89–45.43 | 0.008 |
| Second trimester | 5.94 | 2.04–25.25 | 0.004 | 5.44 | 1.73–24.19 | 0.009 |
| Third trimester | 4.45 | 1.54–18.81 | 0.015 | 5.01 | 1.62–22.17 | 0.013 |
| Placental lacunae | 1.71 | 1.12–2.61 | 0.012 | 1.73 | 1.07–2.78 | 0.024 |
| Uterovesical hypervascularity | 2.20 | 1.16–4.13 | 0.014 | 1.85 | 0.89–3.78 | 0.095 |
| RBC | 0.71 | 0.47–1.04 | 0.087 | 0.91 | 0.56–1.44 | 0.707 |
| HGB | 0.99 | 0.97–1.00 | 0.069 | 1.01 | 0.99–1.03 | 0.315 |
| PT | 1.58 | 1.27–1.96 | < 0.001 | 1.37 | 1.06–1.78 | 0.016 |
| APTT | 1.08 | 1.03–1.14 | 0.003 | 1.05 | 0.99–1.12 | 0.098 |
| TT | 1.12 | 1.02–1.27 | 0.056 | 1.05 | 0.96–1.19 | 0.431 |
| Fib | 0.83 | 0.70–0.97 | 0.028 | 0.91 | 0.76–1.09 | 0.314 |
| TP | 0.97 | 0.95–0.99 | 0.008 | 0.98 | 0.95–1.01 | 0.157 |
OR, odds ratio; CI, confidence interval; CS, cesarean section; TUD, timing of ultrasound diagnosis; RBC, red blood cell count; HGB, hemoglobin; PT, prothrombin time; APTT, activated partial thromboplastin time; TT, thrombin time; Fib, fibrinogen; TP, total protein.
In Table 4, the multivariate model revealed the TUD in the first trimester was also independently associated with a markedly increased likelihood of hysterectomy (OR 7.88, 95% CI: 1.93–35.75, P = 0.005). Moreover, bladder wall interruption (OR 2.59, 95% CI: 1.12–5.94, P = 0.024), placental lacunae feeder vessels (OR 3.73, 95% CI: 1.33–10.47, P = 0.012), and prolonged APTT (OR 1.19, 95% CI: 1.10–1.29, P < 0.001) were independently associated with increased hysterectomy risk. In contrast, placental lacunae showed an inverse association with hysterectomy (OR 0.44, 95% CI: 0.21–0.87, P = 0.023).
Table 4.
Univariate and multivariate logistic regression analysis for risk factors of hysterectomy.
|
Univariate analysis |
Multivariate analysis |
|||||
|---|---|---|---|---|---|---|
| Variable | OR | 95% CI | P value | OR | 95% CI | P value |
| Parity | ||||||
| 2 | 7.77 | 2.33–48.22 | 0.005 | 6.22 | 1.32–46.35 | 0.036 |
| ≥ 3 | 18.05 | 5.32–112.93 | < 0.001 | 12.43 | 2.05–107.68 | 0.010 |
| Previous CS | ||||||
| 1 | 4.41 | 1.87–12.97 | 0.002 | 2.21 | 0.73–8.19 | 0.195 |
| 2 | 10.25 | 4.15–31.01 | < 0.001 | 3.40 | 0.83–16.28 | 0.103 |
| 3 | 9.17 | 1.63–46.34 | 0.007 | 3.38 | 0.40–27.37 | 0.250 |
| TUD | ||||||
| First trimester | 7.40 | 2.21–28.08 | 0.002 | 7.88 | 1.93–35.75 | 0.005 |
| Second trimester | 2.67 | 1.07–8.09 | 0.052 | 2.12 | 0.74–7.20 | 0.189 |
| Third trimester | 0.79 | 0.31–2.46 | 0.652 | 0.70 | 0.24–2.41 | 0.541 |
| Placental lacunae | 0.57 | 0.31–0.98 | 0.049 | 0.44 | 0.21–0.87 | 0.023 |
| Loss of clear zone | 0.54 | 0.32–0.92 | 0.020 | 0.53 | 0.26–1.07 | 0.072 |
| Bladder wall interruption | 1.76 | 0.95–3.14 | 0.062 | 2.59 | 1.12–5.94 | 0.024 |
| Uterovesical hypervascularity | 1.86 | 0.88–3.68 | 0.086 | 1.28 | 0.51–3.13 | 0.589 |
| Placental lacunae feeder vessels | 3.32 | 1.59–6.71 | < 0.001 | 3.73 | 1.33–10.47 | 0.012 |
| WBC | 1.07 | 0.99–1.15 | 0.067 | 1.06 | 0.98–1.16 | 0.154 |
| HGB | 0.98 | 0.96–0.99 | 0.006 | 0.98 | 0.95–1.01 | 0.336 |
| HCT | 0.94 | 0.90–0.99 | 0.015 | 1.03 | 0.94–1.15 | 0.560 |
| APTT | 1.14 | 1.08–1.22 | < 0.001 | 1.19 | 1.10–1.29 | < 0.001 |
| Fib | 0.71 | 0.56–0.87 | 0.002 | 0.97 | 0.76–1.22 | 0.824 |
OR, odds ratio; CI, confidence interval; CS, cesarean section; TUD, timing of ultrasound diagnosis; WBC, white blood cell count; HGB, hemoglobin; HCT, hematocrit; APTT, activated partial thromboplastin time; Fib, fibrinogen.
3.3. Predictive performance
The model of severe PPH achieved an AUC of 0.752 (95% CI: 0.705–0.799, Fig. 2) and the model of hysterectomy showed higher predictive accuracy with an AUC of 0.837 (95% CI: 0.794–0.880, Fig. 3).
Fig. 2.
Receiver operating characteristic (ROC) curve of the multivariate logistic regression model for predicting severe postpartum hemorrhage (blood loss ≥ 2000 mL).
Fig. 3.
Receiver operating characteristic (ROC) curve of the multivariate logistic regression model for predicting hysterectomy.
4. Discussion
4.1. Main findings
In this large multicenter cohort of patients with PAS, preoperative ultrasound and routinely available laboratory variables enabled risk stratification for severe PPH (AUC 0.752) and hysterectomy (AUC 0.837). Earlier TUD, placental lacunae, and prolonged PT were independently associated with severe PPH, whereas first-trimester TUD, placental lacunae feeder vessels, bladder-wall interruption, and prolonged APTT were independently associated with hysterectomy.
4.2. Interpretation
Our results underscore that specific morphological markers on ultrasound carry distinct predictive weights for severe PPH and hysterectomy. Earlier TUD was strongly associated with adverse outcomes. PAS-related ultrasound signs detected during the first or second trimester were associated with a markedly increased risk of severe PPH, while first-trimester detection was also associated with hysterectomy. Taşkum et al. [12] similarly reported a trend towards a lower risk of hysterectomy with increasing gestational age at diagnosis (aOR 0.78, 95% CI 0.56–1.02, P = 0.07). This suggests that an earlier PAS sonographic manifestation reflects a more aggressive biological phenotype of placental invasion.
Regarding specific signs, the “placental lacunae feeder vessels” was a potent predictor in our cohort, increasing the odds of hysterectomy by 3.73 times. Hussein et al.’s results also indicated that lacunar feeder vessels increase the risk of hysterectomy (OR 5.69, 95% CI 1.77–18.3, p = 0.004) and blood transfusion > 5 RBC units (OR 1.62, 95% CI 1.24–2.11, p = 0.001) in individuals with a history of CS and low-lying placenta/placenta previa [13]. The predictive value of bladder wall interruption for hysterectomy suggests the difficulty of detaching the low segment uterine with the bladder. This is consistent with the findings of Hussein et al. [14], who identified “numerous coherent vessels involving the bladder-serosal interface” as an independent predictor of major hemorrhage during hysterectomy. Similarly, Aryananda et al. [4] reported that intracervical hypervascularity > 50% and bladder wall distortion had a high predictive probability (0.87) for hysterectomy. While Taşkum et al. [12] reported a strong association between “multiple lacunae” and hysterectomy (aOR 48.53), our multivariate analysis showed a more modest association with severe PPH (OR 1.75) and an inverse association with hysterectomy (OR 0.45). The discrepancy may partly reflect differences in definition, as Taşkum et al. [12] assessed multiple lacunae, whereas our multicenter study used a binary presence/absence variable. As more abundant lacunae likely indicate larger blood sinuses and more severe invasion, this simplified definition may have attenuated the observed effect. These findings highlight the need for standardized and quantitative ultrasound assessment to improve preoperative risk stratification in PAS.
Beyond imaging, the maternal systemic state proved to be a modest predictor of adverse outcomes. Our multivariate analysis identified prolonged PT (OR 1.37) and APTT (OR 1.19) as risk factors associated with severe PPH and hysterectomy, respectively. A prolonged APTT may reflects a subclinical consumptive coagulopathy or reduced localized hemostatic capacity. This makes intraoperative hemostasis more challenging, potentially leading clinicians toward a definitive hysterectomy to achieve life-saving control. In contrast, PT served as more robust markers specifically for severe blood loss ≥ 2000 mL. This finding aligns with the results of Gong et al. [15] in a CS population, and we provide evidence specifically in the PAS cohort. Consistent with our previous results [16], the multivariate analysis in this study also found no significant association between preoperative Fib levels and either hysterectomy or EBL ≥ 2000 mL in PAS patients. Although local hyperfibrinolysis has been proposed in PAS [17], circulating fibrinogen concentrations may remain within the normal range before substantial bleeding occurs. Balkaş et al. [18] observed that patients with severe PAS subtypes (increta and percreta) had significantly prolonged PT, which predicted bleeding > 1000 mL. Similarly, our study demonstrated that prolonged PT independently predicted EBL ≥ 2000 mL, reinforcing the consistency of PT as a predictive marker across different severity thresholds. The underlying mechanisms of these coagulation abnormalities remain unclear. These findings suggest that abnormal trophoblast invasion in PAS may lead to the depletion of coagulation factors, resulting in hypocoagulability.
The clinical management of PAS is currently at a crossroads between retrospective diagnosis and prospective risk stratification. Historically, research has predominantly focused on correlating antenatal imaging findings with the depth of myometrial invasion—accreta, increta, or percreta—as determined post-delivery [11]. While these efforts have refined our anatomical understanding, their retrospective nature offers limited utility for risk stratification. It guides the timely mobilization of multidisciplinary teams, activation of massive transfusion protocols, and elective referral to specialized tertiary centers. Although predictive model of MRI has demonstrated satisfactory performance for adverse outcomes [19], its reliance on complex machine learning algorithms and high associated costs limits its widespread applicability. Therefore, extracting predictive information from routinely available medical history, preoperative laboratory tests, and cost-effective ultrasound may represent an efficient approach for prediction. Previous ultrasound scoring systems has been developed to predicting severe PPH and hysterectomy in high-risk populations [20,21]. However, because PAS is uncommon, these studies often included fewer than 100 confirmed cases, and their predictive performance may largely reflect the distinction between PAS and non-PAS pregnancies. Our study included only patients with confirmed PAS and therefore addressed the more clinically challenging question of risk stratification within an established PAS population.
Artificial intelligence has also been applied to predict severe hemorrhage and hysterectomy in patients with PAS. However, many reported models rely heavily on MRI-derived features [22], whereas AI models based on more accessible and lower-cost ultrasound findings and routine preoperative laboratory tests remain limited. Developing and externally validating such models represents an important direction for future research.
4.3. Strengths and limitations
The strengths of this study include its multicenter design, relatively large sample size, and use of readily available preoperative variables, including medical history, ultrasound signs, and routine laboratory tests, thereby supporting broad clinical applicability. To our knowledge, this is among the first multicenter studies to integrate ultrasound and laboratory parameters for predicting severe hemorrhage and hysterectomy in PAS. Several limitations should be acknowledged. The retrospective design may have introduced selection bias and intercenter variability in ultrasound assessment and surgical management. Some ultrasound signs may overlap with uterine scar dehiscence, and the small number of cases with first-trimester detection may have reduced the precision of subgroup estimates. Finally, although the models showed acceptable discrimination, prospective external validation is required.
5. Conclusions
Our findings underscore that an earlier gestational age at the sonographic diagnosis (particularly in the first or second trimester), the presence of placental lacunae, placental lacunae feeder vessels and bladder wall interruption are critical warning indicators. When combined with prolonged preoperative PT or APTT, these markers significantly enhance the identification of high-risk patients likely to experience severe PPH (≥ 2000 mL) or require hysterectomy. Such preoperative stratification is vital for optimizing surgical planning and improving maternal safety.
Author contributions
SZ, YZ and LC contributed to the protocol development, data analysis and manuscript writing/editing. DC, AZ, YC, LD, and WZ contributed to the protocol development and data collection. All authors approved the final version of the manuscript and agree to be accountable for all aspects of the work.
CRediT authorship contribution statement
Shuai Zeng: Writing – review & editing, Writing – original draft, Formal analysis, Data curation, Conceptualization. Dunjin Chen: Writing – review & editing, Methodology, Data curation. Aiqing Zhang: Writing – review & editing, Methodology, Data curation. Yiwen Chong: Writing – review & editing, Methodology, Data curation. Lili Du: Writing – review & editing, Methodology, Data curation. Wanjiang Zeng: Writing – review & editing, Methodology, Formal analysis, Data curation. Yangyu Zhao: Writing – review & editing, Writing – original draft, Supervision, Project administration, Funding acquisition, Conceptualization. Lian Chen: Writing – review & editing, Writing – original draft, Supervision, Project administration, Methodology, Investigation, Funding acquisition, Formal analysis, Data curation, Conceptualization.
Funding
This work was supported by the National Key Research and Development Program of China (2022YFC2704503); the Capital’s Funds for Health Improvement and Research (2024–1–4094); Beijing Natural Science Foundation (7252160) and the special fund of the National Clinical Key Specialty Construction Program, P. R. China (2024).
Declaration of Generative AI and AI-assisted technologies in the manuscript preparation process
DeepSeek was used solely for language and grammatical editing. The authors take full responsibility for all scientific content, data analyses, interpretations, and conclusions, which were independently verified for accuracy.
Declaration of Competing Interest
The authors have no conflicts of interest.
Footnotes
Supplementary data associated with this article can be found in the online version at doi:10.1016/j.eurox.2026.100495.
Contributor Information
Yangyu Zhao, Email: zhaoyangyu@bjmu.edu.cn.
Lian Chen, Email: bysychenlian@126.com.
Appendix A. Supplementary material
Supplementary material Figure S1. Representative intraoperative uterine serosal appearances in patients with placenta accreta spectrum, based on the 2019 FIGO clinical classification. (a) Placenta accreta: no obvious placental bulge or distension over the placental bed, no placental tissue visible through the uterine serosa, and absent or minimal neovascularisation. (b) Placenta increta: abnormal macroscopic changes over the placental bed, including bluish-purple discolouration, a prominent placental bulge, and marked hypervascularity. No placental tissue is visible invading through the serosal surface. (c) Placenta percreta: marked abnormality of the uterine serosal surface, with placental tissue seen to be invading through the uterine wall or extending beyond the serosa
Data availability statement
The datasets generated and analyzed during the current study are available from the corresponding author upon reasonable request.
References
- 1.Peng X., Tan X., Wu Z. Prenatal ultrasound scoring in diagnosis and postpartum outcomes prediction for Placenta Accreta Spectrum (PAS): a systematic review. BMC Pregnancy Childbirth. 2024;24(1):846. doi: 10.1186/s12884-024-07076-9. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 2.Jauniaux E., D’Antonio F., Bhide A., Prefumo F., Silver R.M., Hussein A.M., et al. Modified Delphi study of ultrasound signs associated with placenta accreta spectrum. Ultrasound Obstet Gynecol. 2023;61(4):518–525. doi: 10.1002/uog.26155. [DOI] [PubMed] [Google Scholar]
- 3.Jauniaux E., Ayres-de-Campos D., Langhoff-Roos J., Fox K.A., Collins S. FIGO Placenta accreta diagnosis and management expert consensus panel. FIGO classification for the clinical diagnosis of placenta accreta spectrum disorders. Int J Gynecol & Obste. 2019;146(1):20–24. doi: 10.1002/ijgo.12761. [DOI] [PubMed] [Google Scholar]
- 4.Aryananda R.A., Adu-Bredu T.K., Cininta N.I., Twumasi C., Pranpanus S., Coutinho C.M., et al. Diagnostic ultrasound to inform the surgical approach to cesarean delivery in patients at high risk for placenta accreta spectrum disorders. Am J Obstet Gynecol. 2025;233(6):658.e1–658.e16. doi: 10.1016/j.ajog.2025.08.005. [DOI] [PubMed] [Google Scholar]
- 5.Chen L., Shi H.F., Jiang H., Shi X.M., Wang Y.Y., Zhang A.Q., et al. Correlation of an ultrasonic scoring system and intraoperative blood loss in placenta accreta spectrum disorders: a retrospective cohort study. Biomed Environ Sci. 2021;34(2):163–169. doi: 10.3967/bes2021.022. [DOI] [PubMed] [Google Scholar]
- 6.Zhang B., Liu H., Wang W., Wang J., Jiang Y., Jia W., et al. Impact of maternal preoperative hemoglobin levels on intraoperative hemorrhage risk in placenta accrete spectrum disorders: a retrospective cross-sectional study. Med Sci Monit. 2023;29 doi: 10.12659/MSM.940443. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 7.Wu X., He J., Bai Y., Gan Y., Xu H., Qi H., et al. Risk factors for severe postpartum hemorrhage in placenta accreta spectrum patients undergoing prophylactic resuscitative endovascular balloon occlusion of the aorta during cesarean delivery. Ann Med. 2025;57(1) doi: 10.1080/07853890.2024.2442065. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 8.Collins S.L., Ashcroft A., Braun T., Calda P., Langhoff-Roos J., Morel O., et al. Proposal for standardized ultrasound descriptors of abnormally invasive placenta (AIP) Ultrasound Obstet Gynecol. 2016;47(3):271–275. doi: 10.1002/uog.14952. [DOI] [PubMed] [Google Scholar]
- 9.Chong Y., Zhang A., Wang Y., Chen Y., Zhao Y. An ultrasonic scoring system to predict the prognosis of placenta accreta: a prospective cohort study. Med (Baltim) 2018;97(35) doi: 10.1097/MD.0000000000012111. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 10.Nieto-Calvache A.J., Palacios-Jaraquemada J.M., Aryananda R., Basanta N., Aguilera R., Benavides J.P., et al. How to perform the one-step conservative surgery for placenta accreta spectrum move by move. Am J Obstet Gynecol MFM. 2023;5(2) doi: 10.1016/j.ajogmf.2022.100802. [DOI] [PubMed] [Google Scholar]
- 11.Zeng S., Qu J., Jiang H., Shi H., Yan J., Zhao Y., et al. Decreased plasma APOA1 levels are associated with increased severity of placenta accreta spectrum disorders: a nested case-control study. Front Endocrinol. 2025;16 doi: 10.3389/fendo.2025.1627377. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 12.Taşkum, İ, Çetin F., Sucu S., Bademkıran M.H., Kömürcü Karuserci Ö., Bademkıran C., et al. Predicting the risk of cesarean hysterectomy in the management of placenta accreta spectrum disorders: a new model based on clinical findings and ultrasonography. Arch Gynecol Obstet. 2025;311(1):55–66. doi: 10.1007/s00404-024-07858-8. [DOI] [PubMed] [Google Scholar]
- 13.Hussein A.M., Fox K., Bhide A., Elbarmelgy R.A., Elbarmelgy R.M., Thabet M.M., et al. The impact of preoperative ultrasound and intraoperative findings on surgical outcomes in patients at high risk of placenta accreta spectrum. BJOG. 2023;130(1):42–50. doi: 10.1111/1471-0528.17286. [DOI] [PubMed] [Google Scholar]
- 14.Hussein A.M., Momtaz M., Elsheikhah A., Abdelbar A., Kamel A. The role of ultrasound in prediction of intra-operative blood loss in cases of placenta accreta spectrum disorders. Arch Gynecol Obstet. 2020;302(5):1143–1150. doi: 10.1007/s00404-020-05707-y. [DOI] [PubMed] [Google Scholar]
- 15.Gong J., Chen Z., Zhang Y., Liu Y.-Y., Pu J.-C., Xiong C.-Y., et al. Risk-factor model for postpartum hemorrhage after cesarean delivery: a retrospective study based on 3498 patients. Sci Rep. 2022;12(1) doi: 10.1038/s41598-022-23636-5. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 16.Gong C., Jiang H., Su Y., Yang J., Wei Y., Qiao R., et al. Changes in fibrinogen and D-dimer during complicated and uncomplicated pregnancy. Int J Lab Hematol. 2025;47(1):156–165. doi: 10.1111/ijlh.14382. [DOI] [PubMed] [Google Scholar]
- 17.Jauniaux E., Hussein A.M., Elbarmelgy R.M., Elbarmelgy R.A., Burton G.J. Failure of placental detachment in accreta placentation is associated with excessive fibrinoid deposition at the utero-placental interface. Am J Obstet Gynecol. 2022;226(2):243.e1–243.e10. doi: 10.1016/j.ajog.2021.08.026. [DOI] [PubMed] [Google Scholar]
- 18.Balkaş G., Çelen Role of inflammatory and coagulation biomarkers in distinguishing placenta accreta from placenta previa and associated hemorrhage. J Clin Med. 2025;14(11):3884. doi: 10.3390/jcm14113884. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 19.Zong M., Pei X., Yan K., Luo D., Zhao Y., Wang P., et al. Deep learning model based on multisequence MRI images for assessing adverse pregnancy outcome in placenta accreta. J Magn Reson Imaging. 2024;59(2):510–521. doi: 10.1002/jmri.29023. [DOI] [PubMed] [Google Scholar]
- 20.Zheng W., Zhang H., Ma J., Dou R., Zhao X., Yan J., et al. Validation of a scoring system for prediction of obstetric complications in placenta accreta spectrum disorders. J Matern Fetal Neonatal Med. 2022;35(21):4149–4155. doi: 10.1080/14767058.2020.1847077. [DOI] [PubMed] [Google Scholar]
- 21.Happe S.K., Yule C.S., Spong C.Y., Wells C.E., Dashe J.S., Moschos E., et al. Predicting placenta accreta spectrum: validation of the placenta accreta index. J Ultrasound Med. 2021;40(8):1523–1532. doi: 10.1002/jum.15530. [DOI] [PubMed] [Google Scholar]
- 22.Chen K., Xue M., Meng X., Li G., Wang X., Long Y., Li H. Artificial Intelligence (AI) in the diagnosis and prediction of adverse pregnancy outcomes for Placenta Accreta Spectrum Disorders (PAS): a systematic review and meta-analysis of diagnostic accuracy. Int J Surg. 2026;112:8478–8490. doi: 10.1097/JS9.0000000000004443. [DOI] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
Supplementary Materials
The placental lacunae feeder vessels. The visualization of rapid, turbulent blood flow within placental lacunae on dynamic, real-time ultrasound imaging; characterized by high-velocity swirling that mimics the appearance of boiling water or shifting sands, this sign represents a severe manifestation of placental lacunae
Supplementary material Figure S1. Representative intraoperative uterine serosal appearances in patients with placenta accreta spectrum, based on the 2019 FIGO clinical classification. (a) Placenta accreta: no obvious placental bulge or distension over the placental bed, no placental tissue visible through the uterine serosa, and absent or minimal neovascularisation. (b) Placenta increta: abnormal macroscopic changes over the placental bed, including bluish-purple discolouration, a prominent placental bulge, and marked hypervascularity. No placental tissue is visible invading through the serosal surface. (c) Placenta percreta: marked abnormality of the uterine serosal surface, with placental tissue seen to be invading through the uterine wall or extending beyond the serosa
Data Availability Statement
The datasets generated and analyzed during the current study are available from the corresponding author upon reasonable request.