There were 152 cases of suspected PAS identified from 2010–2018 that underwent hysterectomy at time of delivery with 131 confirmed PAS based on pathology (Fig. 1). Of the 131 (86.2%) confirmed cases, there were 38 (25%) placenta accretas, 62 placenta incretas (40.8%), and 31 (20.4%) placenta percretas. There were seven women who underwent cesarean hysterectomy based on preoperative diagnosis of PAS and extensively counseled that did not have PAS based on pathology (not included in the 152 total cases).. 21 (13.8%) of the suspected cases did not have PAS based on intraoperative findings and did not have a hysterectomy. 24.4% (32) cases were diagnosed with PAS at the time of delivery while 75.6% (99) cases were identified on preoperative imaging studies.
Most patients had at least one prior CD in 30 (79%) of accreta cases, 55 (88.7%) of increta cases, and 31 (100%) of percreta cases (Table 1). A history of prior dilation and curettage was noted in 8 (21.1%) accreta, 12 (19.4%) increta, and 6 (19.4%) percreta cases. AMA accounted for 14 (36.8%) accreta, 25 (40.3%) increta, and 15 (48.4%) percreta cases, respectively.
Table 1
Clinical risk factors by severity of placenta accreta spectrum (PAS). Advanced maternal age defined as ≥ 35 years old.
| Accreta (38) | Increta (62) | Percreta (31) | No PAS (21) |
Previous cesarean* | 30 (81.08%) | 55 (88.71%) | 31 (100%) | 14 (66.67%) |
0 | 7 (18.92%) | 7 (11.29%) | 0 (0%) | 7 (33.33%) |
1 | 17 (45.95%) | 25 (40.32%) | 6 (%) | 9 (42.86%) |
2 | 6 (16.22%) | 17 (27.42%) | 14 (%) | 2 (9.52%) |
3 | 3 (8.12%) | 10 (16.13%) | 3 (%) | 2 (9.52%) |
4 | 2 (5.41%) | 2 (3.23%) | 8 (%) | 1 (4.76%) |
≥ 5 | 2 (5.41%) | 1 (1.661%) | 0 (0%) | 0 (0%) |
Placenta previa** | 23 (69.70%) | 54 (90%) | 22 (75.86%) | 16 (76.19%) |
Previous D&C | 8 (21.05%) | 12 (19.35%) | 6 (19.35%) | 7 (33.33%) |
Advanced maternal age | 14 (36.84%) | 25 (40.32%) | 15 (48.39%) | 7 (33.33%) |
*note, data on previous cesarean section was only available for 37 of the 38 accreta cases, thus those percentages are out of 37 |
**5 previa missing – accreta (out of 33), 2 missing for increta (out of 60), 2 missing for percreta (out of 29), 0 missing for no placenta accreta spectrum (out of 21) |
Twelve cases of suspected PAS did not have ultrasound reports available and fifty cases of suspected PAS did not undergo pelvic MRI prior to delivery (Fig. 2). Sensitivity was 94% for ultrasound and 97.6% for MRI, while specificity for both modalities was 33%. When comparing pelvic MRI to ultrasound for either presence or absence of PAS, there was no significant difference in sensitivity or specificity in identifying PAS when compared to pathology (p = 0.525). When examining the accuracy for the degree of abnormal placentation, MRI correctly identified depth of invasion in 39% while ultrasound was correct in 25% (Fig. 2). Ultrasound more often underestimated the degree of abnormal placentation (72.4%) while MRI underestimation occurred in 30.7%. In patients with confirmed PAS, placenta increta was diagnosed most often; thus, it was used as the referent in the multinomial logistic regression for both imaging modalities. When ultrasound predicted absence of PAS, odds of pathology confirming was 4.49, and not statistically significant. When ultrasound predicted accreta, odds of pathology confirming were 0.91, also not statistically significant. When ultrasound predicted percreta, odds were 8.20, which was statistically significant (p < 0.01) (Table 2). When MRI predicted absence of PAS, the odds of the pathology confirming diagnosis (relative to increta) was 15.78, which was not statistically significant. When MRI predicted accreta, the odds of pathology concordance was 2.50, also not statistically significant. However, when MRI predicted placenta percreta, the odds was 0.89, which was statistically significant (p < 0.01) (Table 3). When either ultrasound or MRI predicted percreta, pathology was more likely to concur. Cohen’s kappa for ultrasound versus pathology, was 0.02836, demonstrating slight inter-rater agreement, but not significant (p = 0.55). Cohen’s kappa for MRI versus pathology was 0.12130, which also falls into the slight agreement category, but not significant (p = 0.08). The ultrasound area under the curve was 0.6471, and the MRI area under the curve was 0.5087, where 0.5 is what would be expected due to chance suggesting a low predictive utility for both MRI and ultrasound.
Table 2
Multinomial logistic regression comparing pathology and ultrasound. Pathology = Increta is the referent (applies to screenings for purposes of comparison). Top number is odds (exp(beta)), values below in parenthesis are the standard error of the beta-coefficient. Columns: ultrasound, Row: Pathology. ***p < 0.01. (US: ultrasound, PAS: placenta accreta spectrum)
| No PAS | Accreta | Percreta |
US No PAS | 4.488 (0.971) | 9.486 (0.815) | 3.008 (1.334) |
US Accreta | 1.633 (0.851) | 0.908 (0.758) | 4.101 (1.100) |
US Percreta | 0.0001 (83.282) | 0.272*** (1.239) | 8.201*** (1.142) |
Table 3
Multinomial logistic regression comparing pathology and MRI. Pathology = Increta is the referent (applies to screenings for purposes of comparison). Top number is odds (exp(beta)), values below in parenthesis are the standard error of the beta-coefficient. Columns: ultrasound, Row: Pathology. ***p < 0.01. (PAS: placenta accreta spectrum)
| No PAS | Accreta | Percreta |
MRI No PAS | 15.776 (1.158) | 3.507 (1.488) | 0.0001 (164.883) |
MRI Accreta | 0.750 (0.903) | 2.499 (0.746) | 0.333 (1.141) |
MRI Percreta | 0.375*** (0.744) | 0.500*** (0.772) | 0.889*** (0.576) |
Mean EBL was greater in patients with placenta percreta compared to increta and accreta (p = 0.04). 70.4% patients required some form of blood products (Table 4). In placenta accreta cases, 65.8% required PRBCS, 23.7% utilized cell saver, and 42.1% received other forms of blood products. For increta cases, 56.5%, 48.4%, and 43.5% needed PRBCs, cell saver, and other blood products, respectively. In percreta cases, 54.8% received PRBCs, 58.1% used cell saver, and 45.2% required other blood products. In patients with placenta accreta, 13.2% required admission to ICU and 21.1% experienced prolonged hospitalization (Table 5). For both increta and percreta cases, 29.0% were admitted to the ICU due to significant fluid shifts, while 37.1% (increta) and 54.8% (percreta) cases required prolonged hospitalization. Mean LOS was statistically significantly longer for patients with placenta percreta, compared to increta and accreta (p = 0.014). Prior CD occurred more often in the percreta cohort (p = 0.002). There were four cases of maternal mortality (two placenta increta, one accreta and one percreta).
Table 4
Blood products required by placenta accreta spectrum (PAS) cases intra-operatively. Divided by packed red blood cells (PRBCs), cell saver, and “other blood products” which includes fresh frozen plasma (FFP), platelets, and cryoprecipitate.
| Accreta (38) | Increta (62) | Percreta (31) | No PAS (21) |
PRBCs | 25 (65.79%) | 35 (56.45%) | 17 (54.84%) | 4 (19.05%) |
Cell Saver | 9 (23.68%) | 30 (48.39%) | 18 (58.06%) | 2 (9.52%) |
Other blood products | 16 (42.11%) | 27 (43.55%) | 14 (45.16%) | 1 (4.76%) |
Table 5
Maternal outcomes by severity of placenta accreta spectrum (PAS). Prolonged hospitalization defined as > 4 days post-partum, post-partum hemorrhage defined as > 1L of estimated blood loss (EBL), “other blood products” include fresh frozen plasma (FFP), platelets, and cryoprecipitate.
| Accreta (38) | Increta (62) | Percreta (31) | No PAS (21) |
Cesarean hysterectomy | 26 (68.42%) | 60 (96.77%) | 31 (100%) | 7 (33.33%) |
Prolonged hospitalization | 8 (21.05%) | 23 (37.10%) | 17 (54.84%) | 3 (14.29%) |
Postpartum hemorrhage | 26 (68.42%) | 45 (72.58%) | 22 (70.97%) | 5 (23.81%) |
PRBCs | 25 (65.79%) | 35 (56.45%) | 17 (54.84%) | 4 (19.05%) |
Cell saver | 9 (23.68%) | 30 (48.39%) | 18 (58.06%) | 2 (9.52%) |
Other blood products | 16 (42.11%) | 27 (43.55%) | 14 (45.16%) | 1 (4.76%) |
ICU admission | 5 (13.16%) | 18 (29.03%) | 9 (29.03%) | 1 (4.76%) |
Maternal death | 1 (2.63%) | 2 (3.23%) | 1 (3.23%) | 0 (0%) |