Materials and Methods
This was a retrospective review of gravidas with ≥1 prior cesarean deliveries who had sonographic confirmation of placenta previa or low-lying placenta in the third trimester at our ultrasound unit from December 1997 through December 2011 and were subsequently delivered at our hospital. This was a single-center study of a nonreferred population. Sonographic images and associated reports used in the clinical management of each patient were stored electronically in a picture archiving and communication system throughout the study period. The terms “prior cesarean delivery” and “placenta previa/low-lying placenta” were queried from our sonographic database, to identify women who met inclusion criteria. The diagnosis of placenta previa was based on the presence of placental tissue covering the internal cervical os. Low-lying placenta was diagnosed when the placenta was within 2 cm from the internal cervical os but did not cover it.
All transvaginal and transabdominal ultrasound images were individually reviewed by study investigators who were blinded to sonography report findings and pregnancy outcomes. If images were unable to be retrieved, the patient was excluded from analysis. Sonographic parameters evaluated from archived images included location of placenta, loss of the retroplacental clear zone, irregularity and thickness of the uterine-bladder interface, the smallest myometrial thickness in sagittal and transverse planes, presence of lacunar spaces, and bridging vessels. The technique and findings for each parameter are depicted in Figures 1-6 . In addition to grayscale imaging, color Doppler was used in assessment of abnormal vasculature. If any portion of the placenta covered the anterior lower uterine segment, the placenta was considered to be anterior. Lacunar spaces were graded according to Finberg and Williams as follows: grade 0, none seen; grade 1, 1-3 present and generally small; grade 2, 4-6 present and tending to be larger and more irregular; grade 3, many throughout the placenta and appearing large and bizarre. Confirmation of morbidly adherent placenta was based on histologic evidence of placental invasion from the hysterectomy specimen. Pregnancies that met study inclusion criteria but did not require a cesarean hysterectomy or have histologic evidence of placental invasion on hysterectomy specimen served as the comparison group.
Statistical analyses included linear logistic regression and multiparametric analyses. Only those parameters with complete data were used in the prediction model. The nonlinearity of the continuous measures was examined using cubic smoothing splines compared to the direct linear contribution. Measures found to offer significant improvement in prediction under nonlinear forms were used for the prediction equation. A receiver operating characteristic curve was derived for each combination of parameters (all subsets regression) to select the combination with the greatest area under the receiver operating characteristic curve. The combination of parameters that gave the largest area under the curve was used to generate a predictive equation, which we termed the “Placenta Accreta Index (PAI).” Each parameter was then given a weighted value based on the coefficients from the estimated regression equation to provide the PAI score from 0-9, with morbidly adherent placenta more likely at higher index scores. The sensitivity, specificity, positive predictive value (PPV), and negative predictive value (NPV) were calculated for each index score, with 95% confidence intervals (CIs), using standard methods for rates and proportions. Univariate analysis included Pearson χ 2 test for frequency measures and Student t test and Wilcoxon rank sum test for continuous measures. For presentation purposes, the receiver operating characteristic curves were smoothed using binormal estimation technique. P values < .05 were judged statistically significant. This study was approved by the institutional review board of the University of Texas Southwestern Medical Center.
Results
During the study period, 190 women with at least 1 prior cesarean delivery were diagnosed with either placenta previa or low-lying placenta during third-trimester sonography and were delivered at our institution. Images from 6 women were not able to be retrieved and thus excluded from the final analysis. Of 184 women in the final analysis, 54 (29%) had histologic confirmation of morbidly adherent placenta, and the remaining 130 women served as the comparison group. Maternal demographic characteristics are shown in Table 1 . There were no differences in age, race/ethnicity, or gestational age at ultrasound between pregnancies with and without histologic evidence of placental invasion. As expected, the number of prior cesarean deliveries was significantly associated with risk for placental invasion.
| Characteristic | Placental invasion, n = 54 | No placental invasion, n = 130 | P value |
|---|---|---|---|
| Age, y | 31.6 ± 5.28 | 31.1 ± 5.82 | .59 |
| Race/ethnicity, n (%) | |||
| Black | 5 (9) | 4 (3) | .30 |
| White | 3 (6) | 10 (8) | |
| Hispanic | 45 (83) | 110 (85) | |
| Other | 1 (2) | 5 (4) | |
| Gestational age at sonography, wk | 33.0 ± 2.7 | 33.7 ± 2.3 | .08 |
| Prior cesarean deliveries, n (%) | < .001 | ||
| 1 | 13 (24) | 87 (67) | |
| 2 | 24 (44) | 31 (24) | |
| ≥3 | 17 (31) | 12 (9) | |
| Prior uterine curettage, n (%) | 8 (15) | 21 (16) | .82 |
Sonographic findings are presented in Table 2 . Anterior placentation was significantly associated with placental invasion, P < .05, as was each of the sonographic parameters we assessed, all P < .001. Using both the categorical measures of our evaluation (loss of retroplacental clear zone, lacunae, bridging vessels, irregularity of uterine-bladder interface) and continuous parameters (myometrial thickness and uterine-bladder interface width), in addition to number of prior cesarean deliveries and placental location, receiver operator curves were constructed using different multiparametric combinations. Of the 184 pregnancies studied, 88 had each of these ultrasound variables assessed and were included in the final model, of which 29 had invasion proven on histologic exam. As shown in Figure 7 , the combination of placental location, smallest sagittal myometrial thickness, lacunae, bridging vessels, and number of cesarean deliveries yielded the greatest area under the receiver operating characteristic curve: 0.87 (95% CI, 0.80–0.95). When compared with the variables of placental location and number of prior cesarean deliveries alone, findings that would not generally require specialized sonography, the addition of smallest myometrial thickness, lacuna, and bridging vessels significantly improved the prediction of the model: P = .03. This is shown in Figure 8 . Through logistic regression modeling, a predictive equation, which we termed the “PAI,” was generated using these 5 parameters, based on the addition or weighting of each parameter. The estimated regression equation is defined as: PAI = e f /(1 + e f ), where f is a linear function of the selected parameters and e is the base associated with the natural logarithm.
| Variable | Placental invasion, n = 54 | No placental invasion, n = 130 | P value |
|---|---|---|---|
| Anterior placentation | 38 (70) | 36 (28) | < .001 |
| Lacunae | < .001 | ||
| Grade 0 | 11 (20) | 62 (48) | |
| Grade 1 | 15 (28) | 52 (40) | |
| Grade 2 | 10 (19) | 12 (9) | |
| Grade 3 | 18 (33) | 4 (3) | |
| Bridging vessels | 35 (65) | 24 (18) | < .001 |
| Sagittal smallest myometrial thickness, mm | 0.9 [0.0,2.2] | 2.4 [0.0,4.0] | < .001 |
| Uterine-bladder interface thickness, mm | 2.4 [0.0,4.0] | 3.3 [2.4,4.0] | .003 |
Incorporating the parameters yields the following equation:
f = − 0.1935 − 0.0404 ( grayscale sagittal myometrial thickness [ mm ] ) − 0.0911 ( if lacuna = 2 ) + 1.234 ( if lacuna = 4 ) + 0.4195 ( if bridging vessels ) + 1.1332 ( if > 1 prior cesarean delivery ) + 0.6772 ( if anterior placentation ) .
Stay updated, free articles. Join our Telegram channel
Full access? Get Clinical Tree
