Predictors for Massive Haemorrhage during Caesarean Delivery Due to Placenta Praevia

Objectives: To describe the incidence and predictive factors for massive haemorrhage during a caesarean delivery in women with placenta praevia. Methods: We performed a retrospective cohort study among consecutive women with a placenta praevia planned for a caesarean delivery between 2001-2011 in one academic centre and two teaching hospitals in the Netherlands. Massive haemorrhage was defined as ≥1000 mL blood loss during caesarean section. We used logistic regression analysis to assess which maternal and pregnancy characteristics were predictors for massive haemorrhage during caesarean delivery. Results: Of 54,794 deliveries, we identified 215 (0.39%) women with placenta praevia who underwent a Caesarean delivery, of whom 94 (44%) had massive haemorrhage at or after surgery. After univariable preselection, multivariable logistic regression indicated that higher maternal age (OR 2.09; 95% CI 1.17 3.74), no Caucasian ethnicity (OR 1.73; 95% CI 0.92 3.27), multiple pregnancy (OR 3.92; 95% CI 0.72 21.28), lower systolic blood pressure during pregnancy (OR 1.03; 95% CI 1.00 1.07) and the placenta located at the anterior wall (OR 2.21; 95% CI 1.20 4.04) were independent predictors for massive haemorrhage during a caesarean delivery in women with placenta praevia. Predicted probabilities varied between 22% and 71%. The model has a reasonable discriminative ability and acceptable calibration. Conclusions: Women with placenta praevia are at high risk for massive haemorrhage during caesarean delivery. The risk profile can be used to identify extreme high-risk women in whom extensive preventive measures are justified.


Introduction
Placenta praevia, defined as a placenta that overlies or is proximate (≤2 cm) to the internal ostium (os) of the cervix, has a prevalence at delivery of 0.3% -0.5% [1] [2].In the United States, maternal mortality occurs in 0.03% of the cases [2].Although the aetiology of placenta praevia is unclear, several risk factors for developing a placenta praevia have been described, such as higher maternal age, multiparity, previous caesarean delivery, previous abortion or miscarriage, smoking and cocaine use during pregnancy [3].
Three small retrospective studies suggest that women in whom the distance between the lower placental edge and the internal cervical os was >2 cm, a so called low-lying placenta, could safely have a vaginal delivery [4]- [6].Placenta praevia requires an elective caesarean delivery, as delivery is often complicated by massive haemorrhage [1] [7].Contractions and cervical effacement and dilatation cause the placenta to separate from the uterine wall and provokes bleeding, which can stimulate contractions [1].
If massive haemorrhage cannot be controlled with conservative methods such as uterotonic agents, then surgical haemostatic sutures, radiologic interventions or eventually an emergency hysterectomy are the only life saving procedures.Recent studies showed that in about 0.5% of all caesarean deliveries a peripartum hysterectomy is done, of which 32% is due to an uncontrollable bleeding caused by a placenta praevia [8] [9].
Identifying women at high risk for massive haemorrhage due to placenta praevia can help to select the cases in which extra precautions should be taken, like availability of an intervention radiologist and extra venous access during the caesarean section.Risk factors described in a little number of other studies were higher maternal age, previous caesarean section, anterior location of the placenta, total placenta praevia, no bleeding before the caesarean section and the use of general anaesthetics [10]- [13].A prediction model has currently not been made.Therefore the aim of this study was to further investigate the incidence and predictive factors and to make a prediction model for massive haemorrhage during a caesarean delivery due to a placenta praevia.

Design and Participants
We performed a retrospective cohort study.We included all women diagnosed with placenta praevia who delivered between January 1 st 2001 and December 31 st 2011.The study was performed in one university medical centre (Academic Medical Centre in Amsterdam) and two teaching hospitals (Onze Lieve Vrouwe Gasthuis in Amsterdam and Spaarne Ziekenhuis in Hoofddorp).
Women who delivered with a placenta praevia were selected from the delivery registration at the labour wards.A placenta praevia was defined as a placenta with a maximum distance to the internal os of 20 mm, diagnosed by transabdominal and transvaginal ultrasound.When a placenta praevia was suspected at the routine transabdominal ultrasound at 20 weeks of gestation, a transvaginal ultrasound was planned between 28 and 32 weeks of gestation to confirm the diagnosis, or earlier if clinically indicated.In our study the last transvaginal ultrasound was used to confirm the diagnosis.We excluded women who delivered vaginally and women who delivered <24 weeks of gestation.We also excluded women with pre-eclampsia, HELLP syndrome, clotting disorders and women using anti-coagulants, as in these disorders vaginal blood loss might have another cause than solely the placenta praevia.

Outcome and Possible Risk Factors
The outcome, major haemorrhage, was defined as ≥1000 mL blood loss during caesarean section.As potential predictors, based on literature search, we considered general and obstetric history, race, age, smoking during pregnancy, Body Mass Index (BMI), use of medication, highest diastolic (DBP) and systolic blood pressure (SBP) during pregnancy, multiple pregnancy, type of placenta praevia (total, partial of marginal), anterior location of the placenta, gestational age at first bleeding, hospitalization, number of hospitalizations, steroids given during pregnancy, tocolytics given during pregnancy, Haemoglobin (Hb) level at start pregnancy, iron suppletion or transfusions given during pregnancy, Hb before delivery and gestational age at delivery.
We also determined which type of anaesthesia was used during the delivery.We believe the choice for loco regional or general anaesthesia is often based on clinical judgement and the degree of emergency rather than it being an independent predictor for massive haemorrhage.Because of the risk of confounding by indication this variable was not included in our further analysis.All data were extracted from the electronic patient charts.

Analysis
We used univariable and multivariable logistic regression analysis to assess whether in women with placenta praevia the potential predictors (mentioned above) were associated with major haemorrhage during caesarean delivery.
Factors with a p-value < 0.20 in the univariable analysis were included in a multivariable model.A multivariable model to identify independent predictors for massive haemorrhage was constructed using backward stepwise elimination.Variables with a p-value < 0.10 remained in the final model.The crude and adjusted Odds Ratios (OR) and their 95% confidence intervals (95% CI) were computed.
We constructed a Kaplan-Meier curve to display the time to delivery in the non-major haemorrhage and the major haemorrhage group.A log-rank test was used to test for differences between groups.
Predictive accuracy and calibration of the model were compared using a ROC curve and the Hosmer and Lemeshow goodness of fit test, respectively.The calibration plot indicates the range of predictions that are made, as well as their reliability.A strongly increased risk of haemorrhage would indicate that preventive measures would be warranted, whereas in low risk predicted patients this would not be necessary.
For the comparison of the baseline characteristics p-values were calculated using the Chi-square test for differences for categorical variables and a t-test for normally distributed continuous variables.For the logistic regression the Wald test was used.All statistical analyses were performed with SPSS version 19.0 (Chicago, IL, USA).

Results
Of 54.794 women who delivered during the 10-year study period at the three hospitals, 243 (0.44%) were diagnosed with placenta praevia.Of these, we excluded 28 women for the following reasons: 13 delivered vaginally, 3 delivered before 24 weeks of gestations, 6 had pre-eclampsia or HELLP syndrome, 3 had clotting disorders and 3 used anti-coagulants.The flow chart is shown in Figure 1.
The 215 patients included in the analysis were divided into two groups, 94 (44%) with massive haemorrhage during delivery and 121 (56%) without massive haemorrhage.In the group with massive haemorrhage the mean blood loss was 2066 mL (median: 1500 mL), in the group without massive haemorrhage this was 550 mL (median: 500 mL).In the massive haemorrhage group, 87 (93%) women underwent one or more interventions to stop the bleeding: haemostatic sutures in 73 women (78%), one embolization of an artery (1.1%) was performed by an intervention radiologist, and two (2.1%) emergency hysterectomies were performed.The emergency hysterectomies were performed for an uncontrollable bleeding and due to a placenta accrete respectively.Moreover, 48 women (42%) had blood transfusion with on average 2 packed cells (range 0 -16).
In six cases, the placenta praevia was accompanied by placenta accrete (2.8%), all identified during the caesarean section, of whom five had massive haemorrhage.Six patients had a previous placenta praevia (2.8%), of which three had massive haemorrhage.
Figure 2 shows the Kaplan Meier curve comparing the gestational age at delivery for women with and without massive haemorrhage.The median gestational age at delivery was 38 weeks + 1 day in the group with and 38 weeks + 2 days in the group without massive haemorrhage (p-value for difference 0.78).
Figure 4 shows the calibration plot of the observed and predicted risks of massive haemorrhage during caesarean section.The observed and the predicted risks are quite similar, the Hosmer and Lemeshow test assessing calibration showed a good fit (p-value 0.83).

Discussion
Hasegawa et al. also found advanced maternal age and a placenta at the anterior wall as predictors for massive bleeding during the caesarean delivery [11].They additionally identified a previous caesarean section as a strong risk factor (OR 7.3, 95% CI 1.9 -28.2).In our study this was not significant in univariable analysis (OR 1.5, 95% CI 0.7 -2.9).Differences in results may be caused by the difference in the definition of major haemorrhage (>2500 mL in the study by Hasegawa et al. vs. ≥1000 mL in our study).Jang et al. reported that a placenta at the anterior wall had a higher incidence of massive bleeding compared to a placenta located at the posterior wall (OR 2.9, 95% CI 1.6 -5.4) [12].This finding resembles with our results (OR 2.3; 95% CI 1.3 -4.2).
Watanabe et al. reported that patients without blood loss before delivery, bled more than those with blood loss before start of delivery (1563 ± 790 ml vs. 1276 ± 688 ml, p = 0.02) [13].That finding was not reproduced in our study.We found no difference between the occurrence of no bleeding and massive haemorrhage at delivery (OR 1.20; 95% CI 0.65 -2.24).The study population of Watanabe et al. was smaller (n = 114) than ours, with wide ranges of blood loss during/after caesarean section observed in patients both with and without warning bleeding.They suggest further research, to confirm their results.
Our univariable analysis showed that patients who received general anaesthesia during caesarean delivery were more likely to have a massive bleeding (OR 2.36; 95% CI 1.31 -4.24, p-value < 0.01).However, we did not include this factor in our prediction model, as the choice of anaesthesia is usually based on clinical factors already present like amount of blood loss at the time of decision, placental localisation and whether or not the delivery is an emergency.Our findings are similar with the findings of Frederiksen et al. [10] and Afolabi et al. [14] who reported that general anaesthesia increased intraoperative blood loss and the need for blood transfusion as compared to regional anaesthesia.This difference can be caused by muscle relaxation and opening of the blood vessels with general anaesthesia.In case of a higher risk for massive haemorrhage, for instance in cases with placenta praevia, general anaesthesia should be avoided.
Strong points of our study are the relatively large sample size.We included all women with a placenta praevia during a 10-year period in three hospitals.Furthermore this study contributes to a better risk estimation en makes the obstetrician more aware of the problems that may be encountered.
Our study has also some weaknesses.The study was retrospective; some cases might have been documented inaccurately and some cases might have been missed due to inaccuracy of diagnosing placenta praevia.However, the retrospective study design allowed us to include a relatively large number of patients with placenta praevia.

Observed probability
Secondly, the observed and the predicted risks are quite similar, but there is some uncertainty caused by the large confidence intervals (Figure 4).This may lead to situations where a patient at high risk is assigned a low predicted probability, and thus forgoes possible preventive interventions.The first and fifth points are not overlapping indicating a patient in the lowest predicted risk category will have lower observed massive haemorrhage than patients in the highest predicted risk category.
Validation can be done by prospective use of the prediction model.The clinical outcome of two groups, one with and one without use of the model, can be compared.After validation the model may lead to adaption of the policy concerning placenta praevia and as a result reduce the occurrence of massive haemorrhage.We recommend at this stage that women with a placenta praevia and predictors, despite the number, should be treated with caution and precautions must be taken, such as the availability of a multidisciplinary team, notifying an intervention-radiologist, extra venous access and enough blood, medication and uterotonics in the operation room in order to be able to stop the bleeding.A multidisciplinary protocol should be made by gynaecologists, anaesthesiologists, radiologists and coagulation specialists to provide optimal care.
The risk for massive haemorrhage ranged from 15% for patients aged under 35 with a Caucasian ethnicity, a singleton pregnancy, the placenta located at the posterior site and a higher systolic blood pressure during pregnancy (123 mmHg ± 12), to 91% for patients aged over 35 with no Caucasian ethnicity, a multiple pregnancy, the placenta located at the anterior site and a lower systolic blood pressure during pregnancy (120 mmHg ± 11).Obstetricians should be aware of this high risk.

Figure 1 .
Figure 1.Flowchart of inclusion of patients.

Figure 2 .
Figure 2. Comparison of the gestational age at delivery for placenta praevia with and without massive haemorrhage.

Table 1 .
Univariable analysis of maternal, pregnancy and intrapartum characteristics assosciated with massive haemorrhage.
* Mean (SD).** This factor wasn't included in our prediction model, as it is the choice of the clinician based on factors already present, rather than an independent factor.a Systolic blood pressure; b Confidence interval; c Body mass index; d Systolic blood pressure; e Diastolic blood pressure; f Hae moglobin; g Caesarean section.

Table 2 .
Factors associated with massive haemorrhage in multivariable analysis.