Severe postpartum haemorrhage and mode of delivery: a retrospective cohort study

Authors


C Holm, Department of Obstetrics, Rigshospitalet – Copenhagen University Hospital, Blegdamsvej 9, DK 2100 Copenhagen Ø, Denmark. Email charlotteholm@dadlnet.dk

Abstract

Please cite this paper as: Holm C, Langhoff-Roos J, Petersen KB, Norgaard A, Diness BR. Severe postpartum haemorrhage and mode of delivery: a retrospective cohort study. BJOG 2012;119:596–604.

Objective  To examine the association between intended mode of delivery and severe postpartum haemorrhage.

Design  A retrospective cohort study.

Setting  Material from a nationwide study in Denmark.

Population  Danish women giving birth in 2001–08.

Methods  We compared use of red blood cell transfusion by intended mode of delivery in the total population (n = 382 266), in low-risk nulliparous women (n = 147 132) and in women with a previous caesarean delivery (n = 25 156).

Main outcome measure  Red blood cell transfusion within 7 days of delivery.

Results  In the total population the crude transfusion rates for women with planned caesarean delivery and intended vaginal delivery were 2.24 and 1.75%. After adjustment for maternal age, body mass index, birthweight, smoking, parity, number of infants and previous caesarean delivery, the risk of red blood cell transfusion was significantly lower in women with planned caesarean delivery compared with intended vaginal delivery (odds ratio 0.82; 95% CI 0.73–0.92; < 0.01). In low-risk nulliparous women and in women with a previous caesarean delivery the transfusion rates were lower for planned caesarean delivery compared with intended vaginal delivery before and after adjustment.

Conclusion  Compared with intended vaginal delivery, planned caesarean delivery was associated with a reduced risk of severe postpartum haemorrhage indicated by use of red blood cell transfusion.

Introduction

Postpartum haemorrhage (PPH) is a major cause of maternal morbidity and mortality worldwide. This major public health issue has been further aggravated by the recent unexplained increase in the frequency and severity of PPH.1

Studies have shown that clinical estimates of blood loss are uncertain.2,3

Severe PPH is often defined as blood loss >1000 ml. However, as the estimate is of doubtful validity and may be influenced by other factors such as mode of delivery, the incidence of red blood cell (RBC) transfusion may offer a valid alternative when estimating the risk of severe PPH in relation to intended mode of delivery.

Studies on the relation between actual mode of delivery and the risk of severe PPH show that caesarean delivery is associated with a higher risk of severe PPH,4–6 but from the few studies on intended mode of delivery that have been published7,8 it is difficult to determine if planned caesarean delivery is associated with a higher risk of severe PPH than intended vaginal delivery.

Knowledge of the risk of severe PPH by intended mode of delivery may be more relevant than by actual mode of delivery when discussing preferred mode of delivery with the pregnant woman.

This study was designed to investigate the risk of RBC transfusion by intended mode of delivery in the total population, in nullipara with a singleton uncomplicated pregnancy at term, eligible for vaginal delivery, and in women with a second pregnancy after one caesarean delivery.

Methods

The study was a retrospective cohort study of a large percentage of the Danish obstetric population in 2001–08 (75%). We conducted a database linkage of the Danish Medical Birth Registry and the National Danish Transfusion Database (DTDB).

The Danish Medical Birth Registry (MBR) was established in 1968 registering all births in Denmark, as well as sociodemographic and medical characteristics about the mother, the infant and the delivery.9

All Danish births during the period 2001–08 were assessed for inclusion in the study. We included births at all hospitals using local electronic transfusion databases at the time of delivery (Figure 1). Transfusion data were collected from local blood bank transfusion databases and put on the DTDB. Copies of data from the local databases are sent to the DTDB once every year, where they are used for national bench-marking purposes. By Danish law, and in accordance with the European guidelines for preparation, use and quality assurance of blood components,10 the local blood bank transfusion databases retain specified data on all blood components transfused. The data include: the civil registration number of the woman, the type, blood groups and serial number of blood components, time of delivery of the blood component from the blood bank to the ward and the reporting of the completion of the transfusion with or without complications back to the blood bank.11 Complete traceability of blood components is demanded and intended in all registration and handling procedures. However, mean data coverage is 98–99% on actual transfusion of the component to the woman, because 1–2% of transfusion events are never reported back from the wards to the blood bank. For safety reasons such unconfirmed transfusion events are registered as actual transfusions in the transfusion databases, even though in some instances, the components may have been discarded.

Figure 1.

 Flow diagram showing the total study population, exclusions and subdivisions.

The main outcome variable was any RBC transfusion within 7 days of the delivery, after which time the transfusion rate declined to a minimum. Corresponding to the generally accepted mode of analysis by intention-to-treat we investigated the risk of RBC transfusion by intended mode of delivery. The definitions of intended mode of delivery are based on the Danish coding system. The four codes for caesarean delivery are (1) emergency prelabour, (2) planned (=elective) prelabour, (3) planned caesarean delivery performed in labour, and (4) emergency in labour. Planned is in the coding system defined as scheduled >8 hours before the procedure, regardless of whether this was before or during labour. Intended vaginal delivery included all vaginal deliveries and emergency caesarean deliveries in labour (code 4). Planned caesarean delivery included both prelabour and in labour caesarean delivery (codes 2 and 3). Women undergoing prelabour emergency caesarean delivery (code 1) were excluded, because the intended mode of delivery could not be ascertained through the registry.

A low-risk nullipara subpopulation was defined. It consisted of all nulliparas with a delivery of a single infant at term (gestational age ≥37 weeks), with an uncomplicated pregnancy defined by no pre-eclampsia, placenta praevia or placental abruption.

Statistical analysis

Analysis of the risk of RBC transfusion by intended mode of delivery was calculated and reported as crude and adjusted odds ratios (ORs). Logistic regression was used for the adjusted analysis. The following variables were adjusted for in the multivariate logistic regression model: Maternal age, body mass index (BMI), birthweight, smoking, parity, number of infants and previous caesarean delivery. The variables were identified a priori on the basis of previous literature12–14 and availability in the Medical Birth Registry. All variables were included in one multivariate regression model. Linearity of continuous variables was assessed by including the squares of the variables in the model. In case of non-linearity, data were grouped. Maternal age and birthweight were found to be non-linear, and grouped into the following categories 13–26, 27–29, 30–31, 32–34, ≥35 years and <2500, 2500–3499, 3500–3999, 4000–4499, ≥4500 g. respectively. The BMI was treated as a categorical variable in accordance with generally used categories: <18.5, 18.5–24.9, 25–29.9, ≥30 kg/m2. Smoking was divided into never, former and current smokers. The mother was defined as ‘former smoker’ if she stopped smoking before or when the pregnancy was discovered, and ‘current smoker’ if she reported continued smoking during the pregnancy regardless of the number of cigarettes per day. Parity was grouped into one, two and three or more deliveries and number of infants was grouped into one and two or more. In the comparison of the distribution of demographic variables between groups the Pearson chi-square test was applied. The analysis was performed using Stata 10.1 (StataCorp, College Station, TX, USA). The level of significance was set at a probability value of < 0.05.

Results

The total population consisted of 382 266 deliveries and 7335 (1.92%) received RBC transfusions within 7 days of delivery. A total of 24 357 RBC units were transfused (the median [interquartile range] was 2 units/delivery [2–4 units/delivery]). The low risk nullipara subpopulation consisted of 147 132 deliveries (38.5% of the total) and 3010 (2.05%) received RBC transfusions within 7 days of delivery. There were 25 156 women who had a second delivery after caesarean delivery and 754 (3.00%) received an RBC transfusion within 7 days of delivery.

Maternal and infant characteristics and obstetric details are summarised in Table 1. Considerable differences were observed between the planned caesarean delivery and intended vaginal delivery groups in the total population and the low-risk nullipara population. Multiple gestation, breech presentation and previous caesarean delivery were more frequent in women with a planned caesarean delivery (all < 0.001).

Table 1.   Demographics and obstetric characteristics of the study population
 Total populationLow-risk nulliparas
AllIntended VDPlanned CDAllIntended VDPlanned CD
  1. CD, caesarean delivery; VD, vaginal delivery

  2. *Failed induction, probably the indication for planned caesarean delivery.

N382 666336 24334 201147 132136 9727733
Maternal age in years (±SD) (missing: n = 9)30.7 (4.8)30.5 (4.7)32.3 (4.6)28.8 (4.6)28.7 (4.6)30.3 (4.8)
BMI (prepregnancy) in kg/m2 (±SD) (missing: n = 137 390)24.1 (4.8)24.0 (4.7)24.8 (5.2)23.6 (4.4)23.6 (4.4)23.8 (4.8)
Smoking, n (%)
Never312 708 (81.8)275 585 (82.0)28 001 (81.9)120 945 (82.2)112 658 (82.3)6341 (82.0)
Former6781 (1.8)6113 (1.8)488 (1.4)3771 (2.6)3551 (2.6)168 (2.2)
Current52 603 (13.8)46 324 (13.8)4410 (12.9)19 359 (13.2)18 036 (13.2)963 (12.5)
Unknown10 174 (2.7)8221 (2.4)1302 (3.8)3057 (2.1)2727 (2.0)261 (3.4)
Parity, n (%)
1173 388 (45.4)156 060 (46.4)11 382 (33.3)147 132 (100)136 972 (100)7733 (100)
2137 003 (35.8)118 515 (35.2)14 706 (43.0)
3+69 821 (18.3)59 827 (17.8)7963 (23.3)
Unknown2054 (0.5)1841 (0.6)150 (0.4)
Previous CD, n (%)38 989 (10.2)20 997 (6.2)14 894 (43.5)
Labour (%)
Spontaneous304 323 (79.6)290 272 (86.3)3740 (10.9)123 392 (83.9)120 504 (88.0)987 (12.8)
Induced48 074 (12.6)45 971 (13.7)592* (1.7)17 171 (11.7)16 468 (12.0)177* (2.3)
Actual mode of delivery, (%)
VD304 342 (79.6)304 342 (90.5)0 (0)119 930 (81.5)119 930 (87.6)0 (0)
CD77 924 (20.4)31 901 (9.5)34 201 (100)27 202 (18.5)17 042 (12.4)7733 (100)
Number of infants, n (%)
1373 267 (97.7)330 862 (98.4)31 704 (92.7)147 132 (100)136 972 (100)7733 (100)
2+8999 (2.4)5381 (1.6)2497 (7.3)
Gestational age in weeks (±SD) (missing: n = 1117)277.6 (14.3)279.1 (13.1)269.0 (11.5)281.1 (8.9)281.6 (8.7)273.3 (7.0)
Presentation, n (%)
Vertex361 891 (94.7)329 033 (97.9)23 994 (70.2)139 365 (94.7)134 614 (98.3)3086 (39.9)
Breech/Transverse18 521 (4.8)5596 (1.7)10 073 (29.5)7261 (5.0)1895 (1.4)4617 (59.7)
Unknown1854 (0.5)1614 (0.5)134 (0.4)506 (0.3)463 (0.3)30 (0.4)
Birthweight, g (±SD) (missing: n = 2443)3499 (596)3525 (560)3370 (630)3514 (476)3523 (471)3390 (487)

In the total population the planned caesarean delivery group had a significantly increased risk of RBC transfusion compared with the intended vaginal delivery group (2.24% versus 1.75%, OR 1.29; 95% CI 1.19–1.39; < 0.001). After adjustment for maternal age, BMI, birthweight, smoking, parity, number of infants and previous caesarean delivery, the risk of RBC transfusion was significantly lower in women with planned caesarean delivery (OR 0.82; 95% CI 0.73–0.92; < 0.01). The unadjusted and adjusted ORs for all the variables are shown in Table 2. The actual mode of delivery with the lowest risk of receiving RBC transfusion was intended vaginal delivery with spontaneous onset of labour and spontaneous delivery (1.25%) (Table 3).

Table 2.   Unadjusted and adjusted odds ratios for all variables in univariate analysis in the total population and in the low-risk nullipara population
 Transfusion rate (%)Unadjusted OR [95% CI]Adjusted OR* [95% CI]
  1. CD, caesarean delivery; ref., reference; VD, vaginal delivery.

  2. *Adjusted for maternal age, BMI, birthweight, smoking, parity, number of infants and previous CD for the total population and adjusted for maternal age, BMI, birthweight and smoking in the low-risk population.

Total population
Intended mode of delivery
 VD (ref.)1.7511
 CD2.241.29 [1.19–1.39] < 0.0010.82 [0.73–0.92] < 0.01
Maternal age in years
13–26 (ref.)2.0911
27–291.840.88 [0.82–0.94] < 0.0010.89 [0.81–0.98] P < 0.05
30–311.860.88 [0.82–0.95] < 0.010.96 [0.87–1.06] P = 0.43
32–341.740.83 [0.77–0.89] < 0.0010.94 [0.85–1.04] P = 0.20
≥352.060.98 [0.91–1.05] P = 0.611.10 [0.99–1.22] P = 0.07
BMI (prepregnancy)
<18.5 (ref.)2.3911
18.5–24.91.960.82 [0.72–0.93] P < 0.010.75 [0.65–0.86] P < 0.001
25.0–29.91.990.83 [0.72–0.96] P < 0.010.72 [0.62–0.84] P < 0.001
≥30.02.040.85 [0.73–0.99] P < 0.050.72 [0.61–0.85] P < 0.001
Birthweight, g
<2500 (ref.)4.0911
2500–34991.660.40 [0.36–0.43] P < 0.0010.87 [0.74–1.02] P = 0.09
3500–39991.640.39 [0.36–0.43] P < 0.0011.08 [0.92–1.27] P = 0.36
4000–44992.270.55 [0.50–0.60] P < 0.0011.57 [1.32–1.86] P < 0.001
≥45003.060.74 [0.65–0.84] P < 0.0012.11 [1.72–2.59] P < 0.001
Smoking, n
Current (ref.)1.9211
Former2.201.15 [0.97–1.35] P = 0.101.04 [0.85–1.28] P = 0.72
Never1.660.86 [0.80–0.92] P < 0.0010.95 [0.86–1.05] P = 0.35
Parity, n
1 (ref.)2.4411
21.450.59 [0.56–0.62] P < 0.0010.41 [0.38–0.44] P < 0.001
≥31.520.62 [0.58–0.66] P < 0.0010.34 [0.31–0.38] P < 0.001
No. of infants, n
1 (reference)1.8011
≥27.014.12 [3.79–4.49] P < 0.0014.77 [4.11–5.54] P < 0.001
Previous CD, n2.881.64 [1.54–1.75] P < 0.0012.58 [2.32–2.88] P < 0.001
Low-risk population
Intended mode of delivery
 VD2.0911
 CD1.240.59 [0.48–0.72] P < 0.0010.52 [0.40–0.68] P < 0.001
Maternal age in years
13–26 (ref.)2.0711
27–291.900.92 [0.83–1.01] P = 0.080.89 [0.78–1.00] P = 0.06
30–311.990.96 [0.87–1.07] P = 0.500.91 [0.80–1.04] P = 0.18
32–342.141.03 [0.92–1.16] P = 0.560.99 [0.86–1.15] P = 0.93
≥352.321.13 [1.00–1.28] P = 0.061.09 [0.93–1.27] P = 0.29
BMI (prepregnancy)
<18.5 (ref.)2.6711
18.5–24.92.120.79 [0.65–0.95] P < 0.050.69 [0.57–0.84] P < 0.001
25.0–29.92.210.82 [0.67–1.01] P = 0.070.67 [0.54–0.83] P < 0.001
≥30.02.120.79 [0.62–0.99] P < 0.050.63 [0.50–0.80] P < 0.001
Birthweight, g
<2500 (ref.)1.2111
2500–34991.461.21 [0.81–1.82] P = 0.361.38 [0.77–2.44] P = 0.28
3500–39992.091.75 [1.16–2.62] P < 0.052.03 [1.14–3.61] P < 0.05
4000–44993.532.99 [1.99–4.51] P < 0.0013.41 [1.91–6.08] P < 0.001
≥45005.164.45 [2.90–6.82] P < 0.0014.68 [2.56–8.55] P < 0.001
Smoking, n
Current (ref.)2.1011
Former2.201.05 [0.84–1.31] P = 0.660.98 [0.75–1.28] P = 0.87
Never1.660.79 [0.70–0.89] P < 0.0010.95 [0.82–1.10] P = 0.51
Table 3.   Transfusion rates in the total population and in low-risk nullipara by intended and actual mode of delivery
Intended modeof deliveryOnset of deliveryActual mode of deliveryTransfused (%)OR [95% CI]
  1. CD, caesarean delivery; VD, vaginal delivery.

Total population
Intended VD All5894 (1.75)
Spontaneous labourAll4574 (1.58)
Spontaneous delivery (reference)3008 (1.25)1
Operative vaginal925 (3.33)2.72 [2.52–2.93] < 0.001
Emergency CD710 (2.92)2.37 [2.19–2.58] < 0.001
Induced labourAll1320 (2.87)
Spontaneous delivery776 (2.32)1.87 [1.73–2.03] < 0.001
Operative vaginal286 (5.35)4.45 [3.93–5.04] < 0.001
Emergency CD284 (3.73)3.05 [2.69–3.45] < 0.001
Planned CDAll 767 (2.24)
PrelabourCD657 (2.17)1.75 [1.61–1.91] < 0.001
LabourCD110 (2.79)2.26 [1.86–2.74] < 0.001
Induction (failed)CD21 (3.55)2.90 [1.87–4.49] < 0.001
Low-risk nullipara
Intended VD All2856 (2.09)
Spontaneous labourAll2361 (1.96)
Spontaneous delivery (reference)1436 (1.63)1
Operative vaginal660 (3.21)2.01 [1.83–2.20] < 0.001
Emergency CD312 (2.35)1.45 [1.28–1.65] < 0.001
Induced labourAll495 (3.01)
Spontaneous delivery255 (2.59)1.61 [1.40–1.84] < 0.001
Operative vaginal149 (4.74)3.00 [2.53–3.57] < 0.001
Emergency CD103 (2.73)1.69 [1.38–2.07] < 0.001
Planned CDAll 96 (1.24)
PrelabourCD79 (1.18)0.72 [0.57–0.91] P<0.01
LabourCD17 (1.62)1.00 [0.61–1.61] = 0.99
Induction (failed)CD3 (1.69)1.04 [0.33–3.26] = 0.95

In the low-risk nullipara subpopulation the transfusion rate in the planned caesarean delivery group was significantly lower compared with the intended vaginal delivery group, namely 1.24% versus 2.09% (OR 0.59; 95% CI 0.48–0.72; P < 0.001). When the planned caesarean delivery group was stratified into breech and non-breech presentation the transfusion rates were 1.06 and 1.52%. After adjustment for maternal age, BMI, birthweight and smoking the risk reduction remained almost the same (OR 0.52; 95% CI 0.40–0.68; < 0.001). The unadjusted and adjusted ORs for the variables in the low-risk nulliparas are shown in Table 2.

The lowest risk for RBC transfusion in the low-risk nullipara was planned caesarean delivery before onset of labour (1.18%) (Table 3).

The influence of previous mode of delivery on the RBC transfusion rate is shown in Table 4. In the total population the highest transfusion rate (3.23%) was seen for intended vaginal delivery after a previous caesarean delivery (vaginal birth after caesarean), and the lowest transfusion rate (1.00%) was for intended vaginal delivery after previous vaginal delivery. Following a previous vaginal delivery, a planned caesarean delivery was associated with a significantly higher risk of RBC transfusion compared with an intended vaginal delivery (OR 1.60; 95% CI 1.29–1.99; < 0.001). Following a previous caesarean delivery, the risk of RBC transfusion was significantly lower in the planned caesarean delivery group compared with an intended vaginal delivery (OR 0.67; 95% CI 0.56–0.79; < 0.001).

Table 4.   Transfusion rates in all women by intended mode of delivery and previous actual mode of delivery
First deliverySecond delivery
Actual mode of deliveryIntended mode of deliveryTransfusion rateOdds ratio [95% CI]
Vaginal deliveryCaesarean1.60% (92/5751)1.60 [1.29–1.99] < 0.001
Vaginal1.00% (1047/104 342)
Caesarean deliveryCaesarean2.18% (195/8955)0.67 [0.56–0.79] < 0.001
Vaginal3.23% (458/14 173)

We found that women planning a caesarean for the second delivery had a significantly higher risk of receiving an RBC transfusion if they had a previous caesarean delivery compared with a previous vaginal delivery (OR 1.37; 95% CI 1.07–1.76; < 0.05), and those women intending a vaginal delivery after a previous caesarean delivery had a three-fold higher transfusion rate compared with the intended vaginal delivery after a previous vaginal delivery (OR 3.29; 95% CI 2.95–3.68; < 0.001).

Discussion

The present study found that planned caesarean delivery is associated with a reduced risk of severe PPH, as indicated by the use of RBC transfusions, compared with intended vaginal delivery. The use of both a national registry and local databases enabled us to study a large cohort, and by using RBC transfusion as an indicator of severe PPH, we may have a better indicator than imprecise clinical estimates of blood loss and different cut-off levels for defining severe PPH (500 or 1000 ml) used in most studies.

Estimates of blood loss are influenced by mode of delivery. Usually, it is easier to estimate the blood loss at a caesarean delivery compared with a vaginal delivery, because the amount suctioned is measured and the surgical swabs can be weighed. However, inclusion of amniotic fluid means that the blood loss at a caesarean delivery before rupture of membranes may be overestimated. By selecting RBC transfusion as an indicator of severe PPH we have chosen a measure that is influenced by haemoglobin levels and the clinical condition of the woman, both of which are probably less influenced by mode of delivery.

An inherent weakness in our study is the observational design and register-based information on a limited number of variables. Hence, an uncomplicated pregnancy was defined by no pre-eclampsia, placenta praevia or placental abruption, and many other complicating factors possibly affecting the risk of receiving an RBC transfusion were not analysed. Also, it seems plausible to assume that risk factors for receiving RBC transfusions are unevenly distributed between women intending caesarean delivery and vaginal delivery, with the highest burden of comorbidity and risk factors present in those planning a caesarean delivery. This assumption was supported by the adjusted analysis, because adjustment lowered the OR of RBC transfusion with planned caesarean delivery in both the total and the low-risk nullipara populations. We were only able to adjust for factors routinely registered in the birth registry and under-adjustment seems likely. Further supporting the assumption, the transfusion rate in the planned caesarean delivery group in the low-risk nullipara population was lower among those with breech presentation compared with those with a cephalic presentation. For cephalic presentation we do not know the indications, and this might have been a medical or pregnancy complication increasing the risk of PPH. For breech presentation there is usually no other indication than the presentation itself and the group might be comparable to caesarean delivery on maternal request. Hence, the true risk reduction by planned caesarean delivery may be even larger than what has been found in this study. Another potential weakness is that at caesarean delivery, when the anaesthetist is present and intravenous fluids are already running, there may be a different threshold for RBC transfusion. It seems likely that the threshold might be lower in the caesarean delivery situation, again suggesting a possible underestimation of the risk reduction by caesarean delivery in this study.

Furthermore, a potential confounder in using RBC transfusion as indicator of severe PPH is antepartum anaemia, which increases the need for RBC transfusion; however, antepartum anaemia is probably not associated with intended mode of delivery.

Several studies on RBC transfusion have found that caesarean delivery is associated with a higher risk of severe PPH compared with vaginal delivery.12,13,15–17 These studies compared actual mode of delivery. In contrast, our study analysed intended mode of delivery in accordance with the principle of intention-to-treat. Few studies on intended mode of delivery have been made. One reported no difference in risk of RBC transfusion between caesarean delivery without labour and spontaneous onset of labour.8 and a Canadian register-based study with extraordinarily low transfusion rates confirms our findings, reporting the risk of haemorrhage requiring transfusion as significantly lower in low-risk planned caesarean delivery (0.2‰) compared with planned vaginal delivery (0.7‰).7

We found as most, but not all other studies have, that when attempting vaginal delivery after a previous caesarean delivery there is an increased risk of receiving transfusion compared with prelabour caesarean delivery after a previous caesarean.8,17–19

A strength in our study is that we retrieved data from an original primary source, the blood banks. Some studies rely on secondary coding and hospital discharge registers, which may be incomplete.13,16 We therefore had the possibility of conducting the analyses on transfusions given within 7 days postpartum. Previous international studies used transfusion within 24 hours and 3 days postpartum as indicators of PPH.20,21 This influenced the total rate because only 73.2% of the transfused women in the present study had RBC transfusion within 1 day and 94.0% within 3 days postpartum.

In our study the overall transfusion rate was 1.92%, which is higher than reported from Australia, Canada, France and Norway,13,16,20,22,23 but lower than the overall transfusion rates in single-centre surveys from Oxford, UK24 and California, USA,25 and in a Dutch two-centre study.26

Transfusion practices vary between countries, and Australia, Canada, France and Norway have low population transfusion rates.27 Denmark has a very high use of RBC,27 which theoretically makes the measure RBC transfusion a sensitive indicator of severe PPH, but perhaps also less specific. Also, regional differences and individual variations in physicians’ thresholds for transfusion may influence transfusion rates. Furthermore, a general tendency to transfuse two units at a time may contribute to the high frequency of two-unit transfusion events in this study. However, we still consider these variations to be smaller, and more independent of delivery mode, than the variation associated with clinical bleeding estimates. We can, however, not exclude that the relatively high transfusion rate in Denmark may also be the result of a high rate of severe PPH especially in women with intended vaginal delivery.

In the light of the lack of evidence of benefit of RBC transfusions for purposes other than the immediate restoration of haemodynamic stability, tissue oxygenation and haemostasis during bleeding, and the increasing amount of evidence of immunological and cardiovascular side effects,28–30 we think that it is of clinical significance to include risk of RBC transfusions when evaluating the safety of the mode of delivery. Studies on transfusion practice, guidelines for prophylaxis and treatment of PPH and their implementation in clinical practice are required to explain the national differences and the association between mode of delivery and PPH, and—most importantly—to reduce severe PPH and the risks associated with RBC transfusion.

Conclusion

Planned caesarean delivery is associated with a reduced risk of severe PPH indicated by use of RBC transfusion. The results of this study may be used in the counselling of pregnant women about the risk of severe PPH by intended mode of delivery in first and second pregnancies. It should be emphasised, however, that individual risk assessment is still important and that the need for RBC transfusion is only one of several fetal and maternal factors that should be considered when choosing mode of delivery.

Disclosure of interest

All authors declare that they have no disclosures to report.

Contribution to authorship

CH designed the study, performed analyses and drafted the manuscript. JLR and BRD contributed substantially to the study design, analyses and revision of the manuscript. AN and KBP contributed with discussion of analyses and revision of the manuscript. All authors contributed to the final manuscript.

Details of ethics approval

The study was approved by the Danish National Data Protection Agency.

Funding

No financial support.

Acknowledgements

Special thanks are due to the Danish National Transfusion Database for data collection and collaboration, to Steen Rasmussen at the Danish Medical Birth Registry for merging of data, and to Elisabeth Anne Wreford Andersen and Susanne Rosthøj, Department of Biostatistics, University of Copenhagen, for help with the statistical analyses.

Ancillary