European Journal of Obstetrics & Gynecology and Reproductive Biology 151 (2010) 134–139

Contents lists available at ScienceDirect

European Journal of Obstetrics & Gynecology and

Reproductive Biology
journal homepage: www.elsevier.com/locate/ejogrb

Antenatal and intrapartum prediction of shoulder dystocia

Manish Gupta a,*, Christine Hockley b, Maria A. Quigley b, Peter Yeh a, Lawrence Impey a

a
Oxford Fetal Medicine Unit and Nuffield Department of Obstetrics and Gynecology, The John Radcliffe Hospital, 4th Floor, Women’s Centre, Oxford OX3 9DU, United Kingdom
b
National Perinatal Epidemiology Unit, University of Oxford, United Kingdom

ARTICLEINFO ABSTRACT

Article history: Objectives: To (1) develop algorithms to calculate the risk of shoulder dystocia at individual deliveries;
Received 11 August 2009 (2) evaluate screening for shoulder dystocia.
Received in revised form 5 March 2010 Study design: Retrospective analysis of 40284 consecutive term cephalic singleton pregnancies using a
Accepted 29 March 2010
‘train and test’ method. Four models were derived using logistic regression and tested (birthweight
alone; birthweight and other independent antenatal variables; birthweight and all independent
Keywords: antenatal and intrapartum variables; and all independent variables excluding birthweight).
Antenatal Results: Shoulder dystocia occurred in 240 deliveries (0.6%). Birthweight was the most important risk
Intrapartum
Prediction
factor although 98 cases (41%) occurred in babies weighing <4.0 kg. Birthweight and maternal height
were the only independent antenatal variables; for intrapartum use, only these and instrumental
Shoulder dystocia
delivery were independent. The antenatal model could calculate an individual’s risk; the intrapartum
model could also calculate the risk if an instrumental delivery were undertaken. Both showed 0.7%
women to have a risk of shoulder dystocia of >10%. Although the antenatal model had high predictability
(area under curve 0.89), it was no better than birthweight alone and had a sensitivity of 52.4%. Where
birthweight was excluded, prediction of shoulder dystocia was poor.
Conclusion: Antepartum and labour calculation of the risk of shoulder dystocia is possible. Whilst greatly
hindered by the inaccuracy of estimating weight, it allows due weight to be given to factors which may
already be influencing clinical practice. However, shoulder dystocia cannot be predicted with sufficient
accuracy to allow universal screening.
© 2010 Published by Elsevier Ireland Ltd.

1. Condensation The ideal would be prevention, rather than emergency

Independent antenatal and intrapartum risk factors for
shoulder dystocia can be used to calculate an individual’s risk
but effective screening is not possible.
2. Introduction
Shoulder dystocia at delivery is defined as occurring when
additional obstetric manoeuvres to release the shoulders are
required. Though rare, with an incidence of 0.2–2.0%, it is an
obstetric emergency with the potential for severe morbidity
and even mortality. Management relies on treatment rather
than prevention [1,2]. This is difficult: in a review of 56 fatal
cases of shoulder dystocia referred to in the Confidential Enquiry
into Stillbirths and Deaths in Infancy in 1994–1995 the mean
time between delivery of the head and body of the fetus was
5 min [3].
* Corresponding author. Tel.: +44 01865 751697; fax: +44 01865 851154.
E-mail address: dr_m_gupta@hotmail.com (M. Gupta).
management of shoulder dystocia. The best-known risk factor
is fetal macrosomia [4] but this is difficult to predict even with
ultrasound [5]. ACOG guidelines recommend that babies esti-
mated to weigh >5000 g should be delivered by caesarean section
and in diabetics the weight limit should be 4500 g [6]. UK
guidelines do not recommend caesarean section in non-diabetics
whatever the estimated fetal weight [1]. Therefore whilst
compliance with ACOG Guidelines will prevent very few cases,
the UK guidelines attempt virtually no antenatal prevention at all
and consider that ‘shoulder dystocia is.. .a largely unpredictable
and unpreventable event.’ In spite of this, the current situation is
that obstetric and maternal concern about fetal size is contribut-
ing to an increasing induction and caesarean section rate [7].
Further, successful medicolegal defence of a case of shoulder
dystocia is difficult [8].
If reliable risk factors other than birthweight were considered, it
might be that shoulder dystocia is more accurately predicted,
whilst preventing the virtual reliance on estimation of birthweight.
The aims of this paper are to (1) identify independent antenatal
and intrapartum risk factors for shoulder dystocia, (2) re-evaluate
the potential effectiveness of these in preventing shoulder

0301-2115/$ – see front matter © 2010 Published by Elsevier Ireland Ltd.

doi:10.1016/j.ejogrb.2010.03.025
 M. Gupta et al. / European Journal of Obstetrics & Gynecology and Reproductive Biology 151 (2010) 134–139 135

dystocia, and (3) develop individualised risks for shoulder dystocia
in a pregnancy both antepartum and during labour.
3. Materials and methods
This is a retrospective cohort study of all live deliveries in the
Oxford area between 01/04/1995 and 31/12/2002; after this the
database was unfortunately changed. Only singleton vaginal
cephalic deliveries at 36 or more completed weeks were included.
Mothers with pre-existing or gestational diabetes and those with
previous shoulder dystocia were further excluded from the study
because of their very high elective caesarean section rate.
Pregnancy data are entered into a database (OXMAT) by the
midwife, and after discharge by coding personnel. Shoulder
dystocia is coded according to the International Classification of
Diseases (ICD-10). This dataset has been extensively validated and
used: only where fields relied on data prior to the index pregnancy
(such as previous shoulder dystocia) was accuracy not high.
The following maternal characteristics were noted: ethnic
group, age, parity, gestation at delivery, height, weight and body
mass index (BMI), use of oxytocin, labour induction, epidural
analgesia, electronic fetal monitoring (EFM), length of first stage of
labour, length of second stage of labour and instrumental delivery.
The following neonatal characteristics were included: shoulder
dystocia, gender, birthweight, Apgar score <5 at 1 min, admission
to the neonatal unit, neurological sequelae and neonatal mortality.
Data were not available for neonatal brachial plexus injuries.
Each woman was randomly assigned to one of two groups (the
train and test datasets) such that the number of cases of shoulder
dystocia was the same (N = 120) in each group. The women in the
train dataset were used to derive the models and those in the test
datasets were use to validate the models. In the train dataset,
logistic regression was used to obtain four different models which
could be used to predict a woman’s risk of shoulder dystocia.
Lemeshow goodness of fit test. Statistical analyses were carried out
using STATA software, version 10.0 (Stata Corporation, College
Station, USA).
Ethics committee approval was granted (COREC) in September
2005.
4. Results
There were 52130 deliveries during the period. After exclusion
of 544 vaginal breech births, 1769 babies from multiple
pregnancies, 116 with diabetic mothers, 92 with prior shoulder
dystocia, 2560 born before 36 weeks, and 9389 born by caesarean
section (not mutually exclusive groups), there were 40284 births
available for analysis. There were 240 occurrences of shoulder
dystocia (risk 0.6%). In one there was a fatal outcome; in another
there were neonatal seizures.
The data were randomly split into two halves: the train and test
datasets. In the train dataset, the risk of shoulder dystocia
increased markedly as birthweight increased, from 0.1% in babies
with a birthweight of 3500 g or less to 10% in babies with a
birthweight above 4500 g. Approximately 41% of babies delivered
with shoulder dystocia weighed less than 4000 g. In univariate
analysis, shoulder dystocia was significantly associated with all
factors we examined except maternal ethnic group, maternal age,
parity and use of an epidural (Table 1). No difference was observed
between forceps or ventouse delivery.
A logistic regression model (Model A) using birthweight alone
as a risk factor was fitted using the train data. When this model was
Table 1
Potential maternal and perinatal risk factors for shoulder dystocia using the train
dataset.
Risk factor Total N = 20142a Shoulder p-Value
dystocia N =120a

Model (A) included birthweight alone as risk factor. The Birthweight (g) <0.001
subsequent three models included additional risk factors with a <3000 3259 2 (0.1%)
view to improving and comparing the resulting predictability. 3000–3500 7633 7 (0.1%)
3501–3999 6700 40 (0.6%)
Model (B) included all significant (p < 0.05) independent antenatal 4000–4500 2238 39 (1.7%)
risk factors (maternal height) with birthweight, to evaluate >4500 312 32 (10.3%)
antenatal screening, and create individualized risks for women
Gestation 0.023
in the antenatal period. 36–38 weeks 3490 10 (0.3%)
Model (C) included all significant (p < 0.05) independent 39–40 weeks 10744 67 (0.6%)
antenatal and intrapartum variables (maternal height and instru- >40 weeks 5908 43 (0.7%)
mental delivery) together with birthweight. This was to allow Sex of baby 0.013
individualized risks to be calculated intrapartum when contem- Male 10149 74 (0.7%)
plating instrumental delivery. Female 9993 46 (0.5%)

Model (D) included all significant (p < 0.05) independent Risk factor Total N = 20087 Shoulder p-Value
antenatal and intrapartum factors when birthweight was excluded dystocia N = 119
(maternal height, BMI, parity, gestation, baby’s sex, length of 2nd Ethnic group 0.382
stage of labour, and instrumental delivery). This was created to White 18558 108 (0.6%)
inform intrapartum decision-making as above, but not relying on Asian 647 5 (0.8%)
Afro-Caribbean 241 3 (1.2%)
anticipated birthweight. We did this because of the unreliability of
Oriental 174 2 (1.2%)
clinical or ultrasound estimation of fetal weight. Other/mixed 467 1 (0.2%)
The predictive power of each model was assessed in the train
Maternal age 0.157
dataset and a score was obtained for each woman by substituting
<20 939 4 (0.4%)
the woman’s data into the model equation. For each score value, 20–24 2618 11 (0.4%)
we estimated the sensitivity and specificity that would occur if all 25–30 7055 53 (0.8%)
women above a certain cut-off were classified as shoulder dystocia 31–40 9236 52 (0.6%)
and the women below the score were classified as not shoulder >40 294 0 (0.0%)

dystocia. We then identified the score cut-offs which yielded low Risk factor Total N = 15150 Shoulder p-Value
false positive rates (1%, 3% and 5%) as would be required in clinical dystocia N =85
practice. Receiver operator curves (ROC) of sensitivity plotted BMI 0.004
against (1 — specificity) were produced for the models shown and <18.5 434 1 (0.2%)
tables showing the groups at risk for particular cut-offs of model 18.5–24.9 9192 37 (0.4%)
25.0–29.9 3797 33 (0.9%)
scores were created. There was not strong evidence of poor fit in
>=30.0 1727 14 (0.8%)
any of the models presented, as assessed using the Hosmer–
136 M. Gupta et al. / European Journal of Obstetrics & Gynecology and Reproductive Biology 151 (2010) 134–139

Table 1 (Continued ) fitted to the test data and used to generate an ROC curve, the area
under the curve was 0.8419, suggesting high predictive power. The
Risk factor Total N = 16361 Shoulder p-Value
sensitivity of this model when applied to the test dataset for
dystocia N = 98
different false positive rates is shown in Table 3. In order to be used
Maternal height (m) as a screening tool, however, it would need to accurately identify a
<=1.63 2393 16 (0.67%) 0.009
large proportion of true shoulder dystocia cases, whilst having a
1.64–1.66 7365 57 (0.77%)
>1.66 6603 25 (0.38%) low false positive rate. The sensitivity was low (46%) even with a
false positive rate of 5% (Table 3).
Parity 0.895
Model B, based on the train dataset, was fitted using all
0 8376 46 (0.6%)
1 7305 45 (0.6%) antenatal factors from Table 1 including birthweight. Only
2 2982 19 (0.6%) maternal height and birthweight remained statistically significant
3+ 1479 10 (0.7%) (Table 2). This model was fitted to the test data and used to
Risk factor Total N = 19468 Shoulder p-Value generate an ROC curve. The model is good at predicting shoulder
dystocia N = 235 dystocia, with an area under the curve of 0.8884. However, there
EFM 0.007 was no increase in sensitivity of this model compared with the
Not performed 9270 42 (0.5%) birth weight only model (Table 3).
Continuous 10198 77 (0.8%) Model C, based on the train dataset, was fitted using all
Risk factor Total N = 19468 Shoulder p-Value antenatal and intrapartum factors from Table 1. Only maternal
dystocia N = 235 height, birthweight and instrumental delivery remained statisti-
Oxytocin augmentation 0.071 cally significant (Table 2). This model was fitted to the test data and
No oxytocin 15525 87 (0.6%) was good at predicting shoulder dystocia, with an area under the
Oxytocin 3943 32 (0.8%) curve of 0.8954. Whilst there was no increase in sensitivity of this
Type of delivery <0.001 model compared with Models A and B (Table 3), this model
SVD 16589 76 (0.5%) allowed us to calculate the risk of shoulder dystocia following an
Forceps 1706 21 (1.2%) instrumental delivery whilst taking into account all preceding
Ventouse 1847 23 (1.3%) variables.
Labour Induced 0.037 Model D, based on the train dataset, was fitted using all
No 15390 82 (0.5%) antenatal and intrapartum factors from Table 1 but excluding
Yes 4752 38 (0.8%)
birthweight. The factors which remained statistically significant
Length of first stage of labour (min) were gestation, baby’s sex, maternal height and BMI, parity, length
Mean (sd) 312.3 (215.1) 351.8 (236.1) 0.0704 of second stage of labour and instrumental delivery (Table 2). This
Length of second stage of labour (min) model was fitted to the test data, but was poor at predicting
Mean (sd) 54.7 (58.9) 81.1 (70.2) <0.001 shoulder dystocia, with an area under the curve of 0.6916 (Table 3).
Epidural 0.612
These models may be used to predict a woman’s risk of having a
No 15335 89 (0.6%) baby with shoulder dystocia as ey/(1 + ey), (Table 4). For models A, B
Yes 4807 31 (0.6%) and C, 98% of women had a risk of shoulder dystocia of <5% and 1%
a
N = 20087 overall and N = 120 shoulder dystocia cases unless indicated had a risk between 5 and 10%. For model D, 99.7% of women had a
otherwise. risk of shoulder dystocia of <5% and 0.3% had a risk between 5

Table 2
Crude and adjusted odds ratios for having shoulder dystocia using the train dataset.

Crudea OR (95% CI) Model Bb adjusted OR Model Cc adjusted OR Model Dd adjusted OR

(95% CI) (95% CI) (95% CI)

Birthweight (g) 1.0030 (1.0026–1.0034) 1.0034 (1.0030–1.0039) 1.0035 (1.0030–1.0039) –

Antepartum
Gestation (weeks) 1.25 (1.08–1.45) – – 1.27 (1.06–1.51)

Baby’s sex
Female 1 1
Male 1.59 (1.10–2.30) 2.02 (1.28–3.18)
BMI 1.06 (1.03–1.10) – – 1.05 (1.01–1.09)
Maternal height (cm) 0.9637 (0.9360–0.9923) 0.9031 (0.8757–0.9312) 0.9010 (0.8731–0.9298) 0.9574 (0.9278–0.9879)

Parity
0 1 – – 1
1 1.12 (0.74–1.69) 2.95 (1.67–5.19)
2 or more 1.18 (0.74–1.89) 3.56 (1.83–6.90)

Intrapartum
Instrumental delivery
No 1 – 1 1
Yes 2.72 (1.88–3.96) 2.86 (1.86–4.41) 2.14 (1.14–4.03)
Length of 2nd stage of labor (min) 1.006 (1.004–1.008) – – 1.01 (1.002–1.010)
a
Birthweight only.
b
Antenatal factors only and birthweight. Factors remaining significant (p < 0.05) were maternal height (continuous) and birthweight (continuous).
c
Antenatal and intrapartum factors and birthweight. Factors remaining significant (p < 0.05) were maternal height (continuous), birthweight (continuous) and
instrumental delivery.
d
Antenatal and intrapartum factors excluding birthweight. Factors remaining significant (p < 0.05) were maternal height (continuous), BMI (continuous), parity, gestation
(continuous), baby’s sex, length of 2nd stage of labour, and instrumental delivery.
 M. Gupta et al. / European Journal of Obstetrics & Gynecology and Reproductive Biology 151 (2010) 134–139 137

Table 4

All ‘known’ factors—birthweight excluded. Factors remaining significant (p < 0.05) were maternal height (continuous), BMI (continuous), parity, gestation (continuous), baby’s sex, length of 2nd stage of labour, and instrumental
0.016 (12.4: 6.6–20.6%) [1.6: 0.8–2.7%]
0.031 (5.2: 1.7–11.6%) [3.2: 1.1–7.4%]

0.020 (9.3: 4.3–16.9%) [1.9: 0.9–3.7%]

Coefficients produced using the train dataset.

Cut-off point (sensitivity %: 95% CI)

Model Aa Model Bb Model Cc Model Dd

Intercept —16.5491 1.3747 —1.6075 —10.4871

Birthweight (g) 0.0030 0.0034 0.0034

Antepartum
Gestation (weeks) 0.2367
[PPV %: 95% CI] Baby’s sex
Female 0.7019
Model Dd

Male 1
0.6916

BMI 0.0479
Maternal height (cm) —0.0102 —0.1042 —0.0436

Parity
0 1
1 1.2691
2 or more 1.0803

Intrapartum
All ‘known’ factors and birthweight. Factors remaining significant (p < 0.05) were maternal height (continuous), birthweight (continuous) and instrumental delivery.
Cut-off point (sensitivity %:

0.0755 (23.8: 38.7–58.5%)

0.0365 (39.1: 29.7–49.1%)

0.0239 (48.6: 16.0–33.1%)

Instrumental delivery
95% CI) [PPV %: 95% CI]

No 1 1
Yes 1.0514 0.7608
[13.1: 8.6–18.7%]

[7.7: 5.6–10.3%]

Length of 2nd stage 0.0064

[5.9: 4.4–7.7%]

of labour (min)
Sensitivity for given false positive rates using test data. Comparison of performance of birthweight (continuous) only model A with models B, C and D.

Antenatal factors only and birthweight. Factors remaining significant (p < 0.05) were maternal height (continuous) and birthweight (continuous).

a
The predicted probability of shoulder dystocia = ey/(1 + ey) where y is the sum of the
Model Cc

0.8954

coefficients, e.g. Model C: for a woman with baby’s birthweight estimated to be

4500 g, a height of 150 cm, who had an instrumental delivery, the predicted
probability of shoulder dystocia = exp{—1.6075 + (0.0034 × 4500) — (0.1042
× 150) + 1.0514}/1 + {exp(—1.6075 + (0.0034 × 4500) — (0.1042 × 150) + 1.0514}
= exp(—0.8861)/1 + exp(—0.8861) = 0.292 = 29%.

Table 5
Percentage risk of shoulder dystocia using the test dataset.
Cut-off point (sensitivity %:

0.0225 (52.4: 42.4–62.2%)

Percentage risk Model A Model B Model C Model D

0.068 (22.9: 15.2–32.1%)

0.034 (39.0: 29.7–49.1%)

95% CI) [PPV %: 95% CI]

of shoulder no. (%) no. (%) no. (%) no. (%)

dystocia
[12.8: 8.4–18.4%]

[7.6: 5.5–10.2%]

<0.25% 8962 (54.6) 9905 (60.3) 10353 (63.0) 4659 (30.7)

[6.3: 4.8–8.1%]

0.25 to <0.5% 3024 (18.4) 2539 (15.5) 2342 (14.3) 4857 (32.0)
0.5 to <1.0% 2208 (13.4) 1876 (11.4) 1656 (10.1) 3709 (24.4)
Model Bb

1 to <5% 1975 (12.0) 1787 (10.9) 1737 (10.6) 1900 (12.5)

0.8884

5 to <10% 192 (1.2) 219 (1.3) 225 (1.4) 48 (0.32)

10 to <20% 52 (0.32) 73 (0.4) 78 (0.5) 3 (0.02)
20 to <30% 10 (0.06) 19 (0.12) 25 (0.15) 1 (0.01)
30 to <40% 4 (0.02) 7 (0.04) 4 (0.02) 0 (0)
>=40% 2 (0.01) 4 (0.02) 9 (0.05) 1 (0.01)

and10%. The distribution of the risks for the women in these

Cut-off point (sensitivity %:

0.0315 (36.2: 27.0–46.1%)

0.0215 (45.7: 36.0–55.7%)

(95% CI) [PPV %: 95% CI]
0.060 (23.8: 16.0–33.1%)

models is shown in Table 5.

[12.8: 8.5–18.3%]

5. Comments
[7.1: 5.0–9.6%]

[5.5: 4.1–7.2%]

In this study we identify independent risk factors for shoulder

Model Aa

dystocia, develop an algorithm for calculating an individual’s risk

0.8419

of shoulder dystocia in different clinical scenarios, and evaluate the

integration of risk factors as an antenatal screening test.
In this population, only birthweight and maternal height were
independent antenatal risk factors; other factors such as gestation,
gender, BMI and parity were only independent when birthweight
was not considered (Model D). Of intrapartum variables (Model C),
only instrumental delivery was an independent risk factor.
Area under the ROC curve

Predictably, birthweight is the most important factor. Previous

analyses have concluded that a critical birthweight of 4 kg may be
Screen positive rate

Birthweight only.
(=1 — specificity)

used [9]. Yet more than 13% of all babies were this weight or more
and even for babies weighing more than 4500 g the risk is only about
10%. Further, 41% of neonates born with shoulder dystocia weigh less
than 4000 g (Table 1). These findings are broadly consistent with
delivery.
Table 3

others [10]. We found no association with induction of labour, nor

1%

3%

5%

with African ethnic origin, in contrast to Cheng et al. [11], although

b

d
a

c
138 M. Gupta et al. / European Journal of Obstetrics & Gynecology and Reproductive Biology 151 (2010) 134–139
our numbers in this group are small. Other potential risk factors
previously reported but on which we have no data include lack of
prenatal care [12], fundal height [13], and a high abdominal to head
size ratio [14], fetal growth velocity [15], shoulder dimensions [16]
and previous shoulder dystocia [17].
In clinical practice, individual risk factors frequently influence
clinical decision-making. Although RCOG Guidelines [1] do not
include recommendations regarding estimated fetal weight, those
from the ACOG do [6]. The failure to integrate other reported risk
factors such as maternal height, despite being influenced by them in
clinical decision-making, could lead to high intervention. This means
that screening using multiple risk factors is worth attempting.
In fact, our integrated models B and C perform no better than
that using birthweight alone (Model A): any improvement in
sensitivity at low false positive rates is not statistically significant.
This contradicts some previous attempts at screening. Mazouni
et al. [18] calculated the maternal height to infant weight ratio, but
this assumes a particular function of risk which may not always be
appropriate. Belfort et al. [13] used birth weight, 1 hour glucose,
and operative vaginal delivery. The most promising attempt,
suggesting better prediction than we achieve, is by Dyachenko
et al. [19]. This case-control study of 498 cases of shoulder dystocia
(including 90 with neonatal injury) and 622 controls predicted
50.7% of cases of shoulder dystocia with brachial plexus injury for a
2.7% false positive rate. Their algorithm is marketed as a web-based
tool allowing a calculation of whether the pregnancy is at high or
low risk of shoulder dystocia with neonatal injury (http://
www.lmsmedical.com) although it is unclear what proportion of
women will be advised they are at high risk. Their study has more
cases than ours and uses shoulder dystocia with neonatal injury.
However, prediction was based on a complex function of
gestational age, parity, birthweight and maternal height, which
may have caused over-fitting. We used data from a large
retrospective cohort study of a defined population to obtain a
simple predictive model which was a good fit to the data. We have
attempted to obtain unbiased estimates of model validity by using
half the data to derive the models and the other half of the data to
validate the models.
The principal limitation of any antenatal screening for shoulder
dystocia is a result of the importance of fetal weight. Estimated
weight, even using ultrasound, correlates poorly with actual
birthweight [5], particularly with the larger than average fetus
[20]. This means that all screening that includes an estimation of
fetal weight will be considerably less accurate than is projected by
actual birthweight. This problem also applies to previous
published analyses [13,18,19]. Even if this were not the case,
our data suggest that delivering all the babies weighing over 4.5 kg
by elective caesarean section (the current ACOG recommendation
[6]) would reduce the incidence of shoulder dystocia by only a
quarter. That we cannot improve on birthweight, either by adding
more reliably estimated factors to the model or by using only these
factors and not birthweight (Model D), suggests that whole
population antenatal screening for shoulder dystocia is not viable,
at least in a non-diabetic population.
It is important to differentiate between antenatal assessment of
risk, to decide whether elective caesarean should be performed, and
intrapartum assessment, as the morbidity of intrapartum caesarean
is higher [21]. To apply a screening test during labour would only be
worthwhile if it performed far better than antepartum, and we found
Model C to perform no better than Model B. What both models do
allow, however, is a calculation of risk for an individual. Such a
strategy is not going to lead to significant reduction in the incidence
of shoulder dystocia, but it will allow identification of those at very
high risk (Table 5). Further, Model C will allow calculation of the risk
if an instrumental delivery is undertaken. This may be important
because of fears over shoulder dystocia where such a delivery is
indicated, and could lead to caesarean section or summoning of
appropriate personnel.
There are limitations of our study in addition to the problem of
fetal weight. First, the diagnosis of shoulder dystocia is subjective,
but an overall incidence of 0.6% is unlikely to represent an over-
diagnosis. Second, data on brachial plexus injury were not
available. This is unfortunate, but such injuries follow excessive
traction, frequently resolve and are not the only adverse
consequence of shoulder dystocia. Third, in common with other
series, we did not have data on all potential risk factors. We
excluded diabetics because the incidence of elective caesarean
section in this group is very high in the Oxford area. Fifth, our
models are likely to perform better on the test dataset than they
would in another population, because the train and test datasets
have been drawn from the same population.
Shoulder dystocia is a rare but potentially disastrous complica-
tion of labour. Discussions regarding it are commonplace in
women with large babies or in shorter women; to fail to discuss
these risks has been considered negligent [8]. Our models allow
quantification of an individual’s risk, either before labour or when
contemplating instrumental delivery, by taking into account
factors that may, in clinical practice, be given inappropriate
weighting. We do not, however, deliver an acceptable antenatal
screening test for shoulder dystocia: the false positive rate is too
high, and the reliance on the knowledge of birthweight is too great.
Acknowledgements
We would like to thank all the participants and the midwives
for their hard work in caring for women in labour and entering the
information, thus allowing the creation of such an invaluable
database.
References
[1] Draycott TJ, Fox R, Montague IA, Royal College of Obstetricians and Gynaecol-
ogists Shoulder Dystocia: RCOG Guideline Number 42, December 2005.
[2] Gherman RB, Chauhan S, Ouzounian JG, Lerner H, Gonik B, Goodwin TM.
Shoulder dystocia: the unpreventable emergency with empiric management
guidelines. Am J Obstet Gynecol 2006;195:657–72.
[3] Hope P, Breslin, Lamont L, et al. Fatal shoulder dystocia: a review of 56 cases
reported to the Confidential Enquiry into Stillbirths and Deaths in Infancy. Br J
Obstet Gynaecol 1998;105:1256–61.
[4] Acker D, Sachs B, Friedman E. Risk factors for shoulder dystocia in the average
weight infant. Obstet Gynecol 1986;67:614–61.
[5] Rouse DJ, Owen J. Prophylactic caesarean delivery for fetal macrosomia
diagnosed by means of ultrasonography—a Faustian bargain? Am J Obstet
Gynecol 1999;181:332–8.
[6] Sokol RJ, Blackwell SC. American College of Obstetricians and Gynecologists.
Committee on practice bulletins—gynecology. Shoulder dystocia. Int J Gynecol
Obstet 2003;80:87–92.
[7] Thomas J, Paranjothy S. Royal College of Obstetricians and Gynaecologists. In:
Clinical Effectiveness Support Unit. National Sentinel Caesarean Section Audit
Report. RCOG Press; 2001.
[8] Clements R. Obstetric brachial plexus injury. Clin Risk 2006;12:3–11.
[9] Langer O, Berkus MD, Huff RW, Samueloff A. Shoulder dystocia: should the
fetus weighing greater than or equal to 4000 g be delivered by cesarean
section? Am J Obstet Gynecol 1991;165:831–7.
[10] Bahar AM. Risk factors and fetal outcome in cases of shoulder dystocia
compared with normal deliveries of a similar birthweight. Br J Obstet Gynecol
1996;103:868–72.
[11] Cheng YW, Norwitz ER, Caughey AB. The relationship of fetal position and
ethnicity with shoulder dystocia and birth injury. Am J Obstet Gynecol 2006
Sep;195(3):856–62.
[12] Sheiner E, Levy A, Hershkovitz R, et al. Determining factors associated with
shoulder dystocia: a population-based study. Eur J Obstet Gynaecol Reprod
Biol 2006;126:11–5.
[13] Belfort MA, Dildy GA, Saade GR, Suarez V, Clark SL. Prediction of shoulder
dystocia using multivariate analysis. Am J Perinatol 2007;24:5–10.
[14] Cohen BF, Penning S, Ansley D, Porto M, Garite T. The incidence and severity of
shoulder dystocia correlates with a sonographic measurement of asymmetry
in patients with diabetes. Am J Perinatol 1999;16(4):197–201.
[15] Kernaghan D, Farrell T, Ola B, Owen P. Fetal size and growth velocity in the
prediction of shoulder dystocia in a glucose impaired population. Eur J Obstet
Gynecol Reprod Biol 2009 (Epub ahead of print).
 M. Gupta et al. / European Journal of Obstetrics & Gynecology and Reproductive Biology 151 (2010) 134–139
[16] Verspyck E, Goffinet F, Hellot MF, Milliez J, Marpeau L. Newborn shoulder
width: a prospective study of 2222 consecutive measurements. Br J Obstet
Gynecol 1999;106:589–93.
[17] Overland EA, Spydslaug A, Nielsen CS, Eskild A. Risk of shoulder dystocia in
second delivery: does a history of shoulder dystocia matter? Am J Obstet
Gynecol 2009;200(May (5)):506.e1–6.
[18] Mazouni C, Porcu G, Cohen-Solal E, et al. Maternal and anthropomorphic
risk factors for shoulder dystocia. Acta Obstet Gynecol Scand 2006;85: 567–
70.
139
[19] Dyachenko A, Ciampi A, Fahey J, Mighty H, Oppenheimer L, Hamilton EF.
Prediction of risk for shoulder dystocia with neonatal injury. Am J Obstet
Gynecol 2006 Dec;195(6):1544–9.
[20] Weiner Z, Ben-Shlomo I, Beck-Fruchter R, Goldberg Y, Shalev E. Clinical and
ultrasonographic weight estimation in large for gestational age fetus. Eur J
Obstet Gynecol Reprod Biol 2002;105:20–4.
[21] Murphy DJ, Liebling RE, Verity L, Swingler R, Patel R. Early maternal and
neonatal morbidity associated with operative delivery in second stage of
labour: a cohort study. Lancet 2001;358:1203–7.