Labour characteristics and neonatal Erb’s palsy


Dr WR Cohen, Department of Obstetrics and Gynecology, Jamaica Hospital Medical Center, 89-06 135th Street, Suite 6A, Jamaica, NY 11418, USA. Email


Objective  To determine risk factors for Erb’s palsy, with a focus on graphic labour patterns.

Design  A case–control study.

Setting  New York City.

Sample  A total of 45 consecutive cases of Erb’s palsy and 90 controls.

Methods  Pregnancies and labours of neonatal Erb’s palsy cases were compared with 90 controls using univariate and multiple logistic regression analysis.

Main outcome measures  Erb’s palsy and shoulder dystocia.

Results  Mothers of children with Erb’s palsy had a higher term body mass index and more gestational diabetes than those of controls. Even cases without diabetes had higher blood glucose values after a 50-g glucose challenge than did controls. Cases had a higher birthweight and a lower ratio of head-to-thoracic circumference than controls. Shoulder dystocia occurred in 67% of cases and in 2% of controls (P= 0.001). Only 46% of labours had a completely normal dilatation pattern. In a multiple logistic regression model, variables independently associated with brachial plexus injury were long deceleration phase of labour, long second stage, high birthweight, black race, and high neonatal or maternal body mass.

Conclusions  Erb’s palsy was frequently preceded by abnormal labour and shoulder dystocia; however, a substantial proportion of cases occurred after normal labour and delivery. Predictive models will be necessary to determine to what extent careful monitoring of the terminal portion of dilatation and of fetal descent and incorporation of maternal body mass and race (all independent risk factors in this study) will help identify fetuses at risk for brachial plexus palsy.


Brachial plexus injury (BPI) is a potentially devastating complication of childbirth, with a prevalence of between 0.5 and 4.4/1000 live births.1–7 Traction injury of the nerve plexus during delivery is believed to be the primary cause, although other aetiologies have been proposed.3,8–10 Many risk factors have been identified. Most prominent among them is fetal macrosomia,5–9 but despite the knowledge of risk factors, the ability to predict BPI in an individual case is quite limited, even when multivariate models are applied.11

It is reasonable to expect that problems during delivery that are related to large fetal size would be often preceded by dysfunctional labour. In fact, few studies have addressed thoroughly the relationship between labour disorders and BPI.12–17 Most concentrated on the second stage of labour, avoiding detailed objective analysis of patterns of dilatation and descent.

Abnormal labour may serve as an indicator of the cephalopelvic relationship, which in turn is a known predictor of shoulder dystocia and BPI. In this study, we sought to identify risk factors in a series of cases of neonatal Erb’s palsy, with a focus on determining whether dysfunctional labour patterns contributed independently of macrosomia and other risk factors to the occurrence of this abnormality.


We studied 45 consecutive cases of neonatal brachial plexus palsy diagnosed after term (>37 completed gestational weeks) singleton vaginal delivery at our institution between 1 January 2000 and 31 May 2004. All diagnoses were verified within 12 hours of delivery by a paediatric neurologist, according to criteria designated by Volpe.18 All babies were otherwise normal. A control group was constituted by selecting from our hospital’s chronological birth log the two-liveborn term singleton vaginal deliveries that preceded each index case. Children with major congenital malformations were not present in the case or control group. The maternal and neonatal medical records of all subjects and controls were obtained, and data were abstracted from them manually.

Labour curves of cervical dilatation and fetal descent were created from the data available in the medical records, according to the method of Friedman19 by one investigator, who was not aware of whether the subject was a case or a control. Labour disorders were classified as prolonged latent phase, protracted dilatation or descent, arrest of dilatation or descent, prolonged deceleration phase, and failure of descent, according to standardised definitions (Appendix).19,20 For purposes of the analysis, we used the time from cervical dilatation of 8 cm until full dilatation to define the deceleration phase of active labour. Labour onset was defined as the time at which uterine contractions were noted to be regular and not more than 5 minutes apart in both spontaneous and induced labour.

Data extracted from medical records were recorded in SQL Server 2000 (Microsoft Corp., Bellevue, WA, USA). Data for analysis were exported to Stata 6 (Stata Corp., College Station, TX, USA). Initial exploratory analysis of the data used Student’s t test for continuous and the chi-square test for categorical variables. Nonparametric Wilcoxon or Kruskal–Wallis tests were used for analysis of continuous labour variables that deviated from normal distributions. Univariate logistic regression was performed to obtain unadjusted odds ratios and Wald test P-values.

Many of the recognised risk factors for Erb’s palsy are interrelated. For example, maternal diabetes, obesity, fetal macrosomia, and abnormal labour are often co-associated, and their independent effects cannot be assessed by univariate analysis. We used multiple logistic regression to control for such potential confounding. Continuous variables were transformed into quartiles (Table 1) to model the shape of the dose-response relationships between exposure levels (first through fourth quartiles) and odds of BPI, using the method of Hosmer and Lemeshow.21 The values of adjacent quartiles with comparable odds ratios were collapsed in the final multiple logistic regression model; normal and overweight subjects were collapsed into a single category (nonobese).

Table 1.  Quartiles of continuous variables from cases and controls
VariableFirst quartileSecond quartileThird quartileFourth quartile
Deceleration phase—nulliparas (minutes)<3636–7475–134≥135
Deceleration phase—multiparas (minutes)<2626–4445–69≥70
Maximum slope dilatation—nulliparas (cm/hour)<1.31.3–2.02.1–3.1≥3.2
Maximum slope dilatation—multiparas (cm/hour)<2.22.2–2.93.0–4.9≥5
Second-stage duration—nulliparas (minutes)<3131–6465–99≥100
Second-stage duration—multiparas (minutes)<88–1415–26≥27
Birthweight (g)<32703270–35443545–3944≥3945
EFW (g)<32003200–33993400–3599≥3600
HC (cm)<33.733.7–34.434.5–35.4≥35.5
TC (cm)<33.033.0–33.433.5–34.8≥34.9
HC/TC ratio<1.001.00–1.021.03–1.05≥1.06
Birth length (cm)<49.049.0–50.750.8–51.9≥52
Neonatal BMI (kg/m2)<12.912.9–14.014.1–14.8≥14.9
Gestational age (weeks)<38.938.9–39.639.7–40.2≥40.3
Maternal weight gain (kg)<1010–14.915–17.9≥18


The 45 cases, all of which were unilateral Erb’s palsies, occurred among 11 001 births. Of these, 8253 were vaginal deliveries, yielding an Erb’s palsy prevalence of 4.1/1000 births and of 5.5/1000 vaginal deliveries.

Preliminary analysis (Table 2) revealed that cases had a significantly higher body mass index (BMI) at term, were more likely to be black than Asian or Hispanic, and had a greater prevalence of gestational diabetes mellitus (GDM) compared with controls. Blood glucose values after a 50-g glucose challenge test (GCT) were significantly higher among cases, even among women without diabetes. Univariate regression confirmed that maternal obesity, gestational diabetes, and black race were associated with greater unadjusted odds of Erb’s palsy (Table 3).

Table 2.  Maternal, neonatal, and labour characteristics
 Controls (n= 90)Erb’s palsy (n= 45)P
  1. NICU, neonatal intensive care unit.

Maternal age (years)27.4 ± 5.926.8 ± 6.40.560
Nulliparous, n (%)41 (45.5)16 (35.5)0.265
Race, n (%)
Asian23 (26)7 (16)0.188
Black22 (24)20 (44)0.018
Hispanic31 (34)15 (33)0.898
White5 (6)0 (0)0.047
Other/unknown9 (10)3 (7)0.815
Gestational age (weeks)39.5 ± 1.439.5 ± 1.20.891
Maternal term BMI (kg/m2)29.9 ± 5.533.2 ± 5.20.002
Weight gain (kg)14.4 ± 6.616.0 ± 7.90.255
Gestational diabetes, n (%)5 (5.7)8 (19.1)0.022
Pregestational diabetes, n (%)01 (2.3)0.104
GCT (mg/dl)
All patients106.1 ± 26.3127.5 ± 43.30.001
Nondiabetics102.9 ± 4.2113.5 ± 4.20.04
Positive GCT/negative OGTT, n (%)3 (3.3)7 (15.5)0.007
Birthweight (g)3421 ± 4103900 ± 466<0.001
EFW (g)3335 ± 353525 ± 490.002
HC (cm)34.2 ± 0.134.9 ± 0.20.003
TC (cm)33.1 ± 0.234.6 ± 0.3<0.001
HC/TC ratio1.03 ± 0.041.01 ± 0.060.009
NICU admission, n (%)6 (6.7)9 (20)0.031
1-minute Apgar <7, n (%)3 (3.3)13 (28.9)<0.001
5-minutes Apgar <7, n (%)03 (6.7)0.010
Induced labour, n (%)12 (13.3)13 (28.9)0.032
Augmented labour, n (%)19 (24.3)10 (31.3)0.460
Epidural analgesia, n (%)30 (33)16 (36)0.848
Shoulder dystocia, n (%)2 (2)30 (67)0.001
Table 3.  Unadjusted odds ratios: maternal and neonatal variables
VariableContrast*Odds ratio95% CIP > [z]
  • *

    Rankings indicate quartiles compared. For all variables, the first quartile was the smallest or shortest; the fourth quartile was the largest or longest.

Maternal term BMIOverweight (26.1–29.0) vs normal (19.8–26.0)2.40.43, 13.540.32
Obese (>29) vs normal (19.8–26)7.41.58, 34.370.01
Maternal weight gainFourth vs first–third1.60.67, 3.770.29
ParityMultipara vs nullipara1.50.73, 3.170.27
Gestational diabetesYes vs no3.81.14, 12.520.03
RaceBlack vs Hispanic/Asian2.41.12, 5.140.02
BirthweightSecond vs first2.60.61, 11.040.20
Third vs first6.51.64, 25.760.01
Fourth vs first16.94.30, 66.64<0.001
EFWThird vs first and second2.91.18, 7.340.02
Fourth vs first and second4.41.69, 11.270.002
Birth length≥50th vs <50th percentile3.11.44, 6.460.004
Neonatal BMIThird vs first and second2.71.07, 6.780.04
Fourth vs first and second12.04.23, 34.280.003
HCThird vs first and second2.30.91, 5.590.08
Fourth vs first and second4.01.60, 9.91<0.001
TCSecond vs first2.20.06, 8.420.230
Third vs first3.91.09, 13.700.04
Fourth vs first11.23.16, 40.09<0.001
HC/TC ratioFirst vs second–fourth7.42.44, 22.24<0.001

Compared with controls, neonates with BPI (Table 2) had a higher birthweight and larger head and thoracic circumferences, as well as a lower ratio of head-to-thoracic circumference (HC/TC ratio). Cases were more likely to have had Apgar scores less than 7 at 1 and 5 minutes and were more often admitted to the neonatal intensive care unit. One case was diagnosed with hypoxic-ischaemic encephalopathy. No controls were thus affected. Significantly more cases than controls had birthweights greater than or equal to 4000 g (47 versus 9%; P < 0.001). Birthweight, estimated fetal weight (EFW), and other measures of fetal size had significantly higher unadjusted odds of Erb’s palsy (Table 3).

Severity of the BPI at initial diagnosis was categorised as mild (55%), moderate (37%), or severe (5%) according to the degree of compromised movement and tone in the affected proximal upper extremity.18 Long-term follow up was not available to determine how many were permanent.

Spontaneous vaginal delivery occurred in 98% of controls and in 93% of cases. One case was a compound presentation. Two controls and three cases were delivered operatively by vacuum extractor or forceps. In two other cases, vacuum delivery failed, but spontaneous vaginal delivery eventually occurred.

Shoulder dystocia was recorded to have occurred in 30 cases (67%) and two controls (2%) (P= 0.001). Among the cases, Erb’s palsy was moderate or severe in 51% of those with shoulder dystocia and in only 18% of those with no documented problems during delivery (P= 0.01).

Labours among BPI cases differed in several respects from controls. Among the cases, labour was induced more often than in controls (28.9 versus 13.3%; P= 0.032). Spontaneous labours tended to be augmented with oxytocin more often in cases, but this difference was not significant. Epidural anaesthesia was used with equivalent frequencies in both groups. Only 46% of cases had completely normal dilatation, as compared with 67% of controls (P= 0.042); only 26% of Erb’s palsy cases had completely normal dilatation and descent patterns. Labours were overall longer in cases than in controls, regardless of parity. The longer first stage (a trend that was significant only in nulliparas) was accounted for by the relative prolongation of the active phase; latent phase durations were similar in both cases and controls. The second stage was also longer in cases (significantly so in multiparas). The relatively longer labours and lower median rates of dilatation in BPI cases reflect their higher incidence of first stage abnormalities. The unadjusted odds of Erb’s palsy were increased by slow rates of active phase dilatation, long deceleration phase, long second stage, active phase abnormalities (protraction or arrest disorders), or a prolonged deceleration phase (Table 4). Although precipitate descent occurred in nearly 40% of BPI cases, it was observed with a similar frequency in controls. Eleven of 14 children with precipitate descent had shoulder dystocia (79%). We found no increased frequency of any labour disorder or any difference in the quantifiable aspects of the labour curves among BPI cases with and without shoulder dystocia.

Table 4.  Unadjusted odds ratios: labour-related variables
VariableContrast*Odds ratio95% CIP > [z]
  • *

    Rankings indicate quartiles compared. For all variables, the first quartile was the smallest or shortest; the fourth quartile was the largest or longest.

Maximum slope dilatationFirst vs second–fourth2.41.04, 5.750.04
Deceleration phase durationThird vs First and second2.30.90, 6.00.08
Fourth vs First and second3.31.34, 8.080.01
Second-stage durationFourth vs First–third2.31.01, 5.150.05
Active phase protraction or arrest disorderYes vs no2.81.15, 6.720.02
Prolonged deceleration phaseYes vs no2.20.75, 6.260.15

When variables associated with Erb’s palsy based on their unadjusted odds ratios were entered into a multiple logistic regression model, those that appeared to be independently associated with BPI were long deceleration phase, long second stage, high birthweight, black race, and high neonatal or maternal body mass (Table 5).

Table 5.  Multiple logistic regression model
VariableContrast*Odds ratio95% CIP > [z]
  • *

    Rankings indicate quartiles compared. For all variables, the first quartile was the smallest or shortest; the fourth quartile was the largest or longest.

Deceleration phase durationFourth vs First–third5.851.67, 20.490.006
Second-stage durationFourth vs First–third3.070.96, 9.820.058
BirthweightFourth vs First–third3.310.77, 14.230.107
Neonatal BMIFourth vs First–third7.791.53, 39.660.013
Maternal BMIObese vs nonobese4.661.42, 15.340.011
RaceBlack vs Hispanic/Asian2.600.77, 8.790.125


Our primary interest was to determine whether abnormal labour was associated with BPI in our series. Fewer than half the BPI cases were preceded by completely normal dilatation, emphasising that BPI is often, but not always, preceded by dysfunctional labour. Active phase abnormalities predominated among the mothers with dysfunctional labours. Most importantly, a long deceleration phase (expressed by the fourth quartiles of our data as ≥135 or ≥70 minutes in nulliparas and multiparas, respectively) increased the adjusted odds of BPI almost six times.

There was not an excess of arrest or protraction disorders of descent in our BPI cases. This should not be interpreted to indicate that aberrant descent patterns did not presage shoulder dystocia or BPI; it may instead reflect the fact that most patients with these labour dysfunctions were delivered by caesarean before there was an opportunity for injury to occur at vaginal delivery. Moreover, the long (fourth quartile) second stage in our sample was associated with increased unadjusted and adjusted odds of Erb’s palsy.

Other investigators have specifically addressed the question of whether dysfunctional labour is a harbinger of BPI or shoulder dystocia. One group found22 that 91% of permanent BPI followed normal labours; but their quantitative labour data were not stratified by parity, and no control group was used. Other authors have also concluded that BPI is not predictable from patterns of cervical dilatation or fetal descent.15

Similarly, Lurie et al.12 found no difference in rates of dilatation or second-stage duration in cases with shoulder dystocia and concluded that protracted labour was not a risk factor for it. However, almost 35% of their shoulder dystocia cases were delivered by vacuum extraction, and labour data were not classified by parity. McFarland et al.13 found no difference in first- or second-stage abnormalities in cases with shoulder dystocia.

Gross et al.11 showed in a multivariate analysis that a prolonged deceleration phase and long second stage contributed to BPI risk, but these were not strong predictors. Several other studies suggested a long second stage as a risk factor for shoulder dystocia14,16,22–24 and for BPI.6,11 Our own results support the association of long second stage and deceleration phase durations with BPI independent of the effects of macrosomia, diabetes, and other factors.

About one-third of our BPI cases occurred in the absence of documented shoulder dystocia. This finding is consistent with those of previous studies1,10,18,25 in which 15–45% of BPI were not preceded by any reported difficulty with delivery. A mathematical model26 suggested that endogenous forces during delivery could be sufficient to injure the nerves in certain situations. It has been postulated27,28 that BPI without shoulder dystocia is an entity distinct from that which occurs with it; the rate of resolution in cases of Erb’s palsy without identified shoulder dystocia was significantly lower in one series,27 but it was higher in another.28

BPI cases with and without shoulder dystocia were comparable in parity, birthweight, incidence of diabetes, newborn body proportions, and dysfunctional labour patterns. The sample size in our study was, however, too small to do meaningful multiple regression modelling to search for independent contributors to BPI not preceded by difficult shoulder delivery.

As noted, babies were on average nearly 500 g larger in BPI cases than in controls, and the HC/TC ratio was significantly lower, suggesting that disproportionate growth of the trunk was a risk factor. This asymmetry, well known to exist in fetuses of diabetic women,29,30 persisted in our series even after the diabetic subjects were excluded from analysis.

Estimated and actual fetal weights were higher in the cases, as would be expected; but the obstetrician underestimated the actual weight in the cases by nearly 400 g and by less than 100 g in the controls. This probably reflects the technical difficulties of assessing weights (clinically or by ultrasound) when the fetal and maternal size are large.

Many of the factors found to be associated with Erb’s palsy in previous studies were confirmed in our sample; but some new insights may be inferred from the data. In the univariate analysis, obese women and women with diabetes who had a high EFW were found significantly more often in the BPI group. All are factors associated with large birthweight and are recognised risk factors for shoulder dystocia and for BPI. Our findings regarding the association of mild glucose intolerance among women without diabetes and the odds of BPI were unexpected, as was the strong association of the deceleration phase of labour with risk.

The mean GCT values in the women without diabetes were 10% higher in the BPI group, and women who had a positive (>135 mg/dl) 50 g GCT followed by a negative oral glucose tolerance test (OGTT) were five times as common in cases than in controls. These observations support the view that there is a continuum of glucose–insulin metabolic effects on fetal growth that may contribute to risks during delivery.

Of interest is that GDM did not have an independent effect on Erb’s palsy in our series. This is not to say that BPI is unrelated to diabetes. Rather, the effects of diabetes seemed to be accounted for by maternal and fetal size and the lengths of the deceleration phase and the second stage of labour. GDM is, nevertheless, often coupled with increased maternal and neonatal body mass and disproportionate growth of the fetal trunk.29–31 Because fetal size and proportion are not known with accuracy before delivery, GDM should be considered an important a priori risk factor for BPI.

Most of our cases occurred in the anterior arm; but five (11%) were found in the posterior presenting arm. Only one of those cases was accompanied by shoulder dystocia, and none had any delay in dilatation. One case was a compound presentation, with the anterior arm prolapsed beside the head. Two of the five cases had precipitate descent. Our observations are consistent with those of Gurewitsch et al.,28 who showed BPI without shoulder dystocia was more likely to occur in the posterior arm.

Previous studies32 have implicated excessively rapid descent in the causation of shoulder injury. We could not document this association in our series. Instrumental delivery, especially from the midpelvis, has also been identified as a risk factor for shoulder dystocia and BPI.2,5,6,23,33 The number of vacuum and forceps deliveries in our sample was too small to assess this association.

The substantial independent effect of a long deceleration phase of labour on the odds of BPI that we observed appears important. It was not assessed as a variable of interest in most previous analyses, but its potentially adverse portent for shoulder injury was noted in two studies.11,34 In fact, prolongation of the terminal aspect of dilatation appears to be a harbinger of problems during the second stage, including the need for caesarean,35 as well as shoulder dystocia and BPI. Our data show that when the deceleration phase is long (even before it exceeds the 95th percentile limits described by Friedman19), concern for the development of BPI should be high, especially if other risk factors are present. Duration of the deceleration phase was not determinable (because of insufficient cervical examinations) in 4% of cases and in 11% of controls.

Black race had a modest influence on BPI, even after controlling for labour-related variables, maternal BMI, and fetal size. Race has not been addressed in previous studies as a possible independent risk factor. There are racial differences in pelvic architecture and in neonatal body proportions that could explain our observations.36–38

A potential bias could exist in our sample because it comes from an unusual cluster of cases over a relatively short interval. The incidence of Erb’s palsy during the 2 year period after the cluster was considerably lower than that of the study period (3.4 versus 4.1/1000 births). The reason for this is unexplained. When attention is focused on a problem by a few cases, increased vigilance sometimes results in the diagnosis of minor degrees of pathology that might otherwise have gone unnoticed. It is impossible to discern whether this occurred in our series. If it did, our overall conclusions would likely not be affected because there is no evidence that mild cases of BPI are caused by mechanisms different from those that produce more significant palsies, and the distribution of severity in our sample is similar to others reported in the literature.

Another possible source of bias in our analysis relates to our use of quartiles to classify our data. While this approach was well suited to our sample size and experimental question,21 it could have resulted in blunting the influence of some variables. For example, because quartiles impose heterogeneity in the highest and lowest birthweight groups, the effect of very high or very low weight will likely be diluted in the analysis.

BPI can be caused by inappropriate delivery technique. We cannot know to what extent poor obstetric technique contributed to our cases because we could not judge retrospectively the quality of delivery efforts, a problem common to most series of shoulder dystocia and BPI cases.


Our analysis of a series of Erb’s palsy cases identified their association with large maternal and fetal body mass and disproportionate growth of the fetal trunk relative to the head. While about one-third of BPI cases occurred in normal weight babies after uneventful labours, thorough graphic analysis of labour showed that BPI was frequently preceded by dysfunctional labour patterns. Multiple logistic regression analysis identified a long deceleration phase and a long second stage as important independent contributors. In other words, even after controlling for other identifiable risk factors, certain aspects of labour served as important clinical risk indicators for BPI. Our results suggest that careful monitoring of the pelvic division of labour (the deceleration phase of dilatation and fetal descent) may help identify fetuses at risk for Erb’s palsy. It should, however, be borne in mind that our study had a descriptive or explanatory design. While it identified factors that were associated with the development of Erb’s palsy, it did not address directly the question of whether the accuracy of BPI prediction is improved by taking into account labour-related variables. That awaits the development and testing of predictive models that use only variables known before the injury occurs.


Definition of terms.19,20

Labour Curve: A plot of the relationships among cervical dilatation, fetal descent, and time elapsed in labour.

Latent Phase: The period of labour from the onset of labour until the acceleration in cervical dilatation seen at the onset of active phase.

Active Phase: The period of labour that follows the latent phase from the acceleration in cervical dilatation until full cervical dilatation.

Deceleration Phase: The terminal portion of active phase dilatation when the cervix is approaching the widest diameter of the presenting part. Generally this occurs between about 8 cm and full dilatation.

Prolonged Latent Phase: Exceeds 20 hours in nulliparas or 14 hours in multiparas

Protracted Active Phase: Linear dilatation in active phase less than 1.2 cm/hour in nulliparas or less than 1.5 cm/hour in multiparas.

Arrest of Dilatation: No progress in dilatation for 2 hours in active phase labour.

Precipitate Dilatation: Active phase dilatation greater than 5 cm/hour in nulliparas or greater than 10 cm/hour in multiparas.

Prolonged Deceleration Phase: Exceeds 3 hours in nulliparas or 1 hour in multiparas.

Failure of Descent: No descent of head from early labour beyond the onset of deceleration phase or second stage.

Arrest of Descent: No progress in second stage for 1 hour.

Protracted Descent: Head descent in second stage less than 1 cm/hour in nulliparas or less than 2 cm/hour in multiparas.

Precipitate Descent: Head descent in second stage greater than 5 cm/hour in nulliparas or greater than 10 cm/hour in multiparas.