Risk factors for neonatal brachial plexus palsy: a systematic review and meta-analysis

Neonatal brachial plexus palsy (NBPP) is reported worldwide in 0.1 to 8.1 per 1000 live births.1-6 Incidence rates vary with study type and the availability of maternal and fetal care.7-9 NBPP is the result of a closed nerve stretch injury to the brachial plexus, mostly occurring during labour. The mechanisms of injury include maternal, obstetric, and infant factors that apply traction on the anatomically vulnerable plexus.

Early management includes parental counselling, family support, splinting, and appropriate and supervised rehabilitation. Neurosurgical intervention is usually undertaken at 3 to 6 months of age in children who have shown little or no significant improvement in the affected muscle groups.2, 10-15 Depending on the long-term clinical course, secondary surgery may be indicated later in life.

The few studies on NBPP prognosis not hampered by selection bias suggest that residual deficits are estimated at 20% to 35%.3, 5, 16-20 This finding is at odds with the previous optimistic view of full recovery in over 90% of affected children.21-26 After neonatal occurrence, the upper limb paresis affects psychomotor development, bone growth, and joint development. Deformities occur and may cause painful arthrosis in adults. Children with incomplete recovery have considerable risk for long-term functional limitations, with financial burden, restricted daily life activities, limited participation, and important overall quality of life implications.16, 27-43

The potential impact of NBPP on the child’s physical and psychological development, and their socio-economic future, highlights the need for the development of preventive strategies. Early identification of modifiable risk factors is therefore imperative. Appropriate management or avoidance of risk factors might contribute to a reduction in NBPP morbidity.

The aim of the present study was to determine and unravel the significant risk factors for NBPP to enable individualized risk assessment and proper counselling of individual women at risk. We conducted a meta-analysis by combining the results from all available studies to: (1) assess the five most frequently reported significant risk factors; (2) calculate their pooled odds ratios; and (3) assess the heterogeneity among studies. In addition, we assessed the incidence and its evolution.

Method

The Cochrane Database of Systematic Reviews was searched to ensure a similar review had not been undertaken. The Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) statement,44 MOOSE guidelines,45 and Grant’s review46 were followed to report this meta-analysis. The protocol was accepted for registration in PROSPERO on 14th November 2017 (CRD42017076642).

Results

Meta-analysis

Twenty-one studies represented a total sample size of 29 419 037. The five most frequently reported significant risk factors were used for meta-analysis: shoulder dystocia, birthweight, (gestational) diabetes, instrumental delivery, and breech delivery. Shoulder dystocia showed the highest association with NBPP (odds ratio [OR] 115.27; 95% confidence interval [CI] 81.35–163.35; 12 studies; n=25 825 074)7, 15, 62, 65, 71, 73, 75, 77-81 but with considerable across-study heterogeneity (I²=92%) (Fig. S3a, online supporting information). A sensitivity analysis with exclusion of three studies65, 71, 78 resulted in low heterogeneity (I²=19%; n=24 566 515) with similar strength of association (OR 113.58; 95% CI 95.35–135.29).

Macrosomia (≥4000g) was the second-highest associated factor (OR 9.75; 95% CI 8.29–11.46; 13 studies; n=2 435 212)15, 64, 67, 69-72, 74, 76-80 (Fig. S3b). The substantial heterogeneity (I²=70%) was reduced to 34% (n=2 955 694) by excluding two studies15, 70 and resulted in an odds ratio of 9.14% and 95% confidence interval 8.28 to 10.10. When subdividing the birthweight into three subcategories (4000–4499g, 4500–4999g, and ≥5000g), compared with normal birthweight of not more than 3999g, the odds ratios were respectively 6.32 (95% CI 5.48–7.29; six studies;64, 67, 72, 76-78 I²=59%), 20.77 (95% CI 16.86–25.58; five studies;64, 67, 72, 77, 78 I²=75%), and 55.21 (95% CI 49.7–61.35; five studies;64, 67, 72, 77, 78 I²=0%).

(Gestational) diabetes as a risk factor with an odds ratio of 5.33 (95% CI 3.77–7.55; I²=59%; n=1 651 281) was explored in 10 studies15, 65, 71, 72, 74, 75, 78-81 (Fig. S3c). Excluding two studies74, 75 reduced heterogeneity to 23% (n=1 274 770) and an odds ratio of 4.21 (95% CI 3.12–5.68).

With an odds ratio of 3.8 (95% CI 2.77–5.23; seven studies;7, 15, 67, 71, 72, 77, 78 I²=77%; n=1 849 398) instrumental delivery, containing vacuum and/or forceps delivery, occurred as a risk factor for NBPP (Fig. S3d). Sensitivity analysis affirmed three studies15, 67, 78 accountable, with an odds ratio of 2.50 (95% CI 1.81–3.45; n=25 240; I²=0%) after exclusion. For exploratory reasons a separate pooled odds ratio was calculated for vacuum (OR 6.03; 95% CI 3.61–10.07; seven studies;62, 65, 68, 75, 76, 79, 80 n=25 666 748; I²=96%) and forceps delivery (OR 4.97; 95% CI 1.95–12.66; three studies;62, 65, 68 n=25 284 297; I²=98%).

The last significant risk factor was breech delivery, with a pooled odds ratio of 2.49 (95% CI 1.67–3.7; 11 studies;7, 62, 63, 65-67, 72, 76-79 n=26 780 930; I²=70%) (Fig. S3e). The substantial heterogeneity reduced by excluding three studies62, 78, 79 (n=1 401 993; I²=0%), resulting in a higher odds ratio of 3.35 (95% CI 2.39–4.70).

Caesarean section emerged as a protective factor for NBPP, with a pooled odds ratio of 0.13 (95% CI 0.11–0.16; 11 studies;7, 62, 64, 65, 67, 68, 72, 74, 75, 77, 78 n=28 530 257; I²=41). Heterogeneity was moderate (Fig. S3f). The heterogeneity reduced to 0% by excluding one study62 (n=4 370 832; I²=0%; OR 0.15, 95% CI 0.13–0.17).

The five risk factors demonstrate an asymmetrical funnel plot, which could assume publication bias. The limited number of studies per risk factor, however, inhibits an appropriate causal interpretation (Fig. 1).59 According to the assessment by the GRADE (Table 1), Caesarean section presented high-quality evidence, shoulder dystocia, and birthweight moderate quality, whereas (gestational) diabetes and instrumental delivery showed low-quality evidence. Breech delivery had a very low confidence rating. Strengths of the studies responsible for upgrading were the large magnitude of effect. An important limitation leading to downgrading was inconsistency of results for each outcome and additionally imprecision for results of breech delivery. All findings are summarized in Table 2.

Table 1. Grading of Recommendations Assessment, Development and Evaluation (GRADE) assessment of risk factors for neonatal brachial plexus palsy

	Start	Downgrade					Upgrade			End
		Limitations (RoB)	Inconsistency of results	Indirectness of evidence	Imprecision	Publication bias	Large magnitude of an effect	Dose–response gradient	Effect of plausible residual cofounding
Shoulder dystocia	LQ	Low	↓ −1	OK	OK	OK	↑ +2	/	OK	MQ
(Gestational) diabetes	LQ	Low	↓ −1	OK	OK	OK	↑ +1	/	OK	LQ
Birthweight	LQ	Low	↓ −1	OK	OK	OK	↑ +2	/	OK	MQ
Instrumental delivery	LQ	Low	↓ −1	OK	OK	OK	↑ +1	/	OK	LQ
Breech delivery	LQ	Low	↓ −1	OK	↓ −1	OK	↑ +1	/	OK	VLQ
Caesarean section	LQ	Low	OK	OK	OK	OK	↑ +2	/	OK	HQ

• RoB, Cochrane risk-of-bias tool for randomized studies; LQ, low quality; ↓, lower it down; OK, quality met criteria; ↑, grade it up; /, not applicable; MQ, moderate quality; VLQ, very low quality; HQ, high quality.

Table 2. Summary of findings for the five most significant risk factors

Outcomea	Number of studiesb	Total population	Number of cases within risk group	Number of cases in total population	OR (95% CI) for NBPP	I2 (%)	Strength of recommendationc
Shoulder dystocia	12	25 825 074	6563	33 984	115.27 (81.35–163.35)	92	Moderate
Macrosomia	13	2 975 874	3896	5892	9.75 (8.29–11.46)	70	Low
(Gestational) diabetes	10	1 651 281	175	3209	5.33 (3.77–7.55)	59	Moderate
Instrumental delivery	7	1 849 398	824	3112	3.8 (2.77–5.23)	77	Low
Breech delivery	11	26 780 930	183	34 960	2.49 (1.67–3.7)	70	Very low

• ^a Risk factors implemented in the meta-analysis.
• ^b Included in the meta-analysis.
• ^c Based on the Grading of Recommendations Assessment, Development and Evaluation (GRADE) quality of evidence assessment. OR, odds ratio; CI, confidence interval; NBPP, neonatal brachial plexus palsy.

The pooled overall incidence of NBPP was 1.74 per 1000 live births (95% CI 1.56–1.94) (Fig. S4, online supporting information). The subgroup meta-analysis demonstrated a significant difference (p=0.013) in the pooled proportion of the event, between the three timeframes (1987–1995, 1996–2005, and 2006–2015). The forest plot showed highly heterogeneous individual study results within each timeframe. The estimated pooled proportions were 0.0021 (95% CI 0.0018–0.0025) for the first timeframe, 0.0019 (95% CI 0.0014–0.0025) for the second, and 0.0012 (95% CI 0.0009–0.0017) for the third. Although a considerable overlap was found between the first and second, and second and third, confidence intervals, a decrease in proportion could be concluded between the first and third timeframes.

Discussion

Consistent with previous research,83-86 shoulder dystocia was found to be the major risk factor for NBPP. Three studies were responsible for the considerable heterogeneity.65, 71, 78 First, we assigned the shoulder dystocia incidence variety as a possible source: shoulder dystocia is imprecisely coded, often underreported, and even unrecognized. The prospective design of Backe et al.65 and perhaps a more rigorous reporting of Backe et al.65 and Ouzounian et al.71 might explain their relatively high incidence of shoulder dystocia compared with previous published research. In contrast, the remarkably low incidence in the large retrospective case–control study of Mollberg et al.78 is explained by the authors as a possible underreporting of shoulder dystocia in Sweden. Second, an observable lower association of NBPP and shoulder dystocia was noted in two studies,60, 67 which could result from their immediate resort to the recommended shoulder dystocia delivery technique73, 87 in the case of shoulder dystocia.

Macrosomia, confirmed only after delivery of the neonate, is the second risk factor. The 70% heterogeneity was mostly caused by two studies.15, 70 As both were prospectively performed in Asia, their remarkably higher incidence of macrosomia might be explained by population bias. Sixty per cent of infants with macrosomia in the study by Najafian and Cheraghi70 were of Arab ethnicity. Dawodu et al.15 revealed a higher frequency of maternal diabetes, shoulder dystocia, and macrosomia in Arab versus Western populations. Both studies had a strong association of NBPP and macrosomia. Interestingly, only 2% of infants with macrosomia in the study by Najafian and Cheragi70 were associated with NBPP compared with 70% in that by Dawodu et al.15 Possibly, their birthweight distribution was different, with those in the study by Najafian and Cheragi70 tending to have the lower weight range: 85% birthweight between 4000 and 4449g, 13% between 4500 and 4999g, and only 2% having at least 5000g. Obstetric practices might also differ: all cases of NBPP in the study by Dawod et al.15 were delivered vaginally. Therefore, geographical differences, study design, and population bias might have been responsible for the heterogeneity. In agreement with previous studies,88-90 calculated odds ratios of NBPP expand powerfully with increasing birthweight: 6.32 for birthweight group 4000 to 4499g, 20.77 for 4500 to 4999g, and particularly 55.21 for those with a birthweight of at least 5000g. Birthweight of at least 5000g seems to be a commonly found risk factor, with I²=0%.

In accordance with previous reports,63, 89, 91-93 maternal diabetes is the third significant risk factor. Two studies,74, 75 showing higher odds for NBPP in the diabetic group, were responsible for the 59% heterogeneity. As the incidence of NBPP, respectively diabetes, in these studies was comparable with the included studies, other factors should be accountable. We believe that birthweight, ratio of head-to-abdominal circumference of the fetus, maternal weight, and age difference between diabetic versus non-diabetic groups are important characteristics possibly responsible for population bias. Disproportionate growth of trunk, measured as a smaller ratio of head circumference to abdominal circumference, is well-known in women with diabetes.94-96 Unfortunately, the included studies published none of these data. In the large population-based study by Freeman et al.,75 61% of diabetic mothers underwent Caesarean section, a finding also reported in other studies.97-99 The reason for Caesarean section was not mentioned but, knowing the diabetes diagnosis, the obstetrician may have preferred Caesarean section for fear of shoulder dystocia or NBPP.

Instrumental delivery (vacuum extraction and/or forceps delivery) augments the risk for NBPP 3.8-fold, with vacuum extraction showing an odds ratio of 6.03. The 77% heterogeneity was attributable to three studies.15, 67, 78 Different reporting numbers, depending on the use of maternal versus neonatal records, result in a probable source of bias. Instrumental delivery is often noted in the mother’s record as a procedure and only in the newborn infant’s record as a diagnosis if its use resulted in an injury. Possible population bias following different prevalence of other risk factors such as diabetes, macrosomia, shoulder dystocia, or increased length of second stage of labour in the studied population might cause heterogeneity. The different studies, despite reporting adjusted results, did not examine the same confounders. In addition, no information could be extracted on the sequential use of instruments, and certainly temporal trends are important.

Our meta-analysis confirms the previously reported NBPP risk of breech delivery.100-105 NBPP incidence and severity are increased with the tendency to develop more upper palsy and a higher percentage of bilateral palsy.106 Five out of 11 articles7, 65, 71, 76, 79 had a wide confidence interval, mainly because they included relatively few patients. Three studies were liable for the 70% heterogeneity.62, 78, 79 The contribution of Okby and Sheiner79 of 21% was possibly due to the presence of only one breech-delivered infant with NBPP. The low risk factor prevalence of 0.12% in the study by Abzug et al.62 might be explained by a higher rate of Caesarean section of 27%. In contrast, the study by Mollberg et al.78 presented 2.51% breech deliveries with a small complication rate, which could be ascribed to a more homogenous and better selection for trial of vaginal breech delivery.

Caesarean section seems to be a protective factor, with an odds ratio of 0.13. Low heterogeneity was reduced to 0% by excluding one study.62 The high prevalence of Caesarean section in the study by Abzug et al.62 might be the contributor.

The pooled overall incidence of NBPP was 1.74 per 1000 live births. The subgroup meta-analysis demonstrates a significant decrease of incidence over time (1987–1995, 1996–2005, and 2006–2015). The heterogeneity could be partly explained by the difference in type of obstetric care, in Caesarean section rate, and in average birthweight of neonates in different geographical regions.

In practice, clinicians are confronted with a combination of different risk factors, which are interrelated (Fig. 2). The risk of shoulder dystocia is highly correlated with fetal macrosomia,75, 107 a history of previous shoulder dystocia7, 108 or macrosomia, maternal diabetes and obesity,109 induction of labour,73 abnormalities of labour, the number and type of manoeuvres used,73 and instrumental delivery.75, 110-112 Infants with macrosomia who are vaginally delivered are at risk of shoulder dystocia, with the highest association in the birthweight group greater than 5000g.78 Macrosomia is an important risk factor for instrumental delivery and increases the risk of Caesarean section. Most studies confirm the higher incidence of maternal obesity, macrosomia, disproportional fetal growth, abnormal labour, and shoulder dystocia among diabetic pregnancies.96, 113, 114 Other major risk factors for macrosomia are increased maternal age (>35y) and obesity, a positive history of previous macrosomia, prolonged pregnancy, and multiparity.64, 70, 113, 115, 116 Given the potential for severe neonatal and long-term morbidity of NBPP, as well as the significant medico-legal implications, it is essential to recognize patients who are at high risk (for instance by using a risk factor scoring system). However, prevention of NBPP is a challenge owing to the poor predictive value of the risk factors. Estimation of fetal weight, often used as a measure of risk stratification, remains, even with modern ultrasound equipment and in experienced hands, an inaccurate task.116 A sensitivity and specificity of respectively 60% and 90% for identifying fetuses of at least 4000g makes it insufficiently reliable.86, 90 Therefore, the target in antenatal prevention of NBPP should be on nutrition and physical activity guidance,117, 118 strict glycaemic and weight control in case of (gestational) diabetes, and appropriate counselling of women at high risk. Counselling should focus on the risks and alternatives to vaginal delivery.

Intrapartum risk management should include presence of consistent intrapartum guidelines, ensuring availability of regularly and well-trained staff at delivery and minimizing manoeuvres with excessive traction on the fetal neck. For macrosomia, recent research suggests that early-term induction significantly reduces the risk of shoulder dystocia.119-121 The clear association between NBPP and macrosomia in diabetic pregnancies, with higher occurrence of total palsies114 and therefore more permanent disability,116 is a major concern in the prevention of NBPP. Elective Caesarean section should be offered where there is an estimated fetal birthweight greater than 5000g in non-diabetic pregnancies90, 122 and greater than 4500g in those with diabetes according to guidelines.83, 90 In all others, a trial of labour is recommended. This requires preparedness for operative delivery, shoulder dystocia, and newborn asphyxia.114 In the case of shoulder dystocia, the risk for severity of NBPP is critically linked to a timely recognition, and increases with the number and types of manoeuvres required.73 Several authors123-127 reported evidence that a systematic approach with simulation training of specific manoeuvres can reduce cases of NBPP significantly.127, 128 For mothers with a previous shoulder dystocia history, an estimated fetal weight less than the previous dystocia delivery, or a lack of history of permanent NBPP, a trial of vaginal delivery may be reasonable.109

Independent or sequential use of forceps and vacuum extractor greatly increases the risk of NBPP, especially among obese and diabetic pregnancies.78, 129 The choice between vacuum and forceps has recently shifted, with a preference for vacuum as the instrument of first choice.130-133 Caesarean section performed after failed instrumental delivery carries increased risk.29 Therefore, obstetricians should aim to complete all operative vaginal delivery safely with a single instrument.134

The routine to deliver almost all term breech cases by elective Caesarean section is a continuing debate. In fact, there are only three randomized controlled trials that have investigated neonatal outcomes in term breech by mode of delivery.135-137 A Cochrane review138 reported a reduced perinatal or neonatal mortality among singleton infants delivered by a planned Caesarean section. However, neither elective Caesarean section nor vaginal delivery for breech presentation is risk free.105 Eligibility criteria, balancing all risks and benefits, for vaginal breech delivery should be set at the national level to guide best practice.

Although Caesarean section seems to be protective, it does not totally prevent NBPP68, 139 and carries significant neonatal and maternal morbidities (an increased risk for repeat Caesarean section, abnormal placentation, uterine rupture, and ectopic pregnancy).140, 141 This opens the commonly raised issue of the cost:benefit ratio.142-144 For the USA population, more than 1000 elective Caesarean sections, costing US$4 million to US$8 million, would be required to prevent one case of NBPP.75, 86 By comparison, the lifetime costs for a case of NBPP, excluding potential loss in productivity and earning capacity, is estimated at more than US$1 million.143 Caesarean section is only indicated in selected cases: women with previous children with permanent NBPP,23, 145 macrosomic pregnancies complicated by diabetes,74, 142, 143 and a high global risk identified by the obstetrician.

The decreasing incidence might be the result of augmented awareness of the problem and improved obstetric techniques and strategies. This encourages further research to determine predictable and modifiable risk factors. A future worldwide meta-analysis of incidence could be important to evaluate geographical differences and their influence on the risk of NBPP.

Conclusion

Shoulder dystocia, macrosomia, maternal diabetes, instrumental delivery, and breech delivery are the main risk factors for NBPP, with shoulder dystocia presenting the highest risk. Prevention remains difficult owing to the unpredictability of these factors and their often labour-relatedness. Moreover, many risk factors are interrelated. Caesarean section is a protective factor for NBPP. The incidence of NBPP has decreased over time, possibly as a result of increased awareness of the risk and improved obstetric strategies. In the view of its lifelong impact, the risk for NBPP, its severity, and morbidity should be further lowered. The focus needs to be on risk stratification, antenatal counselling, enhanced labour surveillance, and simulation training. This study highlights the need for further research to determine predictable and modifiable risk factors with emphasis on their relationship. As this meta-analysis integrates results of known risk factors, future research should also focus on undiscovered ones.