Association of Parity and Previous Birth Outcome With Brachial Plexus Birth Injury Risk

Abstract

Objective:

To evaluate the association of maternal delivery history with a brachial plexus birth injury (BPBI) risk in subsequent deliveries, and to estimate the effect of subsequent delivery method on BPBI risk.

Methods:

We conducted a retrospective cohort study of all livebirth deliveries occurring in California-licensed hospitals from 1996-2012. The primary outcome was recurrent BPBI in a subsequent pregnancy. The exposure was prior delivery history (parity, shoulder dystocia in a previous delivery, or previously delivering an infant with BPBI). Multiple logistic regression was used to model adjusted associations of prior delivery history with BPBI in a subsequent pregnancy. The adjusted risk (AR) and adjusted risk difference (ARD) for BPBI between vaginal and cesarean delivery in subsequent pregnancies were determined, stratified by prior delivery history, and the number of cesarean deliveries needed to prevent one BPBI was determined.

Results:

Of 6,286,324 infants delivered by 4,104,825 individuals, 7,762 (0.12%) were diagnosed with a BPBI. Higher parity was associated with a 5.7% decrease in BPBI risk with each subsequent delivery (aOR 0.94, 95% CI 0.92, 0.97). Shoulder dystocia or BPBI in a previous delivery were associated with 5-fold (0.58 vs 0.11%; aOR 5.39, 95% CI 4.10, 7.08) and 17-fold increases (1.58% vs. 0.11%; aOR=17.22, 95% CI 13.31, 22.27) in BPBI risk, respectively. Among individuals with a history of delivering an infant with a BPBI, cesarean delivery was associated with a 73.0% decrease in BPBI risk (0.60% vs 2.21%; aOR=0.27, 95% CI 0.13, 0.55), compared with an 87.9% decrease in BPBI risk (0.02 vs 0.15%; aOR =0.12, 95% CI 0.10, 0.15) in individuals without this history. Among individuals with a previous history of BPBI, 48.1 cesarean deliveries are needed to prevent one BPBI.

Conclusions:

Parity, previous shoulder dystocia, and previously delivering a BPBI infant are associated with future BPBI risk. These factors are identifiable prenatally and can inform discussions with pregnant individuals regarding BPBI risk and planned mode of delivery.

Precis

Parity, previous shoulder dystocia, and previously delivering an infant with a brachial plexus injury are associated with future brachial plexus birth injury risk. Cesarean delivery partially mitigates this risk.

Introduction

Brachial plexus birth injury (BPBI) is a birth condition resulting from traction on the nerve roots of the brachial plexus during labor and delivery. It is among the most common birth injuries, occurring in approximately 1.5 per 1000 livebirths,^1–3 and presents as upper extremity weakness or paralysis. Although the majority of affected infants recover spontaneously, up to 30% have incomplete or absent neurologic recovery (“permanent BPBI”), ^4–8 with persistent motor weakness, sensory deficits^4–9 and impaired musculoskeletal development, including joint contractures and skeletal dysplasia,^10,11 due to chronic denervation during growth. Long-term, children may experience functional limitations,^12–14 pain,^15–17 and psychosocial impairments,^12,14,18 persisting into adulthood.¹⁹

It is often assumed that pregnancy history (parity, previous shoulder dystocia, or history of BPBI) is associated with BPBI risk; however, this has not been definitively established. The evidence for the effect of parity on BPBI risk is inconsistent, with some prior studies demonstrating an increased BPBI risk with higher parity, while others demonstrate a decreased risk.^20–29 Similarly, there are only small case series evaluating the effect of shoulder dystocia in a prior delivery or previous BPBI on BPBI risk in subsequent pregnancies.^24,30,31 Although pregnancy history is a potentially potent source of information regarding risk of BPBI, an evidence gap remains regarding the relationship between prior pregnancy history and future BPBI risk. Understanding this relationship could inform preventive strategies and ultimately, provide insight into the mechanism by which BPBIs occur.

The purpose of this study was to evaluate if previous pregnancy history, including parity, history of shoulder dystocia, or BPBI in a previous pregnancy, was associated with BPBI risk in subsequent pregnancies. Additionally, we sought to estimate the reduction in BPBI risk with cesarean delivery in subsequent pregnancies.

Methods

This study was approved by the University of California Davis Institutional Review Board and the California Committee for the Protection of Human Subjects. Using the Linked Birth Files from California’s Department of Health Care Access and Information (HCAI) created to facilitate research on pregnancy outcomes,^32–38 we conducted a retrospective cohort study of all maternal-livebirth infant pairs whose childbirth occurred in a non-federal California-licensed hospital from 1996 to 2012, accounting for 98% of California births.³⁸ The data were compiled by HCAI from the California Inpatient Discharge Dataset, Birth Certificate Dataset, and the Vital Statistics Birth Cohort File. Data included maternal demographic and health data for 9 months prior to and 12 months after delivery, linked to the infant’s demographic and health data from birth through the first year of life.^32–36 Individuals who delivered more than once during the study interval were identified by a unique identifier. Previous studies report the accuracy of this dataset for maternal factors, intrapartum events, and obstetric complications.^39–41

The study cohort included all maternal-infant pairs in which the infant had an ICD-9 code indicating “livebirth” (V30, V31, V33, V34, V36, V37, V39). Stillbirth infants and unlinked mothers or infants were excluded. Demographic factors in the dataset were obtained from inpatient discharge data, and included: maternal age, infant sex, and race and ethnicity for both mother and infant as reported by the mother. Missing or unknown ethnicity was imputed as ‘Non-Hispanic’ for both the mother and infant. Mothers under 13 and older than 50 years were excluded. Age was classified as < 19 years, 20 – 34 years, or > 35 years. Missing maternal age at a given birth was imputed using the age at previous or subsequent births and the date of birth of the infant. The primary outcome, recurrent BPBI in a subsequent pregnancy, was identified using ICD-9 codes (767.6 or 953.4). The exposure was prior delivery history (parity, shoulder dystocia in a previous delivery, or previously delivering an infant with BPBI,) and present delivery method (vaginal or cesarean), which were available in the dataset. Shoulder dystocia in a previous delivery, and previous infant with BPBI were identified by using unique maternal identifiers and evaluating all deliveries associated with that mother during the study interval.

Descriptive statistics were calculated for maternal and infant demographic factors and maternal pregnancy history. The incidence of BPBI was determined for: primiparas, all multiparas, multiparas without a history of shoulder dystocia or BPBI in a previous delivery, multiparas with shoulder dystocia only (not BPBI) in a previous delivery, and multiparas who delivered a previous infant with BPBI.

Multiple logistic regression was used to model adjusted associations with BPBI. When examining parity as the exposure, we adjusted for race, ethnicity, age, delivery method (vaginal or cesarean) and delivery year, because these covariates are identifiable prenatally and have been associated with BPBI in previous studies.^1–3,28 The analysis was restricted to individuals with more than one birth over the study interval when modeling BBPI risk associated with shoulder dystocia or BPBI in a previous delivery. These models adjusted for parity and the above listed covariates. Effect modification of mode of delivery by maternal BPBI history was evaluated by adding to the model an interaction term for the product of these binary indicators. Maternal clustering effects on standard errors, which we found previously to be negligible,³ were not considered, for computational feasibility.

The method of recycled predictions⁴² was used to compare the adjusted predicted probability of BPBI for vaginal and cesarean deliveries, using predictive margins to statistically balance the covariate distributions between these two delivery modes.⁴³ This method allows us to estimate an average probability (adjusted risk) of BPBI with different delivery methods and different prior delivery histories, while holding the other variables (race, ethnicity, age, and delivery year) constant. The adjusted risk (AR) and adjusted risk difference (ARD) for BPBI between delivery modes were determined for four groups: all multiparas, multiparas without a history of shoulder dystocia or BPBI in a previous delivery, multiparas with shoulder dystocia only in previous deliveries (without BPBI), and multiparas who delivered previous infant(s) with BPBI. The number of cesarean deliveries needed to prevent (NNP) one BPBI for each group was calculated as 1/ARD. All statistical analysis was performed using SAS® software version 9.4 for Windows®. (SAS Institute Inc, Cary, NC). Significance was established at p < 0.05.

Results

The study cohort included 6,286,324 infants born to 4,104,825 individuals. 1,334,954 individuals contributed a single birth to the cohort and 2,769,871 contributed multiple births. The mean number of infants per individual was 1.53 ± 0.83. 7,762 infants were diagnosed with BPBI (1.23 BPBI per 1000 livebirths). Maternal and infant demographic characteristics are included in Table 1; the association of these demographic characteristics with BPBI has been published previously.³ Pregnancy and delivery characteristics are included in Table 2. BPBI incidence and probability by pregnancy history and delivery method are included in Table 3.

Table 1:

Demographic Characteristics of the Study Cohort

Maternal Characteristics N=6,286,324
	Mean (SD)
Age (years) n=6,286,324	28.3 (6.2)
	No. (%)
Age Category n=6,286,324
< 19 years	506,964 (8.06%)
20-34 years	4,905,359 (78.03%)
> 35 year	874,001 (19.3%)
Race n=6,286,324
Asian	718,588(11.43%)
Black	325,949(5.19%)
Native American	28,509(0.45%)
White	4,095,319(65.15%)
None of the above	1,117,959(17.78%)
Ethnicity n=6,239,875
Hispanic	2,723,422 (44.06%)
Non-Hispanic	3,516,453 (55.94%)
Infant Characteristics
Total	No. (%)
Sex n=6,286,324
Male	3,181,664 (50.61%)
Female	3,104,612 (49.39%)
Race n=6,286,321
Asian	688,770 (10.96%)
Black	319,012 (5.07%)
Native American	27,506 (0.44%)
White	4,126,162 (65.64%)
None of the above	1,124,871 (17.89%)
Ethnicity n=6,286,324
Hispanic	2,667,685 (42.44%)
Non-Hispanic	3,618,639 (57.56%)

Table 2:

Pregnancy History and Delivery Characteristics

Pregnancy History and Delivery Characteristics
	Total No. (%)	Number with BPBI	BPBI Incidence per 1000 (95% CI)
Total livebirths	6,286,324	7,762	1.23 (1.21, 1.26)
Parity
Primiparas	3,516,453 (55.94%)	4,585	1.30 (1.27, 1.34)
Multiparas	2,769,871 (44.06%)	3,177	1.15 (1.11, 1.19)
Number of previous deliveries
0	3,516,453 (55.94%)	4,585	1.30 (1.27, 1.34)
1	1,757,348 (27.96%)	2,120	1.20 (1.16, 1.26)
2	696,987 (11.09%)	718	1.03 (0.96, 1.11)
3	217,198 (3.46%)	239	1.10 (0.96, 1.24)
4	64,713 (1.03%)	62	0.96 (0.72, 1.20)
5+	33,625 (0.53%)	38	1.13 (0.77, 1.49)
Previous Shoulder Dystocia	9072 (0.14%)	53	5.84 (4.28, 7.41)
Previous BPBI	3,872 (0.052%)	62	16.01 (12.06, 19.97)
Delivery Method in Subsequent Pregnancy *
Vaginal	4,176,399 (66.44%)	5,723	1.37 (1.34, 1.41)
VBAC	71,000 (1.13%)	128	1.80 (1.49, 2.12)
Instrumented vaginal	359,483 (5.72%)	1501	4.18 (3.97, 4.39)
Instrumented VBAC	9,772 (0.16%)	44	4.50 (3.18, 5.83)
Primary Cesarean	915,149 (14.56%)	269	0.29 (0.26, 0.33)
Repeat Cesarean	754,507 (12.00%)	97	0.13 (0.10, 0.15)

^*There were 14 individuals for whom the delivery method was unknown

Table 3:

Probability of Brachial Plexus Birth Injury by Pregnancy History and Delivery Method

	Vaginal delivery			Cesarean Delivery
	Number of BPBIs/Total	Incidence per 1000 (95% CI)	Integer Rounded Probability	Number of BPBI/Total	Incidence per 1000 (95% CI)	Integer Rounded Probability
Primiparas	4361/2,598,937	1.68 (1.63, 1.73)	1 in 596	224/917,506	0.24 (0.21, 0.28)	1 in 4,098
Multiparas	3,035/2,017,717	1.50 (1.46, 1.56)	1 in 665	142/752,150	0.19 (0.16, 0.22)	1 in 5,291
Previous BPBI	52/2,353	22.11 (16.16, 28.04)	1 in 45	9/1,510	5.96 (2.08, 9.84)	1 in 168
Previous shoulder dystocia but no previous BPBI	41/6,318	6.49 (4.51, 8.47)	1 in 154	0/2,094	0	-
No prior shoulder dystocia or BPBI	2,942/2,009,046	1.46 (1.41, 1.52)	1 in 683	133/748,546	0.18 (0.15, 0.21)	1 in 5,618

The 14 individuals with an unknown mode of delivery were excluded from this table.

The “Previous BPBI” category includes individuals with a previous BPBI with or without previous shoulder dystocia. There were 655 deliveries in this row with previous BPBI with shoulder dystocia. Of these deliveries, there were 12 BPBI cases in 351 vaginal deliveries and 0 BPBI cases in 304 cesarean deliveries

Parity was associated with a 5.7% decrease in the odds of BPBI with each subsequent delivery (AOR=0.94, 95% CI 0.92, 0.97). Shoulder dystocia in a previous delivery was associated with 5-fold increased odds of subsequent BPBI (AOR = 5.39, 95% CI: 4.10, 7.08), while a history of previous BPBI was associated with a 17-fold increase in the odds of BPBI in subsequent pregnancies (AOR=17.22, 95% CI: 13.31, 22.27). (Table 4) We identified a significant interaction between history of BPBI and delivery method (p=0.0325), indicating that the effect of this history on the risk of subsequent BPBI varied by delivery method. Among individuals with a history of BPBI, subsequent cesarean delivery was associated with a 73% decrease BPBI risk (aOR=0.27, 95% CI 0.13, 0.55) compared with vaginal delivery. Among individuals without a history of BPBI, cesarean delivery in a subsequent pregnancy was associated with an 87.9% decrease in BPBI (aOR =0.12, 95% CI 0.10, 0.15) (Table 4).

Table 4:

Odds Ratios for Brachial Plexus Birth Injury by Pregnancy History

Adjusted Odds Ratios for BPBI by Pregnancy History
Entire Cohort	No BPBI (n=6,278,562)	BPBI (n=7,762)	OR (95% CI)	aOR (95% CI)
Increasing Parity, mean±SD	0.68±0.96	0.61±0.92	0.93 (0.90, 0.95)	0.94 (0.92, 0.97)
Multiparas Births Only	No BPBI (n=2,766,694)	BPBI (n=3,177)	OR (95% CI)	aOR (95% CI)
Previous Shoulder Dystocia, # (%)	9,014 (0.33)	53 (1.67)	5.19 (3.95, 6.82)	5.39 (4.10, 7.08)
Previous BPBI, # (%)	3,802 (0.14)	61 (1.92)	14.23 (11.02, 18.37)	17.22 (13.31, 22.27)
No previous BPBI (n=2,766,004)	2,762,888 (99.89)	3,116 (0.11)
Cesarean	750,507 (27.16)	133 (4.27)	0.12 (0.10, 0.14)	0.12 (0.10, 0.15)
Vaginal	2,012,381 (72.84)	2983 (95.73)	ref	ref
Previous BPBI (n=3863)	3802 (98.42)	61 (1.58)
Cesarean	1501 (39.48)	9 (14.75)	0.27 (0.13, 0.54)	0.27 (0.13, 0.55)
Vaginal	2301 (60.52)	52 (85.25)	ref	ref

All percentages are column percentages. For mode of delivery, the column percentages are computed separately within the multiparas birth subgroup defined by whether there is a history of previous BPBI.

aOR is adjusted Odds Ratios

Rows report aOR for each pregnancy history factor. A separate multiple logistic regression model was fit for each pregnancy history factor. All models included age, race, ethnicity, year of birth, and parity as independent variables. For mode of delivery, the multiple logistic regression model was fit on all multiparas births and included a main effect for previous BPBI and an interaction term for previous BPBI and cesarean delivery.

Each reported odds ratio is statistically significant (p<0.05). In addition, the interaction term for previous BPBI and cesarean delivery was statistically significant, implying that the adjusted odds ratios for Cesarean delivery differ for deliveries from a mother with a previous BPBI and those from a mother with no previous BPBI.

The adjusted risk (AR) of BPBI was lower in cesarean compared with vaginal deliveries in all groups: all multiparas, multiparas with neither shoulder dystocia nor BPBI in a previous delivery, multiparas with previous shoulder dystocia only, and multiparas with previous BPBI (Table 5). Among all multiparas, the adjusted risk difference (ARD) of cesarean compared with vaginal delivery was 131 fewer BPBIs per 100,000 births, indicating that 758 cesarean deliveries are needed to prevent one BPBI. Among multiparas without prior shoulder dystocia or BPBI in a previous delivery, 785 cesarean deliveries are needed to prevent one BPBI. Among multiparas with prior shoulder dystocia, 185 cesarean deliveries are needed to prevent one BPBI. Among individuals with a history of an infant with BPBI, 48 cesarean deliveries are needed to prevent one BPBI. For all groups, the ARD between cesarean and vaginal delivery was significant (p>0.001) (Table 5).

Table 5:

Adjusted Risk and Adjusted Risk Difference of Brachial Plexus Birth Injury in Vaginal and Cesarean Deliveries by Previous Delivery History among Multiparas

Adjusted Risk and Risk Difference of BPBI in Vaginal and Cesarean Deliveries by Previous Delivery History (per 100,000 livebirths)
	Number	AR Cesarean delivery (SD)	AR Vaginal Delivery (SD)	ARD (95% CI)	NNP	P value
All Multiparas	2,769,871	18.9 (12.0)	150.3 (93.3)	131.4 (131.3, 131.5)	757.6	<0.0001
No prior shoulder dystocia or BPBI	2,757,596	18.3 (4.5)	145.8 (36.1)	127.4 (127.4, 127.5)	784.9	<0.0001
Previous shoulder dystocia	8412	78.4 (16.7)	620.2 (131.1)	541.8 (539.4, 544.3)	184.6	<0.0001
Previous BPBI	3863	306.7 (72.7)	2387.3 (552.2)	2080.6 (2065.5, 2095.8)	48.1	<0.0001

AR = adjusted risk (estimated as predictive margins), ARD= adjusted risk difference, SD = standard deviation, NNP = Number Needed to Prevent (number of Cesarean deliveries to prevent 1 BPBI)

Discussion

BPBI is often considered an “unpredictable event,”^20–22,44 because many risk factors are not identifiable prenatally (shoulder dystocia, prolonged labor, instrumented delivery, fetal macrosomia)^45–48 and those that are (maternal obesity, pregnancy weight gain, gestational diabetes) are poorly predictive of BPBI.^45,46,49,50 Our findings indicate that prior delivery history is a valuable source of information regarding future BPBI risk. We found that both history of shoulder dystocia and history of prior BPBI were strongly associated with BPBI in subsequent births. In addition, parity is associated with a 6% decrease in BPBI risk with each subsequent pregnancy. Although cesarean delivery is associated with a lower BPBI risk, it does not eliminate BPBI risk, and the protective effect of cesarean delivery varies by prior delivery history.

Based on small historical case series, it is often assumed that previous history of shoulder dystocia or BPBI infant is associated with an increased risk of BPBI^20,24,30 but there are no prior large longitudinal cohort studies rigorously evaluating the association of shoulder dystocia or BPBI in a previous delivery with BPBI risk in a subsequent pregnancy, to our knowledge. This evidence gap is likely due to the fact that most single institution case series are unlikely to have sufficient numbers of individuals who have delivered a single or multiple infants with BPBI and do not follow individuals longitudinally, so are unable to follow subsequent deliveries or identify individuals who gave birth to an affected infant at a different hospital. Given that up to 50% of affected families pursue litigation in BPBI deliveries, ^53–55 it is plausible that individuals who deliver an affected infant deliver subsequent children at a different facility. Because our dataset includes all births at California-licensed hospitals, we are able to ascertain additional births associated with an individual even if they occurred at a different institution, as long as that subsequent birth occurred in California.

Previous studies evaluating the relationship of parity and BPBI report conflicting results,^{20–22,28,29,51,52} including a reported increased risk in primiparas,^26,28 in multíparas,⁵² or no difference related to parity.^22,2 Contradicting results from prior studies are difficult to interpret. Perhaps parity acts differently in different individuals, and the small, single center (and potentially homogeneous) cohorts in previous studies reflect those differential effects. Alternatively, small cohorts may be underpowered to detect the effect size of parity. In addition, many studies utilize univariate analyses that do not control for covariates, or control for covariates that are not known prenatally (e.g. shoulder dystocia, operative vaginal delivery). Our study design overcomes many of these limitations by including a large, diverse cohort powered to detect small effect sizes, longitudinal data over 17 years that allows for assessment of more subsequent pregnancies, and the capacity to control for important covariates.

We observed that Cesarean delivery decreases but does not eliminate BPBI risk, consistent with previous studies.^56,57 In particular, the probability of an individual with previous BPBI having a subsequent affected infant was 1 in 45 with a vaginal delivery and 1 in 168 with a cesarean delivery. The finding that cesarean delivery does not completely eliminate BPBI risk suggests the possibility of alternative risk factors other than those associated with traversing the birth canal, including in utero injury. ^48,56,60 Augustine et al⁵⁷ identified a higher risk of BPBI in emergent compared with elective cesarean deliveries and hypothesized that unfavorable intrauterine positioning may result in excessive force on the brachial plexus. These findings can inform discussions between clinicians and individuals with this prior delivery history regarding BPBI risk in subsequent pregnancies. Clinician-patient conversations should be based in shared decision-making and include a discussion of the other well-established short- and long-term risks of cesarean delivery, for both the pregnant individual and child.^58,59

Study limitations include the retrospective design and limitations inherent in the use of administrative datasets, including that the dataset used was created for medical billing and resource allocation so some variables potentially relevant to BPBI risk (e.g maternal weight gain, body mass index, medical comorbidities and estimated fetal weight) were not available. We limited the scope of our analyses to characteristics that can be identified prenatally; consequently, we did not analyze the effect of factors such as birth weight and operative vaginal delivery. Our cohort only included infants diagnosed with BPBI at birth, so may have missed infants with mild injuries not diagnosed at birth and likely included infants with BPBI that ultimately resolved. However, because BPBI is commonly diagnosed at birth, is not easily confused with other diagnoses, has unique ICD-9 codes, and does not require confirmatory testing, we believe the accuracy for this diagnosis is high. Another limitation is that in the early years of the dataset multiparas may have delivered prior to onset of data linkage and their prior delivery histories would not be included. Additionally, the risk of a subsequent infant with BPBI delivered vaginally is calculated from those individuals who attempted a vaginal delivery with this prior history, which is likely a select group, and therefore, introduces a selection bias; however, we believe this bias underestimates this risk, as individuals who did not undergo subsequent vaginal delivery may have had more BPBI risk factors and were advised against vaginal delivery.

Strengths of this study include the size of the dataset, allowing us to identify associations that may not be detectable in smaller datasets, as well as the continuous coverage of maternal-infant birth pairs over a 17-year period, allowing us to track individuals over time and identify many more subsequent deliveries than were captured in other studies, even if they occurred at a different institution. The accuracy of this dataset for obstetric complications and birth diagnoses has been characterized in previous studies.^39–41 The use of linked maternal-infant data improves upon studies using infant-only or maternal-only data which often do not provide sufficient information to evaluate the effect of maternal history on infant conditions.^61,62 Additionally, this dataset compiles data from several sources, which allows cross-checking accuracy among overlapping variables and broader coverage from non-overlapping variables. Another strength is the diversity of maternal-infant pairs in our cohort.

This investigation implies that individuals with a history of shoulder dystocia or BPBI in a prior pregnancy are at substantially higher risk of delivering an infant with BPBI compared with those without this history. Clinicians can use this information to counsel patients regarding their risk of delivering an affected infant and discuss the risks and benefits of various delivery strategies. Further work is needed to evaluate the potential benefits of Cesarean delivery in individuals at higher risk of delivering an infant with BPBI, while weighing the well-known risks of Cesarean delivery. Lastly, the finding that cesarean delivery does not eliminate BPBI risk suggests risk factors for BPBI unrelated to trauma during vaginal delivery. Future studies evaluating BPBI occurring in Cesarean deliveries or in individuals who deliver multiple affected infants may provide insight into additional potential risk factors.