HAND/PERIPHERAL NERVE Primary Nerve Repair for Obstetrical Brachial Plexus Injury: A Meta-Analysis Christopher J. Coroneos, M.D., M.Sc. Sophocles H. Voineskos, M.D., M.Sc. Marie K. Coroneos, M.D. Noor Alolabi, M.D. Serge R. Goekjian, M.D. Lauren I. Willoughby, M.D. Achilleas Thoma, M.D., M.Sc. James R. Bain, M.D., M.Sc. Melissa C. Brouwers, M.A., Ph.D. On behalf of the Canadian OBPI Working Group Hamilton and Toronto, Ontario, Canada

Background: Nerve repair may be effective in improving function following obstetrical brachial plexus injury. No previous review has directly compared nerve repair to nonoperative management for similar patients, and no previous analysis has been adequately powered to determine whether nerve repair reduces impairment. Methods: Electronic databases were searched (MEDLINE, Embase, CINAHL, and Cochrane Central). Eligible studies were randomized controlled trials, observational studies, and case series (n > 9); included patients younger than 2 years undergoing nerve repair or nonoperative management of obstetrical brachial plexus injury; and reported functional impairment. Two reviewers independently screened articles using objective a priori criteria. Bias was assessed for each study. Overall quality of evidence was evaluated for each outcome. Results: Among nine cohort studies including 222 patients, nerve repair significantly reduced functional impairment compared with nonoperative management (relative risk, 0.58; 95 percent CI, 0.43 to 0.79; p < 0.001; I2 = 0 percent; absolute risk reduction, 19 percent; number needed to treat, six). Findings are consistent with comparison of similar patients from case series. With operative management, no deaths were reported; major adverse events were reported in 1.5 percent, and minor adverse events were reported in 5.0 percent of cases. Among demographic (all severities) samples managed nonoperatively, residual impairment remains in 27 percent (19 to 36 percent). Conclusions: Nerve repair reduces functional impairment in obstetrical brachial plexus injury. Nonoperative management in patients with a deficit at 3 months of age leads to a high proportion of functional impairment. Mortality is not a common risk of modern pediatric microsurgical nerve repair. Residual impairment with nonoperative management is underestimated in the reported literature.  (Plast. Reconstr. Surg. 136: 765, 2015.) CLINICAL QUESTION/LEVEL OF EVIDENCE: Therapeutic, IV.

O

bstetrical brachial plexus injury occurs secondary to traction on the neck-shoulder angle during labor and delivery.1 Most cases are transient.2 Children with incomplete recovery suffer lifelong impairment, including weakness, From the Division of Plastic Surgery and the Departments of Surgery, Clinical Epidemiology and Biostatistics, and Oncology, McMaster University; and the Department of Obstetrics and Gynecology, University of Toronto. Received for publication November 25, 2014; accepted March 27, 2015. Presented at the 93rd Annual Meeting of the American Association of Plastic Surgeons, in Miami, Florida, April 5 through 8, 2014, where it won third place, resident competition; and at the 68th Annual Meeting of the Canadian Society of Plastic Surgeons, in Montreal, Quebec, Canada, June 24 through 28, 2014, where it was awarded best poster. Copyright © 2015 by the American Society of Plastic Surgeons DOI: 10.1097/PRS.0000000000001629

joint contracture, and limb length discrepancy.3 Obstetrical brachial plexus injury impacts global development and family dynamics.3 Primary management remains controversial. Nerve repair is hypothesized to improve function. Early analysis

Disclosure: The authors have no financial interest to declare in relation to the content of this article. Supplemental digital content is available for this article. Direct URL citations appear in the text; simply type the URL address into any Web browser to access this content. Clickable links to the material are provided in the HTML text of this article on the Journal’s Web site (www. PRSJournal.com).

www.PRSJournal.com

765

Copyright © 2015 American Society of Plastic Surgeons. Unauthorized reproduction of this article is prohibited.

Plastic and Reconstructive Surgery • October 2015 by Narakas4 and Gilbert and Tassin5 established the first indications for nerve repair. These were followed by management algorithms by Terzis et al.,6 Clarke and Curtis,7 and others,8–11 based on the authors’ own series and critical analysis of existing literature. There remains no conclusive synthesis to guide decision-making for nerve repair or nonoperative management. Rationale Commonly cited obstetrical brachial plexus injury literature is limited to cohort studies and case series.12 Nerve repair is typically compared to retrospective controls. Operative indications and timing have not been compared with a randomized controlled trial, despite acknowledgment of its need.13 Previous systematic reviews2,14–18 have poor methodologic quality, do not compare patients with similar prognoses, and are ultimately inconclusive. These systematic reviews conclude that evidence is insufficient to provide management recommendations. Effectiveness of operative intervention is not supported. Despite these conclusions, nerve repair continues to be practiced at specialized centers worldwide. The literature contains cohort studies comparing microsurgical nerve repair versus nonoperative therapy in patients with defined impairment at specific ages. However, no previous systematic review has applied methodologic solutions to pool these existing studies. Moreover, a number of physical assessment scales exist for obstetrical brachial plexus injury. No previous systematic review has pooled outcomes from these different scales, based on their similar definitions of impairment. The dearth of conclusive evidence may have allowed historic attitudes to remain entrenched among primary care providers—“permanent sequelae are rare,”19 and operative repair is “unwarranted.” Recovery is often overestimated by primary care providers, causing guardian distress and delayed specialist referral. However, there are methodologic solutions to address the inherent limitations in the literature. In clinical situations where evidence is unclear, rigorous approaches to knowledge synthesis have the highest capacity to impact practice. The methodologic quality and outcome completeness of existing systematic reviews have not been optimized, do not provide direction to support clinical decision-making, and thus may compromise patient outcomes. The objective of this study was to assess the effect of primary nerve repair versus nonoperative management on physical function in patients with obstetrical brachial plexus injury.

PATIENTS AND METHODS Protocol A protocol was developed a priori; it was not published. Eligibility Criteria Eligible studies were full reports of systematic reviews, randomized controlled trials, observational trials, and case series (n > 9) in English. Studies included infants with obstetrical brachial plexus injury undergoing any nerve repair, or nonoperative management alone, by age 24 months. Pharmacologic therapies (e.g., botulinum toxin) and secondary surgery (e.g., tendon transfers) were excluded. Given inconsistent definitions of recovery in the literature,12 our primary outcome was defined as functional impairment. Secondary outcomes included recovery in demographic samples, and adverse events. Data Sources and Search Electronic search was designed and executed with a health sciences librarian, on February 7, 2013: MEDLINE, 1946 to 2013; Embase, 1980 to 2013; CINAHL, 1982 to 2013; and Cochrane Central, 2013, Issue 1. (See Text, Supplemental Digital Content 1, which shows search criteria, http://links. lww.com/PRS/B397.) Screening was sensitive; any potentially relevant references were included in full-text review. Titles and abstracts were screened in duplicate for eligibility (C.J.C., S.H.V., M.K.C., N.A., S.R.G., and L.I.W.). Full texts of eligible articles were screened in duplicate by the same reviewers, using a pilot-tested eligibility form designed a priori for this study. Reasons for exclusion were recorded. Disagreements were resolved by consensus, and supervisor review (J.R.B., A.T., and M.C.B.). Agreement was calculated using Cohen’s unweighted kappa. Data Extraction and Quality Assessment Data were extracted, and methodologic quality was assessed in duplicate, guided by a pilottested form and data dictionary designed a priori for this study. Disagreements were resolved by consensus and supervisor review. If multiple reports from the same trial or series were present, the most recent report was used. Patients requiring secondary surgery were considered functionally impaired if no other outcome was provided. In articles with multiple outcomes, hierarchy for abstraction was as follows: shoulder, elbow, hand, and global. Randomized controlled trials were assessed with Cochrane Collaboration’s

766 Copyright © 2015 American Society of Plastic Surgeons. Unauthorized reproduction of this article is prohibited.

Volume 136, Number 4 • Obstetrical Brachial Plexus Injury Risk of Bias Tool.20 Nonrandomized studies were assessed with the methodological index for non-­ randomized studies instrument, with eight items applied to noncomparative studies and 12 applied to comparative studies.21 Agreement was calculated using weighted kappa. Data Synthesis and Analysis Comparative studies were analyzed with Review Manager 5.2.22 Relative risks were calculated with 95 percent confidence intervals using a random effects model. Optimal information size was calculated with alpha = 0.05, beta = 0.20, and relative risk reduction  =  0.20 and the nonoperative event rate for each outcome; optimal information size is a power calculation for pooled analyses. Case series were analyzed with proportional meta-analyses, with 95 percent confidence intervals and a random effects model. For operative management, all nerve repair/transfer interventions were pooled. Nonoperative management was defined as the absence of surgery; all rehabilitation therapy and natural history outcomes were pooled. There is insufficient evidence to preclude combining all nerve repair techniques23 and nonoperative protocols.14,17 Sources of heterogeneity were defined a priori: follow-up time (2 years versus shorter), joint assessment (shoulder abduction versus external rotation versus elbow), and injury pattern (upper versus total plexus). Worse outcomes were anticipated with short follow-up, assessment of external rotation, and total plexus injury. Heterogeneity was explored with I224; outcomes with I2 greater than 40 percent were explored with subgroup analyses. Funnel plots were used to explore publication bias for outcomes with at least 10 studies. Comparing all operative outcomes to all nonoperative outcomes would be biased; patients with neurapraxic injuries recover spontaneously within 1 month. These patients are not typically referred to surgeons, and are absent from studies.

Including these patients would favor outcomes in the nonoperative group. Operative outcomes should be compared to nonoperative outcomes for similar patients (e.g., absent elbow flexion at 3 months). A “gray zone” of prognosis in obstetrical brachial plexus injury was previously defined, where “the decision as to the benefits and risks of surgery versus no surgery is not clear.”12 In each study, a group of patients with a defined impairment was identified; these criteria were often surgical indications (e.g., no recovery of biceps at 3 months, failed cookie test at 9 months). Criteria differ between studies. Nonoperative outcomes for patients in the gray zone were used as a control for nerve repair. To pool outcomes across multiple scales, all physical recovery scores were converted to binary functional impairment outcomes. The definition of a “functional” outcome was identified in the source reporting of each scale (Table  1).25–33 All scores below “functional” were considered “impaired.”

RESULTS Search Results Search resulted in 4177 citations, with 391 articles selected for full-text review, and 107 articles were selected for data extraction (Fig.  1). Unweighted kappa for agreement was 0.67. Eight articles were excluded during data extraction for not meeting eligibility criteria. Among 99 articles included in analysis, there were no randomized controlled trials, 17 cohort studies, and 82 case series. Operative outcomes were reported in 39 articles, nonoperative outcomes were reported in 73 articles, and both outcomes were reported in 13 articles. Report Characteristics and Quality Quality assessment is summarized in Table  2. Weighted kappa for agreement was 0.56. Quality of the literature was moderate to poor. Among identified cohort studies, nonoperative control groups

Table 1.  Obstetrical Brachial Plexus Injury Assessments and Corresponding Functional Scores* Assessment Mallet Score Active Movement Scale Narakas Grading System Gilbert Shoulder Classification Gilbert-Raimondi Elbow Classification Raimondi Hand and Wrist Classification Medical Research Council Scale Gilbert and Tassin modified Medical Research Council Scale

Reference

Functional Score

Gilbert et al., 1991 Lin et al., 200926 Narakas, 198527 Haerle and Gilbert, 200428; DiTaranto et al., 200429 DiTaranto et al., 200429 DiTaranto et al., 200429; Gilbert, 199530 Laurent et al., 199331; Boome and Kaye, 198832

III 6 As described III 2–3 III M3

25

Gilbert and Tassin, 198733

M3

*Mallet Score, Gilbert Shoulder Classification, Raimondi Hand, and Wrist Classification are each evaluated in roman numerals; Medical Research Council Scale and Gilbert and Tassin’s modification are each scored numerically with a leading M.

767 Copyright © 2015 American Society of Plastic Surgeons. Unauthorized reproduction of this article is prohibited.

Plastic and Reconstructive Surgery • October 2015

Fig. 1. Flow diagram of search and study selection.

were unequal at baseline, including fewer severe injuries. These groups had a better prognosis, biasing against the operative groups. Furthermore, sample sizes were not calculated in identified cohort studies, and the follow-up periods were variable. Operative series were typically retrospective and not consecutive, but had long follow-up and objective outcome definitions. Only one operative series reported outcomes with a subjective measure. Nonoperative series had large demographic patient samples, but were similarly retrospective, and endpoints were typically subjective. Primary Outcomes Functional Impairment: Direct Comparison of Nerve Repair versus Nonoperative Cohort Studies Outcomes were available for 222 patients from nine cohort studies comparing nerve repair versus nonoperative management. Operative indications, control group characteristics, and outcomes

are described in Table  3.11,31,32,34–38 Typically, studies compared nerve repair versus nonoperative management in infants with absent elbow flexion/ biceps function at 3 to 4 months. Outcomes were measured with objective physical assessments at 1 to 2 years postoperatively. All studies were small. Tendon transfers were considered poor outcomes in two studies.35,37 Gray zone criterion was extrapolated from a median value in one study.38 Nerve repair significantly reduced functional impairment (relative risk, 0.58; 95 percent CI, 0.43 to 0.79; p < 0.001; I2 = 0 percent) (Fig. 2). The comparison was underpowered; optimal information size of 514 was not met. However, all studies were consistent in favoring nerve repair. Functional Impairment: Indirect Comparison of Nerve Repair versus Nonoperative Gray Zone Case Series Operative outcomes were available for 1128 patients from 30 series.11,30,32,34,35,37–61 Operative

768 Copyright © 2015 American Society of Plastic Surgeons. Unauthorized reproduction of this article is prohibited.

Volume 136, Number 4 • Obstetrical Brachial Plexus Injury Table 2.  Quality Assessment Summarized by Outcome*

Functional impairment: operative vs. gray zone nonoperative  Cohort studies

2 4 3 5 18 7 1 10 8

3 4 2 18 7 5 9 8 2

3 2 4 25 3 2 11 3 5

0 0 9 0 1 29 0 3 16

9 0 0 30 0 0 18 0 1

0 5 4 0 13 17 0 10 9

2 4 3 13 10 7 6 6 7

Sample Size Adequate Control Contemporary Baseline Equivalence Analyses

Loss to Follow-Up

Follow-up Period

Unbiased Assessment

Endpoints

222

9

Scale, 9; subjective, 0

>24: 4 24: 17 24: 8 24: 9 24: 9

Primary Nerve Repair for Obstetrical Brachial Plexus Injury: A Meta-Analysis.

Nerve repair may be effective in improving function following obstetrical brachial plexus injury. No previous review has directly compared nerve repai...
487KB Sizes 2 Downloads 10 Views