The Cochrane Handbook for Reviews of Diagnostic Test accuracy was used as guidance to write the methods (36) and the preferred reporting items for systematic reviews and meta-analyses (PRISMA) (37) guidelines were followed.
Types of studies
Studies involving infants with obstetric brachial plexus injuries that report the findings of preoperative MRI in comparison to surgical exploration of the brachial plexus roots will be included. Case reports will be excluded.
Participants
This review will include studies involving children under five years old with obstetric brachial plexus injuries. All injuries to the brachial plexus that occur during delivery will be included, irrelevant of the aetiology (e.g. shoulder dystocia, forceps delivery etc.) Bilateral injuries will also be included.
Index Test
The role of MRI will be to detect root avulsions, other nerve abnormalities and pseudomeningoceles. In OBPI MRI is typically performed at 3 months in infants with persistent upper limb functional limitations. The MRI acquisition will vary in terms of the physical scanner used (manufacturer and model), field strength, pulse sequences, coil arrangement, gradients, postprocessing techniques and other factors – all of which will impact upon image quality and hence diagnostic accuracy. Variations will also arise due to the subjective nature of image interpretation. A radiologist will review the images and either confirm or exclude the presence of avulsion, other nerve abnormality or pseudomeningocele. Positive findings for avulsion are detected by a lack of continuity or absence of the nerve root between the spinal cord and exit foramen (29). Other nerve abnormalities than may be detected include nerve scarring, neuroma or rupture, and these can also be referred to as post-ganglionic nerve injuries (25, 27). An abnormal contour of the dura and collection of dorsal extraspinous fluid is indicative of pseuomeningocele (38) and is considered a surrogate marker of root avulsion given that rupture of the dura mater suggests that the nerve root is also ruptured, although this has been disputed in some literature (39, 40). The presence of one suspected avulsion is of equal importance to that of any number of avulsions, given that any avulsion would warrant nerve transfer surgery. Due to the lower energy stretching forces that typically cause OBPI other types of severe nerve injury apart from avulsion may also be present and can also require surgery e.g. nerve grafting. Avulsions, other nerve injury and pseudomeningoceles can occur at any spinal level from C4 to T2 and may, in rare cases, occur bilaterally. The ability of MRI to identify patients with no root avulsion is vital as surgical exploration could potentially be avoided.
Target condition
Avulsion of the roots of the brachial plexus is the target condition. The ability of MRI to differentiate between any number of root avulsions and no avulsions will be examined. The secondary target conditions are an abnormal nerve and pseudomeningocele.
Reference standard
Surgical exploration of the brachial plexus is the reference standard for detecting root avulsion and other nerve abnormalities. The operation is performed under general anaesthesia and involves an incision in the supraclavicular fossa which extends towards the deltopectoral groove (29). The operation allows direct inspection of the spinal nerve roots C4-T1. Additional intraoperative tests such as somatosensory evoked potentials (SEPs) and bipolar motor nerve stimulation are included as part of the reference standard. SEPs involve measuring cortical activity induced by applying pulses of varying frequency to the nerve roots. Avulsed nerves will not transmit signals to the brain meaning no activity is detected on an encephalogram. Bipolar nerve stimulation involves applying a current across the nerve which would normally cause the corresponding muscle to contract, however in the case of avulsion, no muscle contraction is observed. These intraoperative tests aid surgeons in the diagnosis of root avulsion.
Search strategy
Electronic searches
EMBASE, PubMed and CENTRAL electronic databases will be searched from inception to present date with no restrictions. The medRxiv and bioRxiv preprint archives will also be searched using medrixr (41). GScraper will be used to further increase coverage by pulling hits from Google Scholar (42). The planned EMBASE and PubMed search strategy was formulated with a search strategist and is presented in appendix 1.
References from published studies
Citations will be imported and de-duplicated using EndNote. References will then will be screened using CitationChaser (43).
Study selection
JB will apply the eligibility criteria to screen titles and abstracts for relevance. Two authors (JB and CH) will independently screen identified citations using Rayyan REF. Disagreements will be resolved by discussion with a third author (GB). Full text of eligible studies will then be screened and subsequently labelled as included or excluded. Reasons for exclusion will be noted.
Data Extraction
For included studies the following data will be extracted using an excel spreadsheet. This will be done independently in duplicate by JB and CH: study identifier; number of participants; sex; participants age at diagnosis, MRI scan and surgery; country of origin; time from birth to imaging and surgery; presence of clavicular/humeral fracture; type of MRI scanner used (including brand, model and field strength); pulse sequence; number of true positives, false positives, true negatives and false negatives relating to the detection of root avulsion, abnormal nerve or pseudomeningocele using MRI. The priority outcome is the detection of root avulsion at the brachial plexus as this is the is most important clinically. Detection of abnormal nerves and pseudomeningocele are secondary outcomes. We will contact authors of studies by email if data is missing or unclear.
Methodological quality assessment
An adapted version of the Quality Assessment of Diagnostic Accuracy Studies (QUADAS-2) (44) tool (Appendix 2) will be used to assess the risk of bias and strength of evidence of the eligible studies respectively at the study level. Two authors (JB and CH) will independently complete the assessment. Disagreements will be resolved by discussion with a third author (GB).
Data synthesis
Analysis will be performed separately for each target condition (root avulsion, abnormal nerve and pseudomeningocele) with the patient and then the nerves being the unit of analysis. Forest plots and receiver operating characteristic plots will be used to display estimates of sensitivity and specificity of the included studies as part of the preliminary analysis. These plots will be generated using MetaDTA (45). Summary sensitivities and specificities will be obtained using a bivariate model for meta-analyses, providing data is sufficient (46). Meta-analyses will be performed using Stata version 15 (47).
Investigations of heterogeneity
Heterogeneity in the diagnostic accuracy of MRI will be examined using meta-regression or subgroup analyses if data permits. Variations in field strength is likely to be a source of heterogeneity (29) and will be investigated. Babies with OPBI who have not undergone surgical exploration (e.g. they are too unwell) may result in underestimation of the number false negatives which in turn could upwardly bias the sensitivity of MRI. Furthermore, the diagnostic accuracy of MRI could be biased by retrospective studies which have recruited an unrepresentable sample of patients.
Sensitivity analysis
The impact of bias will be evaluated via sensitivity analyses. Studies with a high or unclear risk of bias as identified by the tailored QUADAS-2 tool will be excluded.
Assessment of reporting bias
Reporting bias will not be assessed given the lack of sensitive statistical methods (36). Data will be made available on the Open Science Framework.