Design, setting, participants and ethics
This was an observational, cross-sectional study involving patients with AIS who underwent surgical treatment in the same hospital and in whom intraoperative EMG measurements were compared with the implant positions evaluated by CT. The Hospital dos Servidores Estaduais institutional review board approved this study, approval number 19134914.2.0000.5463 .
All consecutive patients who underwent surgery performed by the same surgical team in the same institution from March to December 2013 were included in the study. Patients were excluded if they had scoliosis with a known etiology (i.e., not AIS), if they were undergoing revision surgery, or if no postoperative CT was available for review.
All patients were operated by a posterior approach, with insertion of pedicle screws from the same manufacturer (DePuy, Synthes, Raynham, MA, USA) by the "free hand" technique  and under intravenous anesthesia . Neurophysiological monitoring was employed in all surgeries using the same technique and device.
Variables and measurements
After pedicle screw insertion, EMG was performed by directly stimulating the implants, using a monopolar electrode (cathode) and a subdermal needle electrode inserted into the paravertebral musculature (anode). Stimulation was performed with a frequency of 3 Hz, a duration of 0.1 ms and an increasing intensity until an EMG response could be observed . The maximum intensity used was 30 milli-amperes (mA). For each screw, the lowest intensity able to generate a measurable response was recorded as the EMG threshold for that screw. If no response was observed, a value of 30 was assigned to that screw.
Within three months of surgery, patients underwent CT to evaluate the implant positioning and classification according to the criteria proposed by Abul-Kasim et al. . This grading system was developed to distinguish between lateral, medial and anterior cortical perforations and foraminal perforation and is based on whether the cortical violation is partial or total rather than the length (mm) of the perforation.
Each individual screw to the corresponding pedicle was assessed and classified in both the medial and sagittal planes as follows: normally placed in the medial plane; medial cortical perforation (MCP) grade 1, partially medialized; MCP grade 2, totally perforating the medial pedicular cortex; lateral cortical perforation (LCP) grade 1, partially lateralized but anchored in the vertebral body; LCP grade 2, abutting the outer cortex of the vertebral body and not anchored in the vertebral body; normally placed in the sagittal plane; perforating the inferior underlying neural foramen (INF); or perforating the superior underlying neural foramen (SUP). Additionally, screws classified as MCP grade 1 or MCP grade 2 in the axial plane and as perforating the INF or the SUP in the sagittal plane were considered “at risk for nerve injury” (ARNI), as these screws are closer to neural elements. Screws classified as normally placed and screws classified as LCP, although recognized as misplaced, were considered “no risk for nerve injury” (NRNI).
The total intravenous technique (TIVA) was used to induce anesthesia in the AIS surgeries, and medications that usually do not interfere with intraoperative neurophysiological monitoring (propofol and remifentanil) were administered. All AIS corrections included in this study were performed by the same surgical team, using the same techniques for surgery and electromyography evaluation. All CT scans were evaluated by the same observer (BMG).
Continuous variables are presented as the mean and standard deviation, and categorical data are presented as absolute and relative frequencies. A descriptive analysis of the positioning of all screws was performed. To evaluate the diagnostic accuracy of EMG for predicting screws ARNI, we excluded all screws inserted above T6, as those pedicles have lower reliability for EMG acquisition . A single patient contributed multiple sampling units (screws) to the analysis, resulting in a hierarchical structure of the generated data, with subjects as the primary sampling units and individual screws as the secondary sampling units. The association between the EMG threshold recorded intraoperatively and postoperative screw status, considering the risk for nerve injury, was investigated using generalized estimating equations (GEEs). Only medial cortical perforation of the screw (MCP grade 1 or MCP grade 2) was considered a positive outcome (ARNI) in the axial plane, while both superior or inferior deviation of the screw were considered positive outcomes in the sagittal plane. These statistical models are similar to logistic regression models but take into account the hierarchical structure of the data.
The diagnostic accuracy of EMG for predicting screw malposition was investigated using a receiver operating characteristic (ROC) curve. The curve represents a plot of the sensitivity and specificity at progressive cutoffs of a diagnostic test measured on a continuous scale. Therefore, the area under the curve (AUC) is a measure of the ability of EMG to discriminate between screws ARNI and screws NRNI. Estimates of diagnostic accuracy, including sensitivity (Sn), specificity (Sp), negative predictive value (NPV) and positive predictive value (PPV), were calculated for cutoffs at every 5 mA and are presented with their 95% confidence intervals (95%CI). All statistical analyses were performed using the statistical package STATAÒ 14 (StataCorp, TX/EUA). Associations with p<.05 (two-sided) were considered significant.