Does Spinal Cord Type Predict Intraoperative Neuro-Monitoring Alerts in Scoliosis Correction Surgery? A Systematic Review and Meta-Analysis of Operative and Radiologic Predictors

Abstract

Study Design

Systematic literature review and meta-analysis.

Objectives

Predicting patient risk of intraoperative neuromonitoring (IONM) alerts preoperatively can aid patient counselling and surgical planning. Sielatycki et al established an axial-MRI-based spinal cord classification system to predict risk of IONM alerts in scoliosis correction surgery. We aim to systematically review the literature on operative and radiologic factors associated with IONM alerts, including a novel spinal cord classification.

Methods

A systematic review and meta-analysis was performed as per the Preferred Reporting Items for Systematic Reviews and Meta-analyses (PRISMA) Guidelines. A literature search identifying all observational studies comparing patients with and without IONM alerts was conducted. Suitable studies were included. Patient demographics, radiological measures and operative factors were collected.

Results

11 studies were included including 3040 patients. Relative to type 3 cords, type 1 (OR = .03, CI = .01-.08, P < .00001), type 2 (OR = .08, CI = .03, P <.00001) and all non-type 3 cords (OR = .05, CI = .02-.16, P < .00001) were associated with significantly lower odds of IONM alerts. Significant radiographic measures for IONM alerts included coronal Cobb angle (MD = 10.66, CI = 5.77-15.56, P < .00001), sagittal Cobb angle (MD = 9.27, CI = 3.28-14.73, P = .0009), sagittal deformity angle ratio (SDAR) (MD = 2.76, CI = 1.57-3.96, P < .00001) and total deformity angle ratio (TDAR) (MD = 3.44, CI = 2.27-4.462, P < .00001). Clinically, estimated blood loss (MD = 274.13, CI = −240.03-788.28, P = .30), operation duration (MD = 50.79, CI = 20.58-81.00, P = .0010), number of levels fused (MD = .92, CI = .43-1.41, P = .0002) and number of vertebral levels resected (MD = .43, CI = .01-.84, P = .05) were significantly greater in IONM alert patients.

Conclusions

This study highlights the relationship of operative and radiologic factors with IONM alerts.

Keywords

spine deformity adolescent idiopathic scoliosis magnetic resonance imaging neuromonitoring alert spinal cord injury

Introduction

Scoliosis may be secondary to a number of etiologies including congenital, neuromuscular, syndromic, degenerative and traumatic, however adolescent idiopathic scoliosis (AIS) is by far the most common form accounting for 75-80% of cases.^1,2 In severe or progressive cases, surgery is often required, typically occurs through a standard posterior approach for deformity correction and spinal fusion (PSF); various osteotomies may be utilised in more severe cases.³ Although rare, iatrogenic spinal cord injury (SCI) resulting in sensorimotor disturbance or paraplegia is still one of the most feared complications of spine deformity correction injury.⁴ Surgically, SCI can occur due to intraoperative cord compression, ischaemia, traction or direct mechanical injury, such as with pedicle screw placement or misguided instrumentation.⁵ The incidence of neurological sequalae following deformity correction surgery is estimated to be approximately 1%, according to the scoliosis research society.⁶ Repercussions of SCI extend far beyond the physical ones and include psychological, social, legal and economic ones.⁷

Intraoperative neuromonitoring (IONM) is a valuable tool that provides continuous monitoring of sensorimotor function and can herald a spinal cord injury. Alterations of normal spinal cord signals may indicate an insult to the spinal cord, which is why rapid detection of these changes is crucial so that immediate intervention can take place to prevent irreversible neurologic injury.⁸ Combined transcranial motor evoked potentials (TcMEP) and somatosensory evoked potentials (SSEP) have a high sensitivity and specificity and hence, IONM is now standard practice in many scoliosis centres.⁹

Several radiographic deformity measures have been linked with the risk of IONM alerts.¹⁰ More recently, Sielatycki et al have developed a T2-weighted axial-MRI-based classification system that classifies patients into 1 of 3 cord types with the aim of predicting IONM alerts in thoracic deformity correction surgery.¹¹ The intended role of this classification system is to help surgeons counsel patients and prepare for surgery. Type 1 cords were representative of less severe deformity and less IONM alerts whereas, type 2 cord were ambiguous. On the other hand, type 3 cords correlated with more severe deformity and increased incidence of IONM alerts.¹¹

The ability to predict patients’ risk of IONM alerts would allow surgeons to take necessary preoperative precautions and counsel patients accordingly. This study aims to systematically review the literature on this topic, validate the novel classification system and assess operative and radiographic factors associated with IONM alerts. This is the first systematic review and meta-analysis to consider prognosticators of IONM alerts.

Materials and Methodology

A systematic review and meta-analysis were conducted according to the Preferred Reporting Items for Systematic Reviews and Meta-Analysis (PRISMA) guidelines.¹²

Literature Search

A comprehensive systematic literature search of studies up to the August, 10 2023 was conducted by 2 independent authors. MEDLINE was the primary database used. Combinations of Medical Subject Headings (MeSH) and Embase (EMTREE) headings as well as subheadings were utilised to search databases up to August 10th 2023. Additionally, EMBASE and the Cochrane Central Register of Controlled Trials (CENTRAL) was searched for relevant trials. Finally, to check for any ongoing or unpublished studies, the World Health Organisation International Trials Registry and ClinicalTrials.gov were searched. References of all included studies were examined to identify any other studies that were not captured by the initial search. The search terminology included the following terms: ‘thoracic,’ ‘neuromonitoring,’ ‘intraoperative,’ ‘scoliosis,’ ‘deformity,’ ‘spinal,’ ‘cord,’ ‘injury,’ ‘kyphoscoliosis,’ ‘adolescent,’ ‘cobb angle,’ ‘deformity angle ratio,’ ‘DAR,’ ‘coronal,’ ‘sagittal.’ Appropriate combinations were used to identify any existing studies validating the spinal cord classification system as well studies assessing radiological predictors of IONM alerts in scoliosis surgery.

Inclusion and Exclusion Criteria

All studies comparing radiological parameters in scoliosis patients and corresponding categories of those who/who do not demonstrate IONM alerts were included. All scoliosis aetiologies were included. This included any validation studies of the spinal cord classification system. Studies that did not directly compare both IONM and non-IONM groups were excluded. Studies with missing statistical values, such as mean or standard deviation were excluded following attempts to obtain this data from initial authors. Finally, studies in languages other than English were also excluded.

Primary Outcomes

The primary outcome was to assess the validity of the Sielatycki et al classification system in detecting IONM alerts. This involved assessing the odds of IONM alerts between different spinal cord types. The classification system classifies patients into 3 groups. Type 1 spinal cord: this is a round or circular cord that is symmetrical and CSF visible between the cord and the apical concave pedicle or vertebral body. Type 2 spinal cord: this is a round or circular cord that is symmetrical with no CSF visible between the cord and the apical concave pedicle or vertebral body. Type 3 spinal cord: This is a flattened or deformed cord with no intervening CSF visible.¹¹

Secondary Outcomes

Secondary outcomes assess the relationship between clinical and radiographic measures and IONM alerts. The following radiological measures were assessed: Coronal cobb angle (CA) in degrees, flexibility index (FI) (%) which is the percentage of Cobb angle correction with lateral bending towards the apex, thoracic kyphosis, coronal deformity angle ratio or C-DAR, which is the coronal (PA) Cobb angle divided by the total number of vertebrae within the scoliotic curve, sagittal DAR or S-DAR, which is the thoracic kyphosis divided by the total number of vertebrae within the curve, total DAR or T-DAR which is the sum of C-DAR and S-DAR (angle per level is the unit for DAR). Also of interest was the comparison of intraoperative clinical factors between alert and non-alert patients.

Study Selection

Titles and abstracts of articles identified from the literature search were assessed for relevance. Full texts of relevant articles were assessed to determine eligibility for inclusion. Discussion between authors occurred to resolve any uncertainty in study selection.

Data Management and Analysis

An electronic spreadsheet including include baseline study characteristics, patient demographics as well as primary outcomes and secondary outcomes was created. Data extraction from included studies was performed by 2 authors independently and disagreements were resolved through discussion. Statistical analysis was performed using Review Manager 5.4 software. Extracted data was inputted into the software. Data analysis was performed using the random effects model to account for any heterogeneity present between the studies due to variable study conditions such as patient demographics or surgical technique. Odds Ratio (OR) was used to compare dichotomous outcomes between 2 groups. The OR represents the odds of an IONM event in one spinal cord type relative to another. Mean Difference (MD) analysis was used to compare continuous outcomes between the group with an IONM alert and those without. Significance was established at P < .05.

Heterogeneity Assessment

The Cochran Q test (χ2) was utilised for heterogeneity assessment and an I² value was calculated to quantify inconsistency. I² was interpreted as follows: 0% to 25% = low heterogeneity, 25% to 75% = moderate heterogeneity, 75% to 100% = high heterogeneity.

Sensitivity Analysis

A sensitivity analysis was performed to investigate the impact of individual studies on the significance of a given forest plot for a given outcome and ensure results were not skewed one-way by an outlier study. For each forest plot, 1 study was excluded from the plot at any 1 time to see how that study changes the overall significance. Individual studies or those with a high risk of bias did not independently affect the significance of the data. Additionally, funnel plots were used to assess for any study outliers.

Quality Assessment of Included Studies

As seen in Table 1, quality of the non-randomized studies was assessed utilising the Newcastle Ottawa Scale which uses a star system to analyse selection, comparability and outcome.¹³

Table 1.

Newcastle-Ottawa Scale to Assess the Quality of Non-Randomised Studies.^{10,11,13-18,20-23}

Study	Selection	Comparability	Outcome	Quality
Gupta et al²³	****	**	***	Good
Huang et al²³	****	**	***	Good
Illingworth et al²³	****	**	***	Good
Keshavarzi et al²³	****	**	**	Good
Lee et al²³	****	**	***	Good
Lewis et al¹⁹	****	**	***	Good
Mathew et al²¹	****	**	***	Good
Puvanesarajah et al²²	****	**	***	Good
Samdani et al¹⁶	****	**	***	Good
Sielatycki et al²⁰	****	**	***	Good
Wang et al¹⁶	****	**	***	Good

Results

Search results – PRISMA Our search strategy retrieved a total of 212 studies and after a thorough screening of the retrieved articles the authors identified 11 studies in total which met the eligibility criteria (Figure 1).

Figure 1.

Prisma flow diagram. The PRISMA diagram details the search and selection processes applied during the overview. PRISMA, preferred reporting items for systematic reviews and meta-analyses.¹²

Baseline Characteristics of Included Studies

11 studies with a total of 3040 patients were included. A range of thoracic deformities and aetiologies including AIS, syndromic, neuromuscular, congenital scoliosis were present. Surgical techniques included PSF, vertebral column resection and posterior spinal osteotomy. Demographic data can be seen in Table 2.

Table 2.

Baseline Demographics of Studies Included Studies.^11,14-23

Study	Mean Age	Gender	Population	Surgical Technique	Total Number Patients	IONM Alerts	No IONM Alerts
Study	(SD)	(M: F)	(Types of Scoliosis)	Surgical Technique	N	N	N
Gupta et al²³	NA	NA	AIS/IKS/CS/CKS/CK	PSF/VCR/PSO	291	108	183
Huang et al²³	22.6 (NA)	45:37	IS/CS/NM/SYN	PSF/VCR	82	27	55
Illingworth et al²³	13.7 (3)	96:157	AIS/NM/CS/SYN	PSF	253	47	206
Keshavarzi et al²³	13.9 (2.2)	18:137	AIS	PSF	155	12	143
Lee et al²³	15.4 (3.2)	108:916	AIS	PSF	1024	21	1003
Lewis et al¹⁹	14.9 (NA)	26:71	AIS/SYN	PSF/PSO	97	27	70
Mathew et al²¹	13.8 (2.9)	31:59	AIS	PSF	90	31	59
Puvanesarajah et al²²	28 (17)	30:38	AIS	PSF/VCR	68	21	47
Samdani et al¹⁶	NA	134:542	AIS	PSF	676	36	640
Sielatycki et al²⁰	41.3 (20.3)	39:89	AIS	PSF/PSO/VCR	128	21	107
Wang et al¹⁶	25.9 (NA)	74:102	AIS	VCR	176	36	140

Abbreviations: AIS, adolescent idiopathic scoliosis; CS, congenital scoliosis; CK, congenital kyphosis; KYP, kyphosis; KYS, kyphoscoliosis; NM, neuromuscular; NF, neurofibromatosis; SYN, syndromic; TB, tuberculosis; VCR, vertebral column resection; PSF, posterior spinal fusion; PSO, posterior spinal osteotomy.

Primary Outcomes - Spinal Cord Classification

Three studies including 373 patients assessing the classification system were found (Table 3).

Table 3.

Comparing Rate of IONM Alerts and Radiographic Outcomes in Different Spinal Cord Types.^11,17,20

Outcome	Type	Sielatycki et al²⁰	Mathew et al²⁰	Keshavarzi et al²³
Proportion of alerts	1	6/81 (7%)	1/41 (2.4%)	1/56 (1.7%)
	2	4/32 (12.5%)	2/28 (7.1%)	3/58 (5.1%)
	3	12/15 (80%)	12/21 (57.1%)	8/41 (19.5%)
Mean PA cobb angle (SD)	1	59 (18)	56 (11.1)	53.2 (11.2)
	2	74 (21)	61 (14.3)	58.2 (13.6)
	3	109 (39)	61 (18.9)	64.1 (16.8)
Sagittal cobb/kyphosis (SD)	1	NA	25 (20.4)	NA
	2	NA	30 (12.8)	NA
	3	NA	33 (16.8)	NA
C-DAR (SD)	1	6 (4.1)	8.3 (1.77)	7.8 (1.8)
	2	9.5 (4.8)	9 (1.7)	8.3 (2.1)
	3	16.8 (11.3)	9.7 (2.82)	9.1 (2.7)
Mean S-DAR (SD)	1	11.5 (6.5)	3.2 (2.2)	2.6 (1.1)
	2	16.8 (7.9)	4 (1.8)	2.6 (1.1)
	3	16.7 (7.1)	4.5 (2.6)	2.9 (1.4)
Mean T-DAR (SD)	1	NA	11.3 (3.1)	NA
	2	NA	13.0 (2.6)	NA
	3	NA	14.1 (3.1)	NA

Type 1 vs Type 3

This included 3 studies with a total of 255 patients (Figure 2A). The odds of IONM alerts in type 1 patients were significantly lower relative to type 3 cords (OR = .03, CI = .01-.08, P < .00001). Between the studies, heterogeneity was low (I² = 0%, P = .55).

Figure 2.

(A) Forest plot comparing odds of IONM alerts in type 1 vs type 3 patients. (B) Forest plot comparing odds of IONM alerts in type 1 vs type 2 patients. (C) Forest plot comparing odds of IONM alerts in type 2 vs type 3 patients. (D) Forest plot comparing odds of IONM alerts in type 3 vs non-type 3 patients.

Type 1 vs Type 2

This included 3 studies with a total of 296 patients (Figure 2B). No significant difference in terms of odds of IONM alerts was seen between type 1 and type 2 patients (OR = .46, CI = .16-1.30, P = .14). Heterogeneity across studies was low (I² = 0%, P = .89).

Type 2 vs Type 3

Type 2 cords were found to have lower odds of IONM alerts relative to type 3 cords (OR = .08, CI = .03, P < .00001) (Figure 2C). Heterogeneity between studies was low (I² = 36%, P = .21).

Non-Type 3 vs Type 3

As shown in Figure 2D, a non-type 3 cord has significantly lower odds of IONM alerts relative to type 3 cords (OR = .05, CI = .02-.16, P < .00001). Heterogeneity was moderate (I² = 52%, P = .12).

Secondary Outcomes

Secondary outcomes involved radiographic predictors of IONM alerts (Table 4) and operative factors (Table 5). Table 6 compares the combined weighted means between alert and non-alert groups.

Table 4.

Comparison of Alert vs Non-Alert Group in Radiographic Outcomes Between Studies.^14,16,18-22

Study	Number of Patients		% of Alert Patients With PND	Mean Coronal Cobb Angle (SD)	FI (SD)	Mean Sagittal Cobb Angle (SD)	Mean CDAR (SD)	Mean SDAR (SD)	Mean TDAR (SD)
Gupta et al²³	Alert	108	NA	92 (39)	NA	101.4 (38)	12.4 (7.2)	16.8 (9.4)	NA
Gupta et al²³	No alert	183	NA	88 (41)	NA	87.5 (46)	12 (7.2)	12.7 (8.4)	NA
Huang et al²³	Alert	27	40.7% (11/27)	135 (23.5)	12.4 (8.9)	135.8 (25.4)	NA	NA	NA
Huang et al²³	No alert	55	40.7% (11/27)	115 (27.6)	15 (8.7)	116.9 (27.3)	NA	NA	NA
Illingworth et al²³	Alert	47	6.38% (3/47)	86 (33)	NA	64 (30)	12 (11)	8 (4)	20 (6)
Illingworth et al²³	No alert	206	6.38% (3/47)	80 (24)	NA	51 (30)	11 (4)	6 (4)	17 (5)
Mathew et al²¹	Alert	36	8.33% (3/36)	65.6 (15.7)	NA	36 (17.7)	10.3 (2.2)	5.1 (2.6)	15.4 (3)
Mathew et al²¹	No alert	64	8.33% (3/36)	57.3 (13.6)	NA	26.7 (17.3)	8.5 (1.9)	3.4 (2)	11.8 (3.1)
Samdani et al¹⁶	Alert	36	2.78% (1/36)	61 (13)	NA	24.6 (16.2)	NA	NA	NA
Samdani et al¹⁶	No alert	640	2.78% (1/36)	55 (12)	NA	22.8 (13.7)	NA	NA	NA
Wang et al¹⁶	Alert	41	7.32% (3/41)	79.4 (49)	NA	109.3 (33.9)	12.1 (8)	16.8 (7.9)	78.9 (NA)
Wang et al¹⁶	No alert	140	7.32% (3/41)	57.2 (43.3)	NA	87.1 (35.5)	7.9 (6.2)	11.5 (6.5)	19.3 (NA)
Lewis et al¹⁹	Alert	27	0% (0/27)	84.4 (19.8)	NA	38.7 (23.7)	NA	NA	NA
Lewis et al¹⁹	No alert	70	0% (0/27)	76.6 (15.9)	NA	39.7 (20.4)	NA	NA	NA
Lee et al²³	Alert	21	4.76% (1/21)	83.4 (26.5)	39.1 (18.3)	16.1 (19.4)	NA	NA	NA
Lee et al²³	No alert	1003	4.76% (1/21)	58.8 (8.9)	43.9 (14.8)	12.9 (9.1)	NA	NA	NA
Puvanesarajah et al²²	Alert	21	14.3% (3/21)	NA	NA	NA	11.4 (7.1)	14.6 (5)	26.6 (9.8)
Puvanesarajah et al²²	No alert	47	14.3% (3/21)	NA	NA	NA	21 (8.5)	12.2 (6.4)	8.8 (4.7)
Sielatycki et al²⁰	Alert	22	0% (0/22)	NA	NA	NA	NA	NA	NA
Sielatycki et al²⁰	No alert	106	0% (0/22)	NA	NA	NA	NA	NA	NA
Keshavarzi et al²³	Alert	12	NA	NA	NA	NA	NA	NA	NA
Keshavarzi et al²³	No alert	143	NA	NA	NA	NA	NA	NA	NA

Abbreviations: FI, flexibility index; PND, postoperative neurologic deficit.

Table 5.

Comparison of Clinical Factors Between Patients With and Without IONM Alerts.

Study	Number of Patients (N)		Mean Blood Loss (mL) (SD)	Mean Operation Duration (mins) (SD)	Mean Number of Levels Fused (SD)	Mean Number of Levels Resected (SD)
Wang et al²³	Alert	36	1253 (727)	NA	NA	1.6 (.6)
Wang et al²³	No alert	140	1620 (1135)	NA	NA	1.3 (.5)
Samdani et al²²	Alert	36	1857 (1323)	357 (157)	NA	NA
Samdani et al²²	No alert	640	999 (796)	298 (117)	NA	NA
Huang et al¹⁵	Alert	27	3126 (1592)	532 (141)	NA	1.4 (.6)
Huang et al¹⁵	No alert	55	2805 (1230)	492 (81)	NA	1.3 (.5)
Lewis et al¹⁹	Alert	27	NA	NA	13.4 (1.2)	4.8 (1.5)
Lewis et al¹⁹	No alert	70	NA	NA	12.7 (1.7)	3.6 (1.1)
Lee et al¹⁸	Alert	21	728.6 (524.8)	264.3 (113.6)	12.1 (1.6)	NA
Lee et al¹⁸	No alert	1003	380.4 (215.6)	212.9 (30.8)	10.9 (1.7)	NA

Table 6.

Combined Weighted Means for Alert vs Non-Alert Patients.

Outcome	Weighted Mean for Alert Groups	Weighted Mean for Non-Alert Groups	Statistically Significant Difference (Yes/No)
PA cobb angle	82.9	72.2	YES
Sagittal cobb angle	60.7	51.5	YES
FI	18.0	21.1	NO
C-DAR	11.5	10.2	NO
S-DAR	10.7	7.96	YES
T-DAR	17.9	14.5	YES
EBL (mL)	1619	1345	NO
OD (mins)	370	319	YES
No of levels fused	12.8	11.9	YES
No of levels resected	2.24	1.81	YES

Coronal (PA) Cobb Angle

This analysis included 8 studies with a total of 2689 patients (Figure 3A). Coronal Cobb angle was found to be significantly greater in the IONM alert cohort (MD = 10.66, CI = 5.77-15.56, P < .00001) and study heterogeneity was moderate (I² = 57%, P = .02).

Figure 3.

(A) Forest plot of coronal (PA) cobb angle for patients with and without intraoperative neuromonitoring alerts. (B) Forest plot of sagittal cobb angle (kyphosis) for patients with and without for patients with and without intraoperative neuromonitoring alerts. (C) Forest plot of total-deformity angle ratio (T-DAR) for patients with and without intraoperative neuromonitoring loss.

Thoracic Kyphosis

Figure 3B assesses thoracic kyphosis between the 2 groups based on 8 studies enrolling a total of 2689 patients. Thoracic kyphosis was significantly higher in patients with IONM alerts (MD = 9.27, CI = 3.28-14.73, P = .0009) and study heterogeneity was moderate (I² = 65%, P = .0009).

Flexibility Index (%)

FI was reported in 2 studies with a total of 1106 patients as shown in Figure 3C. No significant difference was seen in mean difference (MD = −3.06, CI = −6.68-.55, P = .10). Heterogeneity was low between studies (I² = 0%, P = .63).

Coronal-Deformity Angle Ratio (C-DAR)

Figure 4A compares C-DAR between patients with and without IONM alerts. This outcome was reported in 5 studies including a total of 878 patients. The MD between the 2 groups was not found to be significantly significant (MD = 1.33, CI = −.16-2.83, P = .08). A medium level of heterogeneity was found amongst studies (I² = 58%, P = .05).

Figure 4.

(A) Forest plot of coronal-deformity angle ratio (C-DAR) for patients with and without intraoperative neuromonitoring loss. (B) Forest plot of sagittal-deformity angle ratio (S-DAR) for patients with and without intraoperative neuromonitoring loss. (C) Forest plot of total-deformity angle ratio (T-DAR) for patients with and without intraoperative neuromonitoring loss.

Sagittal-Deformity Angle Ratio (S-DAR)

In Figure 4B, S-DAR was reported in 5 studies enrolling a total of 878 patients. A statistically significant MD was seen showing more severe S-DAR and sagittal deformity in patients with IONM alerts (MD = 2.76, CI = 1.57-3.96, P < .00001). Medium heterogeneity was present between studies (I² = 50%, P = .09).

Total-Deformity Angle Ratio (T-DAR)

T-DAR was reported in 3 studies with a total of 411 patients as shown in Figure 4C. T-DAR in IONM alert patients was significantly higher as shown by the mean difference (MD = 3.44, CI = 2.27-4.462, P < .00001). Heterogeneity was low between studies (I² = 0%, P = .7).

Estimated Blood Loss (Milliliters)

EBL analysis was included in 4 studies with a total of 1958 patients as shown in Figure 5A. No significant difference was seen in EBL between the groups (MD = 274.13, CI = −240.03-788.28, P = .30). Heterogeneity was high between studies (I² = 87%, P < .0001).

Figure 5.

(A) Forest plot of estimated blood loss for patients with and without intraoperative neuromonitoring loss. (B) Forest plot of operation duration for patients with and without intraoperative neuromonitoring loss. (C) Forest plot for number of fusion levels for patients with and without intraoperative neuromonitoring loss. (D) Forest plot for number of resected levels for patients with and without intraoperative neuromonitoring loss.

Operation Duration (Minutes)

Operation duration was reported in 3 studies with a total of 1782 patients as shown in Figure 5B. Operation duration in IONM alert patients was significantly higher as shown by the mean difference (MD = 50.79, CI = 20.58-81.00, P = .0010). Heterogeneity was low between studies (I² = 0%, P = .89).

Number of Levels Fused

Number of fused levels was reported in 2 studies with a total of 1121 patients as shown in Figure 5C. IONM alert patients had a significantly greater number of fused levels shown by the mean difference (MD = .92, CI = .43-1.41, P = .0002). Heterogeneity was low between studies (I² = 12%, P = .29).

Number of Levels Resected (Complete or Partial)

Number of resected levels was reported in 3 studies with a total of 355 patients as shown in Figure 5D. This involved a posterior column osteotomy in Lewis et al and a vertebral column resection in the other 2 studies.¹⁹ There was a significantly greater number of levels resected in the IONM alert patients as shown by the mean difference (MD = .43, CI = .01-.84, P = .05). Heterogeneity was high between studies (I² = 80%, P = .006).

Quality Assessment Results

Overall, all studies were of high quality according to the Agency for healthcare and research standards. All studies showed a high quality of patient selection as well as clear inclusion and exclusion criteria. Patients originated from the same population with high comparability between groups and standardized patient factors. Attrition bias was low as all patients continued till the end of the study.

Discussion

This meta-analysis shows that type 1 cord have lower odds of IONM alerts relative to type 3 patients. It also shows that similar to type 1 cords, type 2 cords reduce the odds of IONM alerts. Radiographically a number of measures including PA and lateral cobb angle as well as deformity angle ratio were associated with IONM alerts. Operatively, number of levels of intervention and operation duration were significant.

Our findings support those by Sieletycki et al with regards to type 1 and type 3 patients. Sieletycki et al failed to show any significant association regarding type 2 patients, however the pooled analysis demonstrates the decreased odds of alerts in this cohort. Type 3 cords deserve consideration for their higher risk of spinal cord injury during scoliosis correction. Their morphology changes along the curve apex, where the apical concave pedicle may be associated with a poor blood supply to the cord and further exacerbated with intraoperative corrective manoeuvres.¹¹ Additionally, spinal dura mater adhesions may result in lower spinal cord mobility during curve correction hence, a greater likelihood of cord injury.¹¹ Also, many type 3 patients were non-AIS patients who are more prone to complications like blood loss and increased operative duration.²⁴

The strength of this classification system is that it is practical and easy to use. It has shown strong associations with a range of other radiographic measures and can be used in conjunction with other predictors to form a clear image of a patient’s surgical risk.¹⁶ However, it does require patients to have an MRI which depending on the local guidelines may not be routine practice.²⁵ Additionally, IONM alert cases can still occur with type 1 and type 2 patients meaning that although this system is predictive for type 3 patients, another way of predicting alerts in non-type 3 patients is needed. Lastly, many curves, especially degenerative ones have a lumbar apex where the spinal cord ends meaning that this classification cannot be used.

Generally, greater spinal deformity represented by higher radiographic parameters correlated with greater odds of IONM alerts. This included Coronal Cobb angle, thoracic kyphosis, S-DAR and T-DAR. C-DAR and FI were not significant for IONM alerts. Keshavarzi et al, which was the second validation study showed that 50% of patients with a mean thoracic PA cobb angle of 65° or more and a type 3 cord developed IONM alerts. On the other hand, only 13% of patients with a main Coronal cobb angle of 65° or more and without a type 3 spinal cord developed IONM alerts (P = .023).¹⁷ This indicates that having a type 3 cord alone is an accurate predictor of IONM alerts irrespective of cobb angle. It also indicates that Coronal CA alone may not be a reliable predictor of IONM alerts.

The DAR looks at how acutely angulated the CAs are.²³ The relationship between C-DAR and IONM alerts remains controversial. Wang et al found that patients undergoing a vertebral column osteotomy with a C-DAR more than 10 had a greater incidence of IONM alerts (32.9% vs 12.3%).²³ However, this finding was not supported by other studies .¹⁶ Lewis et al 2015 compared patients undergoing posterior column osteotomy who had IONM alerts to those without and found no significant difference between the groups in terms of either Coronal CA or C-DAR.¹⁰ Illingworth et al looked at C-DAR with respect to PSF patients showed no association between C-DAR with IONM alerts.¹⁶ All studies looking at C-DAR in different spinal cord types demonstrated an increase in C-DAR with more severe spinal cord types.

Thoracic kyphosis, S-DAR and T-DAR were all significant factors that relate to IONM alerts. Mathew et al did not demonstrate a significant difference in kyphosis when directly comparing patients with and without IONM alerts; however, the P value was very close to statistical significance (P = .051).²⁰ Another study, did in fact show kyphosis to be significant between alert and non-alert patients (P = .004) which agrees with our meta-analysis.¹⁴ Illingworth et al found that patients with an S-DAR > 7 had significantly greater rates of IONM alerts than those with lower S-DARs.¹⁶ Overall, there is support for the role of sagittal plane deformity as an important risk factor for IONM alerts.

Wang et al showed that 41.1% of patients with a T-DAR greater than 25 had IONM alerts relative to 10.8% in those with lower T-DARs.²³ A recent study showed that 44% of patients undergoing PSF with a T-DAR >27, developed IONM alerts relative to 17% in those with lower T-DARs.¹⁶ Similarly, the high prevalence of IONM alerts in patients with a high T-DAR is also confirmed by our own results. In our review, the weighted averages of TDAR 17.9 in the alert group compared to 14.5 in the non-alert patients. Interestingly, 3-Dimensional calculation of T-DAR using CT scan was found to be more sensitive and specific than radiographic DAR however, this may expose patients to unnecessary radiation.²¹ A study comparing the rate of IONM alerts in patients undergoing PSF and vertebral column osteotomy found that both groups had an IONM alert rate of around 20%.²³ Despite the similar rate of IONM alerts in PSF and vertebral column osteotomy, neurological deficits occur at a much higher rate of 6-13% in VCR relative to PSF.¹⁵ This may be a result of greater spine instability and limited management options for IONM alerts in VCR.

Surgically, operation duration and number of levels of fusion or resection were significantly higher in the IONM alert group but EBL was not. A consensus-based checklist has been designed to help the surgical team manage cases of IONM systematically addressing possible causes such as hypotension, anaesthesia, hypothermia and cord compression.²⁶ Poor cord perfusion is a leading cause which is why correcting this is often the first step in restoring IONM alerts. This involves increasing mean arterial pressure to more than 80 or 90 millimetres of mercury, increasing inspired oxygen concentration, warming the patient and blood transfusion if haemoglobin is low.⁴ Raising blood pressure alone was found to restore IONM signals in in 20% of cases; IONM signals were restored in 60% of patients using a combination of measures.²⁷

Longer operative times and greater blood loss may increase susceptibility to cord hypoperfusion.²² Proportion of curve correction was not significant for IONM alerts.²² Although early SCI due to bleeding and cord ischaemia causes MEP abnormalities, blood pressure readings may not always differ.²⁸ Several studies also demonstrate a strong relationship between number of levels fused or resected and IONM alerts.^10,19 IONM alerts may also be associated with specific surgical techniques. Ponte osteotomy during AIS correction surgery has been linked to greater risk of IONM alerts.²⁹ Feng et al further supported this and identified intraoperative traction as a risk factor for IONM alerts suggesting that such interventions increase the applied forces on the spine and the risk of SCI.³⁰

The clear association of type 3 cords with IONM alerts, as well as other radiographic parameters, means that surgeons must consider how to best prepare for surgery in high-risk patients. The need to reduce deformity correction or use halo traction to gradually reduce the deformity may need to be considered.²⁰ Additionally, surgeons need to be skilled at and ready to perform spinal cord decompression using either pediculectomy or vertebral body resection. In the absence of the necessary skillset, the need to refer patients to more specialised tertiary deformity centres must be acknowledged.²⁰

To the best of our knowledge, this is the first systematic review and meta-analysis to consider clinical and radiographic predictors associated IONM alerts in scoliosis deformity surgery. This review will aid the establishment of further prospective studies and development of guidelines to manage patients with different radiological profiles. On the other hand, limitations do exist. Only 1 out of the 11 studies was a prospective observational study. Additionally, slight variation between studies in patient populations, anaesthesia settings, IONM guidelines, surgical technique and implant size exists between studies. With regards the classification system, only 3 studies assessing this were available. Lastly, assessment of clinical factors is limited as very little data for this was available.

Conclusion

Greater deformity represented by greater radiological deformity measures is associated with greater incidence of IONM alerts. The spinal cord classification is simple, practical and effective. It is evident that type 1 and 2 cords have significantly lower odds of IONM alerts relative to type 3 cords. However, it is important to consider limitations of the classification system including the incidence of alerts in type 1 and 2 patients, need for MRI and futility in curves with lumbar apices. Overall, the classification is a useful tool that can help surgeons better assess the risk of surgery, take necessary preoperative precautions and better counsel patients and/or their families; especially when other radiographic measures are considered simultaneously. It is important to consider the development of clear guidelines relating to preoperative risk stratification, patient counselling and management strategies for high risk patients.

Footnotes

Author Contributions

Data Acquisition, Analysis and Interpretation of Data: AR and AZ. Study Concept and Design: AR. Study supervision: DC, AA, AC, MAB, AS, RS. Manuscript drafting: All authors contributed to drafting the manuscript; all authors read and approved the final manuscript.

Declaration of Conflicting Interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The author(s) received no financial support for the research, authorship, and/or publication of this article.

Data Availability Statement

The datasets generated and analysed during the current study are available from the corresponding author on reasonable request.

ORCID iDs

Abdulrahman O. Al-Naseem

Derek T. Cawley

Appendix

References

Hume

. Scoliosis: to screen or not to screen. British J School Nurs. 2008;3(5):214-218. doi:10.12968/bjsn.2008.3.5.31130

Janicki

Alman

. Scoliosis: review of diagnosis and treatment. Paediatr Child helath. 2007;12(9):771-776. doi:10.1093/pch/12.9.771

Maruyama

Takeshita

. Surgical treatment of scoliosis: a review of techniques currently applied. Scoliosis. 2008;3:6. doi:10.1186/1748-7161-3-6

Devlin

Schwartz

. Intraoperative neurophysiologic monitoring during spinal surgery. J Am Acad Orthop Surg. 2007;15(9):549-560. doi:10.5435/00124635-200709000-00005

Murphy

Mooney

3rd . Complications following spine fusion for adolescent idiopathic scoliosis. Curr Rev Musculoskelet Med. 2016;9(4):462-469. doi:10.1007/s12178-016-9372-5

Quinonez

Pahys

Samdani

Hwang

Cahill

Betz

. Complete paraplegia 36 h after attempted posterior spinal fusion for severe adolescent idiopathic scoliosis: a case report. Spinal Cord Ser Cases. 2021;7(1):33. doi:10.1038/s41394-021-00386-6

Rafter

Vasdev

Hurrelbrink

, et al. Litigation risks despite guideline adherence for acute spinal cord injury: time is spine. Neurosurg Focus. 2020;49(5):E17. doi:10.3171/2020.8.Focus20607

Biscevic

Sehic

Krupic

. Intraoperative neuromonitoring in spine deformity surgery: modalities, advantages, limitations, medicolegal issues - surgeons’ views. EFORT Open Rev. 2020;5(1):9-16. doi:10.1302/2058-5241.5.180032

Charalampidis

Jiang

Wilson

JRF

Badhiwala

Brodke

Fehlings

. The use of intraoperative neurophysiological monitoring in spine surgery. Global Spine J. 2020;10(1 Suppl):104s-114s. doi:10.1177/2192568219859314

10.

Lewis

Keshen

Lenke

, et al.

The deformity angular ratio: does it correlate with high-risk cases for potential spinal cord monitoring alerts in pediatric 3-column thoracic spinal deformity corrective surgery?

Spine (Phila Pa 1976). 2015;40(15):E879-E885. doi:10.1097/brs.0000000000000984

11.

Sielatycki

Cerpa

Baum

, et al. A novel MRI-based classification of spinal cord shape and CSF presence at the curve apex to assess risk of intraoperative neuromonitoring data loss with thoracic spinal deformity correction. Spine Deform. 2020;8(4):655-661. doi:10.1007/s43390-020-00101-9

12.

Moher

Liberati

Tetzlaff

Altman

. Preferred reporting items for systematic reviews and meta-analyses: the PRISMA statement. Ann Intern Med. 2009;151(4):264-269. doi:10.7326/0003-4819-151-4-200908180-00135

13.

Wells

Shea

O’Connell

, et al. The Newcastle-Ottawa scale (NOS) for assessing the quality of nonrandomized studies in meta-analyses. https://www.ohri.ca/programs/clinical_epidemiology/oxford.html. Accessed August 10th 2023.

14.

Gupta

Lenke

Gupta

, et al. Intraoperative neuromonitoring predicts postoperative deficits in severe pediatric spinal deformity patients. Spine Deform. 2023;12(1):109-118. doi:10.1007/s43390-023-00745-3

15.

Huang

Z-F

Chen

Yang

J-F

Deng

Y-L

Sui

W-Y

Yang

J-L

. Multimodality intraoperative neuromonitoring in severe thoracic deformity posterior vertebral column resection correction. World Neurosurg. 2019;127:e416-e426. doi:10.1016/j.wneu.2019.03.140

16.

Illingworth

Siddiqui

Skaggs

Andras

. Deformity angular ratio is associated with neuromonitoring changes without a vertebral column resection: spinal deformity is more influential than type of surgery. Spine Deform. 2023;11(4):951-956. doi:10.1007/s43390-023-00669-y

17.

Keshavarzi

Ramchandran

Spardy

Dobyns

Errico

George

. Utilization of pre-operative MRI to identify AIS patients at highest risk of intra-operative neuromonitoring alert. Spine Deformity. 2023;11(6):1419-1426. doi:10.1007/s43390-023-00710-0

18.

Lee

Hwang

Lee

Cho

Park

. Risk factors and exit strategy of intraoperative neurophysiological monitoring alert during deformity correction for adolescent idiopathic scoliosis. Global Spine J. 2023;14:21925682231164344. doi:10.1177/21925682231164344

19.

Lewis

Wong

IHY

Strantzas

, et al. Responding to intraoperative neuromonitoring changes during pediatric coronal spinal deformity surgery. Global Spine J. 2019;9(1 Suppl):15s-21s. doi:10.1177/2192568219836993

20.

Mathew

Milbrandt

Shaughnessy

Stans

Larson

. Is the axial spinal cord classification predictive of intraoperative neurologic alert for pediatric scoliosis patients? An independent retrospective validation study. Spine Deform. 2021;9(2):395-401. doi:10.1007/s43390-020-00241-y

21.

Puvanesarajah

Marciano

Hassan

, et al.

The deformity angular ratio: can three-dimensional computed tomography improve prediction of intraoperative neuromonitoring events?

Spine Deform. 2022;10(5):1047-1053. doi:10.1007/s43390-022-00518-4

22.

Samdani

Bennett

Ames

, et al.

Reversible intraoperative neurophysiologic monitoring alerts in patients undergoing arthrodesis for adolescent idiopathic scoliosis: what are the outcomes of surgery?

J Bone Joint Surg Am. 2016;98(17):1478-1483. doi:10.2106/jbjs.15.01379

23.

Wang

Lenke

Thuet

Blanke

Koester

Roth

. Deformity angular ratio describes the severity of spinal deformity and predicts the risk of neurologic deficit in posterior vertebral column resection surgery. Spine (Phila Pa 1976). 2016;41(18):1447-1455. doi:10.1097/brs.0000000000001547

24.

Levy

Schulz

Fornari

Wollowick

. Complications associated with surgical repair of syndromic scoliosis. Scoliosis. 2015;10:14. doi:10.1186/s13013-015-0035-x

25.

Bettany-Saltikov

. Imaging in the diagnosis and monitoring of children with idiopathic scoliosis. Open Orthop J. 2017;11:1500-1520. doi:10.2174/1874325001711011500

26.

Vitale

Skaggs

Pace

, et al. Best practices in intraoperative neuromonitoring in spine deformity surgery: development of an intraoperative checklist to optimize response. Spine Deform. 2014;2(5):333-339. doi:10.1016/j.jspd.2014.05.003

27.

Yang

Skaggs

Chan

, et al. Raising mean arterial pressure alone restores 20% of intraoperative neuromonitoring losses. Spine (Phila Pa 1976). 2018;43(13):890-894. doi:10.1097/brs.0000000000002461

28.

Lewis

Gray

Holmes

, et al. Neurophysiological changes in deformity correction of adolescent idiopathic scoliosis with intraoperative skull-femoral traction. Spine. 2011;36(20):1627-1638. doi:10.1097/BRS.0b013e318216124e

29.

Harfouch

Bunyan

Faraidy

Alnemari

Bashir

. Ponte osteotomies increase risk of intraoperative neuromonitoring alerts in adolescent idiopathic scoliosis surgery. Surg Neurol Int. 2022;13:154. doi:10.25259/sni_67_2022

30.

Feng

Qiu

Shen

, et al. Impact of multimodal intraoperative monitoring during surgery for spine deformity and potential risk factors for neurological monitoring changes. J Spinal Disord Tech. 2012;25(4):E108-E114. doi:10.1097/BSD.0b013e31824d2a2f