1 Introduction

Lateral interbody fusion by the oblique approach depends upon the presence of a favourable oblique corridor.^1–4 This corridor is a safe area anterior to the intervertebral disc between the anteromedial border of the left psoas muscle and the lateral border of the aortoiliac vessels.^5,6 An unobstructed oblique corridor (OC) visualized in the pre-operative axial cut magnetic resonance image (MRI) is generally considered as a prerequisite for choosing to proceed with oblique lumbar interbody fusion (OLIF).⁷ However, this assessment is subjective and is influenced by experience. A more objective way of assessing the feasibility of OLIF is by using our recently proposed modified Moro's classification of psoas morphology and the OC grading. A detailed description of these systems are available in our original publication (Ng et al. 2020).⁸

In the modified Moro's classification, the anterior zone described in the original Moro's system is further subdivided into 4 zones named AI-AIV which is basically a mirror image of the equally divided zones I-IV (Fig. 1).^8,9 Using this system, we can precisely locate the anterior border of the psoas in one of the zones. Furthermore, the oblique corridor can be divided into 4 grades (grade 0: no measurable corridor, grade 1: ≤ 1 cm; grade 2: >1 to 2 cm; grade 3: >2 cm) depending upon the OC grading.⁸ By incorporating these two systems, we believe it is simple and easy to assess the feasibility of oblique lumbar interbody fusion in an objective manner. However, these systems remain unvalidated as our original proposal lacks a reliability analysis. Thus, we intend to assess the inter and intra-rater reliability of these systems using kappa statistics among spine surgeons at our institute.

Close

Fig. 1 The modified Moro's classification of Psoas morphology. The anterior zone described in the original Moro's system is further subdivided into 4 zones named AI-AIV which is basically a mirror image of the equally divided zones I-IV

Export to PPT

2 Materials and methods

Twenty-five consecutive lumbar spine MRI scans done on patients with degenerative spine conditions were obtained from the radiology department. The axial cut image of a single disc level between L1 to L5 in each MRI scan was selected by a senior spine surgeon in such a way that a) the set includes anterior varieties of psoas as per the modified Moro's classification and b) all four OC grades. We did not include L5-S1 disc level mainly because the conventional anterior to psoas oblique approach is difficult at this level due to the varied anatomy of the iliac vessels and the overriding iliac crest.^10,11

A calibrating session was held to brief the spine surgeons who frequently perform OLIF at our institute about the modified Moro's classification of psoas morphology and the OC grading using our original work.⁸ All queries were addressed, and the 25 selected images were distributed to five spine surgeons. They had to analyze each image to identify, 1) the feasibility of OLIF at the given level based on their perception, 2) the psoas type on the left side according to the modified Moro's classification and 3) the OC grade.

To maintain the precision and consistency of judgment, surgeons were not put under an obligation to time as to when their responses need to be submitted. Once the responses were received, the 25 given levels were divided as favourable or unfavourable for OLIF depending on the majority agreement and the percentage of agreement was calculated. Similarly, the psoas type and OC grade which most of the surgeons agreed for each image was considered as the actual category and the percentage of agreement for each psoas type and OC grade was calculated. In addition, the overall percentage of agreement was calculated.

For analyzing the inter-rater and intra-rater reliability, responses given by surgeons for the psoas type and OC grade were considered as categorical variables. Inter-rater reliability of both the modified Moro's classification and the OC grading was determined by calculating Fleiss' Kappa. In addition, Cohen's Kappa was calculated between all pairs of surgeons and kappa matrices were generated. The images were reassessed after a month by three of the included surgeons. Their test and retest responses were used to determine the intra-rater (test-retest) reliability by calculating Cohen's Kappa.

Statistical analyses were performed using IBM SPSS V. 23.0 (IBM Corp., Armonk, NY, USA). Reliability was assessed by calculating the percentage agreement and the kappa statistic.¹² Kappa was interpreted as per the recommendation of Landis and Koch (0–0.20 = slight agreement, 0.21–0.40 = fair agreement, 0.41–0.60 = moderate agreement, 0.61–0.80 = substantial agreement, 0.81 or higher = excellent agreement and 1 = absolute agreement).¹³ This study with Ref. No. 2019/00352 was reviewed by our national health care group (NHG) domain specific review board (DSRB) and an exemption was granted.

3 Results

A total of 25 lumbar disc level axial cut MR images belonging to different patients were assessed by five spine surgeons. The selected images were predominantly of L4-L5 level; however, there were 2 images each of L2-L3 and L3-L4 levels. Based on surgeon responses, there were 11 and 14 levels deemed favourable and unfavourable for OLIF respectively with an overall agreement of 93.6%.

As per majority agreement, there were 5 type I, 12 type A-I, 5 type A-II and 3 type A-III psoas (Fig. 2); likewise, there were 4 grade 0 OCs, 6 grade 1 OCs, 7 grade 2 OCs and 8 grade 3 OCs (Fig. 3). The overall percentage of agreement based on the modified Moro's classification and the OC grading was 86.4% and 84% respectively. In addition, the percentage of agreement for each psoas type and OC grade was calculated (Table 1).

Close

Fig. 2 Various types of Psoas among the selected images as per the modified Moro's system. A) Type I; B) Type AI; C) Type AII; D) Type AIII.

Export to PPT

Close

Fig. 3 Different grades of oblique corridor (OC) as per the OC grading system. A) Grade 0; B) Grade 1; C) Grade 2; D) Grade 3.

Export to PPT

In order to analyze the interrater reliability of both the modified Moro's classification and the OC grading, we calculated Cohen's Kappa between all pairs of surgeons. This data was used to form the inter-rater kappa matrix for both the systems (Tables 2 and 3). However, as there were multiple raters, Fleiss' kappa was calculated to assess the absolute inter-rater reliability. The inferred Fleiss' kappa of 0.626 for the modified Moro's classification and 0.648 for the OC grading signifies substantial reliability according to the interpretation of Kappa by Landis and Koch (Table 4).

The intra-rater reliability represented by Cohen's kappa was calculated using the test-retest ratings given by three surgeons. The retest ratings were consistent with previous ratings with an agreement of 80%, 76% and 84% for the psoas classification and an agreement of 80%, 80% and 96% for the OC grade. The inferred Cohen's kappa demonstrated substantial intra-rater reliability for the modified Moro's classification and substantial to excellent intra-rater reliability for the OC grading.

On correlating the psoas type and OC grade with surgeon perception of whether the given levels are favourable for OLIF, we noticed that OLIF was considered not possible whenever the psoas was of type A-II or above and the OC was of grade 1 and below. These results signify the reliability and highlight the clinical utility of both the modified Moro's classification and the OC grading on pre-operative planning for OLIF.

4 Discussion

The biomechanical advantage of placing a larger end to end cage via the lateral approach (direct lateral or oblique lateral) and the sagittal correction that can be achieved by using such a cage are proven to be of clinical importance.^14,15 Concerning the oblique approach to the lumbar spine, it may not be feasible on every patient and on every level, as it is subject to the presence of a favourable OC. The OC, being the space between the anteromedial border of the left psoas muscle and the lateral border of the aortoiliac vessels, can be influenced by the normal anatomical variations in these structures. For the surgeon, approaching the OC will entail dealing with both these structures and hence preoperative planning and patient selection are essential to prevent complications.^16,17 For this reason, our recently proposed modified Moro's classification and the OC grading can be of significant use.⁸ Their combined ability to predict the feasibility of OLIF at a particular level in an objective manner can guide in patient selection.

Both the Modified Moro's classification and OC grading are simple, and measurement-based; hence, the selected spine surgeons (raters) were comfortable on using these systems. Moreover, in the calibrating session, they were explained in detail and were allowed to practice on MRI images to refine their understanding of the systems. All measurements were done digitally on our picture archiving and communication system (PACS). This made it easy to equally divide the disc area into zones I-IV and then to similarly divide the anterior zone into AI-AIV as a reflection of zones I-IV. Even though PACS was used for this, we believe the same can be done manually without difficulty using a measurement scale or even subjectively by mere visualization to get a rough idea of whether an oblique approach is feasible at a particular level.

As the ratings given by the surgeons were considered as nominal variables, we used the Fleiss' Kappa (statistical measure for assessing the reliability of agreement between multiple sets of ratings) for calculating inter-rater reliability. For determining intra-rater reliability, we used Cohen's kappa (statistical measure for assessing the reliability of agreement between two sets of ratings) to analyze the test-retest ratings given by the same rater. Based on kappa statistics, we found both the modified Moro's classification and OC grading to be highly reliable and consistent; hence, adoptable for clinical use to assess the feasibility of OLIF. However, regarding clinical adoption, it should be noted that the MRI images are taken in a supine position and during surgery, the patient is in a right lateral decubitus position. This variation in position can move the aortoiliac vessels rightward to increase the OC or the psoas can change shape to decrease the OC which may influence the validity of patient selection using these systems.¹⁸

5 Conclusion

The inter and intra-rater reliability of the modified Moro's classification and OC grading were analyzed among spine surgeons at our institute. Their ratings for selected lumbar axial cut MR images as per these new systems were consistent and had substantial homogeneity, which implies that the systems are highly reliable. This observation indicates that both these systems are of clinical importance and can predict the feasibility of lateral interbody fusion by the oblique approach at lumbar levels in an objective manner to aid in patient selection for this procedure.