The normality of data was tested by the Shapiro-Wilk test. Normally distributed data were reported as the mean±standard deviation. Non-parametric data were reported as median (interquartile range, IQR). Differences between preoperative and postoperative parameters and between different imaging modalities were compared by the paired t-test. The relationship between MRI parameters and cage position was tested by linear regression analysis. Independent variables with a p-value <0.10 in simple linear regression were further analyzed in a multiple linear regression model. Statistical significance was considered if the p-value <0.05. All tests were two-tailed. Statistical analysis was done using IBM SPSS version 21 (IBM Corp., Armonk, NY, USA).

Table 4 shows the comparison between preoperative and postoperative imaging parameters. In the level that used the cages with 6-degree lordosis, SL increased from 6.5±6.6 degrees to 9.4±5.9 degrees. With a 12-degree lordosis cage, SL increased from 9.5±5.8 degrees to 12.3±6.3 degrees. The increment of SL was the same between the levels that used a 6 or 12-degree cage (p=0.92). LL increased by 5.5±6.8 degrees in the single level OLIF group (95% confidence interval [CI], 2.4 to 8.1 degrees) and 8.2±7.0 degrees in multiple level OLIF group (95% CI, 2.9 to 12.4 degrees) without statistical differences between groups (p=0.23). In the single level OLIF group, utilization of 6-degree lordosis cage or 12-degree lordosis cage resulted in similar LL change (5.4±7.2 degrees and 5.8±5.3 degrees respectively, p=0.91).

The left and right FSCA increased by 19.5% and 15.6% in the postoperative CT scan (Table 4). There was no statistical difference between the left and right side FH and FSCA gain postoperatively (p=0.12 and 0.40).

Spinal canal width increased by 28.8%, SCSA increased by 33.3% compared to preoperative MRI. The left and the right FSCA increased by 43.7% and 45.0%, respectively. Percentage change of the left and the right FSCA did not become different (p=0.87). Foraminal stenosis grades were unchanged in 57 foramens, decreased 1 grade in 46 foramens, decreased 2 grades in 18 foramens, increased 1 grade in three foramens, and unmeasurable due to the metallic artifact in 12 foramens. The median of foraminal stenosis grade was reduced from grade 2 to grade 1 (Table 4). SCSA was unmeasurable due to the metallic artifact in 3 levels (6%). The rate of the metallic artifact was not different between fixation by pedicle screws or cortical bone trajectory screws (p=0.50).

Linear regression analysis showed no correlation between cage parameters and SCSA increment (Table 5). Factors associated with SCSA increment were anterior DH gained and preoperative SCSA. Table 6 demonstrated that cage distance and preoperative FSCA were inversely associated with postoperative FSCA increment while DH change was not associated with postoperative FSCA increment in the multiple linear regression analysis. Anterior implantation of the cage is associated with a greater amount of SL correction (Table 7). Linear correlation between cage position and foraminal stenosis decompression, and SL correction was shown in Fig. 5.

The lateral lumbar interbody fusion (LLIF) could be done by trans-psoas approach or ante-psoas approach. These approaches had proved to adequately decompress the central canal, lateral recess, and neural foramen [1,6-8,16,19,21,22,25]. Radiographic parameters showing the indirect decompression by OLIF in our study were similar to previous studies on LLIF, which were measured by the same imaging modality [6,19,21]. However, our study demonstrated a 30-46% difference in the FSCA decompression between the measurements using CT and MRI. The MRI study after LLIF by Park et al. [21] also found that the increment in the FSCA was 69.5%, which was higher than an average increase of 32.5% to 38.9% using plain radiograph or CT [9]. The disparity in the FSCA change reflects the effect of indirect decompression on soft tissues, which were more challenging to detect on CT. This difference between imaging modalities should be considered when comparing results between studies, especially in systematic review and meta-analysis.

The effectiveness of indirect decompression is reflected in postoperative outcomes as 51% improvement in ODI and 73% improvement in NRS for back pain and 67% for leg pain. However, the correlation could not be found between radiographic parameters and clinical outcomes; the finding consistent with previous studies [3,6,17,24].

The factors that influence the indirect decompression of neural foramen are still controversial [9,21,22]. Park et al. [21] and Lin et al. [12] reported that putting LLIF cage into anterior 1/3 or middle 1/3 of the vertebral body resulted in the similar FSCA decompression. Limthongkul et al. [11] divide LLIF cage positions into anterior and posterior groups and also found insignificant differences between groups. The utilization of plain radiographs and dividing cage positions into groups from the aforementioned studies may be inadequate to detect the actual difference. Our study measures the cage position using axial CT, which can pinpoint the center position of the OLIF cages, so we can demonstrate that putting the cage closer to the posterior part of the vertebral body correlated with better FSCA decompression.

Surgeons performing OLIF usually use the corridor between the aorta and psoas muscle and implement the orthogonal maneuver to reduced the cage trajectory angle but the final position of the cage may not parallel to the posterior vertebral body. The oblique position of the cage means the right end of the cage is closer to the neural foramens than the left end, which may result in the asymmetric decompression of neural foramens. Sato et al. [24] reported that, after OLIF, there were 39% and 21% increments of right and left FSCA respectively. However, they did not report the degree of the obliquity of the cage in their series. Limthongkul et al. [11] reported that putting OLIF cages obliquely at 6 degrees does not affect the differences between right and left neural foramen decompression. Our study suggested that even the cages were oblique at the mean of 14 degrees, both sides of the neural foramen could be equally decompressed (Table 4).

Indirect decompression of the central spinal canal by OLIF had been studied extensively by correlating the distraction of the intervertebral disc with SCSA decompression [12,21,24]. Our finding that expansion of DH associated with an increment in central canal decompression was supported by previous studies. The multivariate analysis showed that cage morphology and position were not associated with the percentage change of SCSA, similar conclusions were reported [11,12,21].

Position of the cage in the anterior part of the vertebral body associated with greater correction of SL [12,21]. Our study has similar results and also suggested that each 1 mm of cage position further from the posterior part of the vertebral body correlated with a better 0.3 degree of SL correction.

Limitations of this study are small sample sizes and retrospective review nature. The MRIs, done at the immediate postoperative, could overestimate the efficacy of the indirect decompression because late cage subsidence may occur and could result in decreased central and neural foramen. CTs and MRIs from our study were also not done at the same time, making a parallax in the comparison between these two modalities.