## Introduction

Transforaminal lumbar interbody fusion (TLIF) has gained wide popularity among the surgical spine community due to its efficacy, safety, and reproducibility, namely in the treatment of lumbar degenerative diseases that failed conservative treatment. Several published studies favor minimally invasive TLIF (MI-TLIF) regarding intraoperative blood loss, length of stay, complication rates, and clinical outcomes over open TLIF (O-TLIF), despite higher radiation exposure1,2.

Minimally invasive spine surgeries have been developed to reduce tissue trauma, decrease complication rates, and improve functional recovery3,4,5,6,7,8. The advances in endoscopic spine surgery made way for new opportunities to minimize tissue aggression further. Recently published meta-analyses regarding the treatment of lumbar disc herniations favored endoscopic discectomy (ED) over microdiscectomy (MD) in clinical outcomes (Oswestry disability index), duration of surgery, length of hospital stay, and lower risk of overall complications. These results opened the perspective that ED could take over the place of MD as the gold standard of care in the management of lumbar disc disease7,8. Technological innovations in endoscopic spine surgery have widened its range of applications beyond lumbar disc herniations. Endoscopic treatment of central and lateral recess stenosis, as well as endoscopic assisted lumbar interbody fusion (LIF) for degenerative lumbar diseases, are increasingly common9,10.

The proposed benefits of an even less invasive technique than MI-TLIF would be further improvement in the advantages over O-TLIF and obviating general anesthesia. Encouraged by the success of ED, endoscopic assisted intraforaminal LIF (iLIF) has been increasing its popularity. Even though several surgical techniques have been described, Kambin’s triangle approach through an intraforaminal facet sparing technique is the most usual and the one that has greater potential to reduce iatrogenic soft and bone tissue trauma10,11,12,13,14,15. Recent studies have shown promising results regarding reduced blood loss, decreased length of stay, clinical outcomes, complications and fusion10,14. However, the comparison between MI-TLIF and iLIF is sparse in the literature, and concerns about the safety and effective benefits of the endoscopic technique remain unanswered.

This systematic review and meta-analysis of MI-TLIF and iLIF were conducted to synthetize and compare the available data in the literature on clinical outcomes, complications and fusion rates.

## Methods

### Literature research

This review and meta-analysis were performed following the preferred reporting items for systematic reviews and meta-analyses (PRISMA) guidelines16. The study protocol was registered in April 2020 with the PROSPERO international prospective register of systematic reviews (Registration No. CRD42020180980, accessible at https://www.crd.york.ac.uk/prospero/).

Electronic systematic research of MEDLINE, Embase, and Cochrane Library databases was performed to identify all relevant studies published from the date of inception to June 15, 2020. The following search strategy was used: (((“spine fusion” OR “lumbar fusion”) AND (“Endoscopy” OR “Endoscopic”)) OR ((“MI” OR “Minimally invasive”) AND (“TLIF” OR “transforaminal lumbar interbody fusion”))). The language of the included studies was restricted to English.

### Selection criteria and data extraction

For this review, “iLIF” procedures were defined as endoscopic assisted lumbar interbody fusion performed through a uniportal intraforaminal access to Kambin’s triangle11,12,13,17. Intraforaminal access implied that minimal, partial, or total resection of the superior articular process (SAP) was performed to allow disc space preparation and cage deployment, while the inferior articular process (IAP) had to be preserved. Studies reporting percutaneous LIF or endoscopic assisted transforaminal lumbar interbody fusion (TLIF) approaches with complete facetectomy were excluded. “MI-TLIF” procedures were defined as surgery performed through a muscle-sparing surgical corridor, created by serial dilators that allowed for a tubular or cylindrical retractor to be docked on the facet joint complex, as reported by Foley and Schwender18,19. Besides anterior support, both techniques implied supplementary same level screw fixation.

The following inclusion criteria were used: 5 or more consecutive patients who underwent iLIF or MI-TLIF for lumbar degenerative diseases; description of the surgical technique; clinical outcome measures reported at a minimum follow-up of 12 months; complications assessment; and imaging assessment of fusion at a minimum follow-up of 12 months. The corresponding authors of studies with insufficient data (i.e., reported mean, standard deviation (SD), number of subjects or events) to extract and pool the predefined primary endpoints were contacted via email for clarification, if otherwise the studies were excluded. Systematic reviews, meta-analyses, technical notes, surgical techniques, biomechanical studies, case reports, and editorials were also excluded.

Two review authors (J.M.S. and H.R.) independently retrieved and screened all titles and abstracts to determine study eligibility. Full-text articles of the relevant abstracts were reviewed by the same two authors. The reference lists of the eligible studies were hand searched for potentially relevant publications. When studies with overlapping samples and outcomes were identified, only the most complete reports included for analysis. Data extraction of the selected studies was performed independently by two review authors (J.M.S. and H.R.) using a standardized data extraction Microsoft Excel form (Microsoft, Redmond, WA). Study characteristics (number of patients, age, body mass index, disease, follow-up, number of levels operated) and outcomes were extracted. The number of subjects, mean, and SD of continuous variables, and cross tabulated frequencies of dichotomous outcomes were also extracted. The following outcomes of interest were defined: a) primary outcomes: clinical outcomes measures at baseline and last follow-up; overall complications; fusion rate; b) secondary outcomes: average surgical time, intraoperative blood loss and hospital length of stay. Clinical outcomes measures reported in at least three studies of each surgical technique were pooled for meta-analysis. Any disagreements related to study selection or data extraction were settled through discussion and consensus with a third reviewer (J.G.C.).

### Risk-of-bias

Two review authors (J.M.S. and H.R.) independently assessed the methodologic quality of the studies according to the methodological index for non-randomized studies (MINORS)20. Items were scored as 0, 1, or 2, whether they were not reported, reported but inadequate or reported and adequate, respectively. For non-comparative studies eight items were evaluated (maximum score of 16), and for comparative studies all 12 items were scored (maximum score of 24). Non-comparative studies with MINORS score $$\le$$ 12 and comparative studies with MINORS score $$\le$$ 20 were considered at high risk of bias. Any disagreements were resolved through discussion and consensus with a third reviewer (J.G.C.).

### Statistical analysis

Intraclass correlation coefficient (ICC) was calculated to quantify inter-rater reliability of the MINORS scores20,21.

Unstandardized mean differences (UMD) were pooled and calculated for continuous outcomes—Visual analogue scale (VAS) back, VAS leg, ODI, surgical time, intraoperative blood loss, hospital length of stay—as follows:$$UMD\left(di\right)={\overline{x}}_{1i}-{\overline{x}}_{2i}, var\left(di\right)=\frac{{sd}_{1i}^{2}}{{n}_{1i}}+\frac{{sd}_{2i}^{2}}{{n}_{2i}}, {w}_{i}= \frac{1}{var({d}_{i})}$$, where $${w}_{i}$$ is the weighting factor, $$di$$ is the unstandardized difference of means, $${n}_{1i}$$ and $${n}_{2i}$$ are number of subjects in groups 1 and 2, $${n}_{i}$$ is $${n}_{1i}$$ + $${n}_{2i}$$, $${sd}_{i}$$ is the pooled standard deviation, $$var\left(di\right)$$ is the variance of difference, and subscript $$i$$ corresponds to study $$i$$.

Pooled prevalence of the dichotomous outcomes was calculated as: $$\overline{p}=\frac{\sum {w}_{i}{p}_{i}}{\sum {w}_{i}}$$, $${w}_{i}= \frac{1}{var({p}_{i})}$$ , where $$\overline{p}$$ was the pooled prevalence, $${p}_{i}$$ was the prevalence of the event in each study, and $${w}_{i}$$ was the weight of each study22,23.

Heterogeneity was estimated using Q statistics. I2 statistics was used to estimate inconsistency among the studies’ results due to heterogeneity rather than chance. Values greater than 50% were considered as substantial heterogeneity. If heterogeneity was present, between studies variation was estimated by calculating $$\tau^{2}$$, which was then used to calculate $${w}_{i}^{*}=\frac{1}{var\left({p}_{i}\right)+\tau 2}$$, a weight term that accounted for variations between studies. Sub-group analysis was performed to explore causes of heterogeneity24,25.

Baseline to last follow-up mean differences of outcome measures reported in at least three studies in each group were calculated and compared to the minimal clinically important difference (MCID) of each outcome: 1.16 for VAS back, 1.36 for VAS leg, and 12.40 for ODI26.

UMD and odds ratio (OR) were estimated using a random effects model, with a 95% confidence interval. Statistical significance was set at p value > 0.05. Meta-analysis was performed between the pooled studies and the comparative study. Pooled analyses were performed using Microsoft Excel (Microsoft, Redmond, WA), the remaining statistical analyses and forest plots were performed using Cochrane Review Manager, version 5.4 (The Cochrane Collaboration, Copenhagen, Denmark)27.

## Results

The literature search is illustrated in the PRISMA flow diagram (Fig. 1). Forty-two studies remained for qualitative and quantitative analyses14,15,28,29,30,31,32,33,34,35,36,37,38,39,40,41,42,43,44,45,46,47,48,49,50,51,52,53,54,55,56,57,58,59,60,61,62,63,64,65,66,67. The mean MINORS score was 12.7 $$\pm$$ 1.29 (10–14) for non-comparative studies, and 20.9 $$\pm$$ 1.9 (17–24) for comparative studies. Six non-comparative studies (MINORS score $$\le$$ 12) and 13 comparative studies (MINORS score $$\le$$ 20) were assessed as a high risk of bias. The intra-class correlation coefficient (ICC) to quantify the degree of agreement between the raters was 0.997.

The corresponding author of the study by Shen14 was contacted by email and provided study’s mean and SD of operative time, estimated blood loss and length of hospital stay.

### Baseline characteristics

The characteristics of the studies are summarized in Table 1. Thirty-six studies, with a total of 2076 patients, were included in the MI-TLIF group; five studies, with a total of 170 patients, were included in the iLIF group; and one prospective cohort study (PCS) with 75 patients comparing MI-TLIF to iLIF was identified.

The mean age was 57.9 ± 11.0 years and 63.4 ± 10.9 years, for MI-TLIF and iLIF, respectively. The proportion of females was 61% for MI-TLIF, and 52% for iLIF. Thirty-two studies reported single-level surgery, two studies (MI-TLIF) reported two level surgeries, while eight studies (six MI-TLIF; two iLIF) included single-level and two or more levels surgeries.

### Operations parameters

Summary changes of surgical time, blood loss and length of hospital stay are portrayed in Tables 2 and 3.

In the pooled studies, surgical time was significantly increased in the MI-TLIF group (UMD 38.1; 95% CI 33.01; 43.23; p value < 0.0001). However, in the meta-analysis no statistically significant difference was observed between both groups (Table. 4).

The mean intra-operative blood-loss of the pooled studies was significantly increased by 128.8 mL (95% CI 118.01; 139.59; p value < 0.0001) in MI-TLIF. In the meta-analysis, there was also a statistically significant mean increase of intra-operative blood-loss (UMD 110.61 mL; 95% IC 70.43; 150.80; p value < 0.0001) in MI-TLIF (Table. 4, Fig. 2).

In the pooled studies, the length of hospital stay had a significantly mean increase of 2.6 days (95% CI 2.27; 2.95; p value < 0.0001) in the MI-TLIF group. In the meta-analysis, there was also a statistically significant mean difference of 2.36 days (95% IC 1.77; 2.94; p value < 0.0001) favoring the iLIF group (Table. 4, Fig. 2).

### Clinical Outcomes

VAS back (33 MI-TLIF studies, n = 1946 vs 4 iLIF studies, n = 70), VAS leg (26 MI-TLIF studies, n = 1518 vs 3 iLIF studies, n = 52), and ODI (36 MI-TLIF studies, n = 2050 vs 6 iLIF studies, n = 170) were reported in three or more studies in each technique and are summarized in Tables 5 and 6. UMD of the pooled results and meta-analysis are summarized in Table 7.

There was no difference in baseline VAS back between pooled MI-TLIF and iLIF groups. At the last follow-up, VAS back was significantly lower in the iLIF group (UMD 0.96; 95% CI 0.81; 1.11; p value < 0.0001). However, the meta-analysis showed no statistically significant difference at either time-point.

VAS leg at baseline was significantly higher in the iLIF group (UMD 1.0; 95% CI 0.73; 1.29; p value < 0.0001), with significant heterogeneity. This statistically significant increase was also observed in the meta-analysis (UMD 0.86; 95% CI 0.37; 1.34; p value = 0.0005). At the last follow-up, VAS Leg was significantly lower in the pooled iLIF group (UMD 0.54; 95% CI 0.38; 0.79; p value = 0.027). However, the meta-analysis showed no statistically significant difference at last follow-up.

VAS back and VAS leg baseline to last follow-up mean improvement was statistically significant (p value < 0.0001), and clinically important (MCID > 1.16 and MCID > 1.36, respectively)26 for both groups.

ODI scores at baseline and last follow-up were significantly higher in the MI-TLIF group (UMD 17.1; 95% CI 15.09; 19.19; p value < 0.0001 and UMD 2.27; 95% CI 0.07; 4.47; p value = 0.042, respectively. Conversely, the meta-analysis showed no statistically significant difference at either time-point.

ODI baseline to last follow-up mean improvement was statistically significant (p value < 0.0001), and clinically important (MCID > 12.40) for both groups26.

### Complications and fusion

The overall complication rate was 4.7% for iLIF studies versus 9.6% for MI-TLIF. The specific complications identified in each technique are summarized in Table 6. Screw malpositioning (1.6%), adjacent segment degeneration (1.5%), and dural tears (1.3%) were the main complications in MI-TLIF, while in the iLIF group cage migration (1.1%), screw malpositioning (0.6%), and infection (0.6%) were the most prevalent complications (Table 8). The pooled OR was 2.15 (95% CI 1.04, 4.43; p value 0.033) when comparing MI-TLIF and iLIF. The meta-analysis showed a borderline OR (OR 2.03; 95% CI 1.01, 4.12; p value = 0.05) favoring iLIF (Table 9).

The pooled fusion rate had a statistically significant OR favoring iLIF (OR 0.05; 95% CI 0.00, 0.78; p value < 0.0001). However, the meta-analysis revealed no difference between iLIF and MI-TLIF fusion rates.

There was no significant difference (OR 2.39; 95% CI 0.58, 9.87; p value 0.23) in overall revision rate, 1.2% and 2.4% in iLIF and MI-TLIF, respectively (Table 9).

### Subgroup analysis

We performed subgroup analysis according to the number of levels operated. Even though ODI at baseline, and VAS back and VAS leg at last follow-up, had statistically significant differences in one level versus two level MI-TLIF, they were not clinically significant. There was no significant difference regarding complications and fusion rates.

Blood loss was significantly increased when two or more levels were operated in either technique (UMD 41.3 mL; 95% CI 27.63, 55.03; p value < 0.0001 for MI-TLIF; UMD 20.55 mL; 95% CI 1.5, 39.6; p value < 0.036 for iLIF). Both one level and two level iLIF subgroups had significantly less blood loss than one level MI-TLIF subgroup (UMD 122.02 mL; 95% CI 110.38, 133.66; p value < 0.0001 for one level iLIF; and UMD 101.47 mL; 95% CI 83.72, 119.22; p value 0.0012 for two level iLIF).

Length of hospital stay was significantly higher in the MI-TLIF subgroup of two or more levels when compared to one-level MI-TLIF (UMD 3.49 days; 95% CI 2.98; 4.0; p value < 0.0001). There was no significant difference between one-level subgroups of MI-TLIF and iLIF (UMD 0.33 days; 95% CI -0.008, 0.74; p value 0.65). Unlike surgical time and blood loss, there was no available data for length of hospital stay in two level iLIF. Comparison between one level MI-TLIF and overall iLIF length of hospital stay revealed a statistically significant decrease favoring iLIF (UMD 2.17 days; 95% CI 1.79; 2.55; p value < 0.0001).

## Discussion

We conducted a systematic review and meta-analysis of the available literature reporting MI-TLIF and iLIF for the treatment of lumbar degenerative diseases. We derived our conclusions based on the meta-results, given the discrepancy in some outcomes of the pooled results and the meta-analysis.

The main finding is that both MI-TLIF and iLIF provide significant clinical improvement and high rates of fusion at a minimum follow-up of 12 months. No significant difference in complication rates was identified. Furthermore, iLIF was associated with significantly less intraoperative blood loss, and reduced length of hospital stay.

Previous studies documented similar benefits of MI-TLIF over open TLIF (O-TLIF), while achieving the same fusion rates, operative time, and decreased complication rates1,2. On the downside, MI-TLIF was associated with increased radiation exposure. According to our findings, iLIF further enhances most of the benefits of MI-TLIF over O-TLIF. The data reported in the retrieved studies did not allow for a comparative analysis on radiation exposure.

Besides efficacy, concerns about procedure-related complications are among the major setbacks for adopting emerging surgical techniques. The present meta-analysis revealed no significant difference in complication rates. The intraforaminal route allows an anatomic approach to the disc, with total or partial preservation of the articular processes. On the one hand, facet preservation provides dural protection, reducing the risk of dural tear, on the other, direct endoscopic visualization of the nerve root allows assessing the need for further foraminal decompression, reducing the risk of nerve root damage. By providing an ultra-minimally invasive approach, trauma to the soft and bone tissues is reduced, which may account for the residual infection rates reported.

The anesthesia protocol and neuromonitoring might also play a role in the sparse number of neurologic complications reported in iLIF procedures, by allowing intra-operative neurological monitoring. Wu15 and Ao62 used general anesthesia and neuromonitoring, Shen14 and Kolcun56 operated consciously sedated patients, Jin65 used local anesthesia supplemented with neuroleptic analgesia for the decompression procedure, with the aim of sensory-motor separation, and epidural anesthesia when the patient complained about unbearable pain during bone harvest, cage insertion, and percutaneous pedicle screw placement procedures. Yang59 used a low-dose epidural anesthesia combined with local anesthesia or general anesthesia based on physical condition and willingness of patients. The vast majority of the retrieved MI-TLIF studies either did not disclosure the anesthesia protocol or used general anesthesia without neuromonitoring. Gao44 reported 75 patients operated under epidural anesthesia.

Studies using a strictly percutaneous approach (pTLIF) similar to iLIF have reported increased rates of post-operative radiculopathy, dysesthesia and transitory muscle weakness68,69,70. However, once endoscopic assistance was precluded, these studies were not included in our analysis. The absence of endoscopic assessment does not allow nerve root visualization or foraminal revision after cage deployment, which may justify the increased complication rates reported. From the 205 patients included in the iLIF studies, only Ao62 reported a patient with decreased muscular strength of quadriceps femoris, grade 4, after surgery at L4L5 level. The patient had significant relief of symptoms after one month of neurotrophic drug treatment and functional exercise.

Even though endplate preparation might be technically challenging in iLIF due to access restraints, direct endoscopic visualization of the endplates allows confirming adequate subchondral exposure62,65, contributing to the high rates of fusion.

In the iLIF publications, the type of cage used varied across studies. Wu15 and Ao62 used static cages, Kolcun56 used mesh cages, and Shen14, Yang59 and Jin65 used expandable cages. In most MI-TLIF studies, static cages were used or the type of cage was not disclosed, except in the study by Kim45 that used expandable cages. Data on sagittal alignment was insufficient, namely on iLIF studies, to derive any conclusions.

Despite overlapping and additional benefits of iLIF compared to MI-TLIF, it is worth mentioning that endoscopic spine surgeries have a steep learning curve. Knowledge and understanding of foraminal anatomy and its landmarks are of utmost importance. Besides the anesthetic technique and neuromonitoring, continuous spatial orientation of surgical instruments during the procedure is mandatory to avoid nerve root injury, either by direct trauma or excessive retraction. Pre-operative planning with magnetic resonance imaging (MRI) is crucial to assess the nerve root trajectory and eventual anatomic variations, mostly at L5S1 level where the dorsal root ganglion may be particularly endangered10,14. Proper training and previous experience in lumbar transforaminal endoscopic discectomy and decompression are advised.

### Strengths and limitations

Intraforaminal endoscopic assisted fusion is an emerging surgical technique. No systematic review and meta-analysis that synthesizes the available data comparing it to MI-TLIF has been published. Recently, Wagner10 published a review and technical note on uniportal endoscopic assisted fusion. However, despite providing a comprehensive description of the endoscopic assisted intraforaminal transkambin technique, the studies’ procedures were heterogeneous and not standardized. Our review followed the PRISMA guidelines, with a prospective design that specified the main technical details of the procedures and the outcomes and minimum follow-up to be included. Statistical analysis was performed according to previously validated statistical methodology22,23,24,25.

The endoscopic assisted intraforaminal transkambin approach for LIF has been termed as percutaneous endoscopic LIF (abbreviated as PELIF15,65 or PETLIF59,62) or endoscopic MIS-TLIF56. However, the term TLIF is misleading once endoscopic assisted intraforaminal transkambin LIF is a facet sparing procedure, as opposed to the TLIF procedure where a total facetectomy is performed. Also, the term PELIF is ambiguous once it may accommodate different approaches to perform LIF, as long as it is done percutaneously and with endoscopic assistance, either uniportal or biportal. We believe that by referring to the intra-foraminal route instead of the traditional transforaminal route, the term iLIF is more accurate and allows a comprehensive definition of the procedure.

The main limitations of our study are a consequence of the quality of the literature retrieved: low levels of evidence and heterogeneity bias. Only one prospective cohort study comparing both techniques was included, and no randomized control trials. Also, the difference in fusion definition across studies, either by evaluating X-rays or computerized tomography (CT) scans, is a source of heterogeneity and eventual bias on fusion rate assessment. The main clinical outcomes reported were pain scales (VAS back, VAS leg) and a functional/disability scale (ODI). The low prevalence of Health-Related Quality of Life (HRQoL) questionnaires prevented further evaluation and comparison of both techniques.

## Conclusion

iLIF and MI-TLIF for the treatment of lumbar degenerative diseases provide significant clinical improvement and high fusion rates at 12 months or later, without significant difference in complication rates. iLIF has significantly less intraoperative blood loss and reduced length of hospital stay in comparison to MI-TLIF.