INTRODUCTION
Minimally invasive spine surgery (MIS), boasting reduced surgical blood loss and shorter hospital stays, has become a leading technique for degenerative spine disease and is increasingly being utilized for nondegenerative conditions, including intradural pathology [1,2]. With increasing familiarity and greater utilization of MIS techniques for nondegenerative spine disease, formal comparison to open techniques is necessary, although difficult in the absence of high-quality evidence. The utility of MIS techniques for intradural extramedullary (IDEM) spine pathology is largely based on observational data, consisting primarily of single surgeon or single institution case series. In the absence of randomized controlled trials, pooled observational data has proven useful to characterize reported multi-institutional experiences with MIS and provide a preliminary basis for comparison to open techniques. While MIS continues to be the subject of numerous reviews, significant heterogeneity often exists between reported surgical techniques broadly classified as minimally invasive. We therefore systematically reviewed published MIS data, specifically looking at tubular retraction in the context of IDEM spinal pathology, and present our case series data, which we contributed to a pooled analysis.
MATERIALS AND METHODS
1. Case Series
We retrospectively reviewed consecutive patients, treated by a single surgeon, who underwent MIS using a tubular approach for IDEM pathology between 2006 and 2018. Demographic information was collected, including patient age, sex, presenting symptom, and pathology. Perioperative details, including operative time, length of hospitalization, and estimated blood loss (EBL) were collected in addition to outcome measures, characterizing degree of resection, complications, neurologic status, and follow-up. All patients underwent standard pre- and postoperative imaging, including magnetic resonance imaging (MRI) and/or computed tomography. Postoperative imaging provided tumor surveillance (when applicable) and confirmed the absence of wound complications (e.g., pseudomeningocele, seroma, etc.).
2. Surgical Technique
Patients underwent general anesthesia and were positioned prone on a radiolucent table. Fluoroscopy was used for localization. A unilateral approach was used for exposure. A paramedian skin incision was made approximately 1–1.5 cm off midline and carried through the dorsal fascia. Soft tissues and musculature were sequentially dilated to accommodate the largest diameter of the lesion, per preoperative imaging. A fixed or expandable table-mounted tubular retractor was placed over the dilators, which were removed (MetRx, Quadrant retractor, Medtronic, Inc., USA). An operating microscope was used in all cases. Either a unilateral laminotomy, extended hemilaminectomy, or sublaminar bilateral decompression was performed. For the extended unilateral approach, midline structures were removed, including the anterior spinous process and ligamentum flavum. For sublaminar bilateral decompression, ligamentum flavum was removed and medial facets/lateral recesses decompressed bilaterally. Dura was opened sharply using a bayonetted knife. 4-0 Nurolon suture (Ethicon, Inc., USA) was used for dural retraction, secured through the retractor to the drapes under tension. A smaller diameter tube of similar length was then inserted for additional radially oriented retraction of the dural sutures. The “double tube” technique was recently described in a similar fashion [3]. Standard micro-neurosurgical techniques were used for intradural exploration and tumor resection, including use of ultrasonic aspiration, when necessary. All specimens were submitted for histopathologic review. Primary dural closure was initially performed using the discontinued U-clip device (Medtronic, Inc.), with subsequent dural closures performed using 4-0 Nurolon suture. Suturing through a tubular retractor was aided by bayoneted needle drivers and dedicated knot pushers (Scanlon International Inc., USA). Dural closure was bolstered by dural sealant to reduce the risk of spinal fluid leak or pseudomeningocele. The tubular retractor was then removed, confirming hemostasis before a multilayered closure. Drains were not used. Patients were allowed to mobilize on the day of surgery.
3. Systematic Review and Pooled Analysis
A systematic review was performed with reference to PRISMA (Preferred Reporting Items for Systematic Reviews and Meta-analyses) guidelines [4]. PubMed was used to query National Library of Medicine database resources, including MEDLINE, PubMed Central, and the NCBI (National Center for Biotechnology Information) Bookshelf. A search was conducted for relevant English articles published between 1990 and 2021 using the following keywords: minimally invasive, MIS, spine, tumor, intradural extramedullary, IDEM (last accessed September 19, 2021). Publications detailing cohorts with intradural extramedullary spine tumors treated using MIS techniques were included. We included only publications specifying the use of tubular retraction. Endoscopic and hybrid techniques, including “mini open” techniques, were excluded. Case reports detailing experience with a single patient were also excluded. References within selected articles were similarly reviewed for eligibility. For inclusion, publications reported, at a minimum, tumor pathology and extent of resection. When available, we additionally collected patient demographics and perioperative details, including operative time, EBL, length of hospitalization, and complications. Technique specifics, including tubular retractor type/size and methods of dural closure, were collected when available. From reports also detailing experience with extradural and intramedullary pathology, details from only IDEM tumor patients were extracted. Similarly, only MIS data was included from reports also detailing experience with open techniques.
Articles were screened and preliminarily reviewed by 1 author (AB), although ultimately reviewed by all authors. Our university statistics colleagues (see acknowledgments) assisted with data extraction and descriptive statistics. Cohort data from our case series was pooled with data extracted from the systematic review for pooled analysis.
4. Statistical Analysis
Descriptive statistics were performed using Microsoft Excel (Microsoft, USA). For continuous variables, series means were collected (when reported) or calculated from available subset data specific to the cohort of interest. Series means were weighted as a proportion of the pooled cohort for analysis and reported alongside extracted component range values. In the absence of consistent patient-level data, series means were used to calculate pooled median values. Pooled continuous data is displayed graphically in box plots. Categorical data is reported for individual case series and as relative frequencies within the pooled cohort. Pooled categorical data is displayed graphically in bar charts. Heterogeneity across studies was not assessed.
RESULTS
1. Case Series
Between 2006 and 2018, 24 patients (14 males, 10 females) underwent MIS for intradural spinal pathology at Northwest Medical Center (USA). Patient demographics, clinical outcomes, and perioperative measures are summarized in Table 1. Average age was 64 years, ranging from 29 to 85 years. All patients presented with pain, sensory abnormalities, motor deficits, bowel/bladder complaints, or a combination thereof. Mean time from symptom onset to diagnosis was 14 months. Histopathology revealed meningioma in 8 patients, schwannoma in 7 patients, metastatic disease in 3 patients, and arteriovenous fistula, lipoma, teratoma, ganglioneuroma, arachnoid cyst, and tethered cord in 1 patient each. There were 12 patients (50%) with thoracic pathology, 10 patients (42%) with lumbar pathology, and 2 patients (8%) with cervical pathology. Mean operative time was 149 minutes, ranging from 90 to 240 minutes. Gross total resection (GTR) was achieved in 100% of cases and confirmed by postoperative MRI. No patients experienced complications related to the surgical technique, including cerebrospinal fluid (CSF) leak, pseudomeningocele, or need for reoperation. Average length of hospital stay was 2.2 days. Mean EBL was 23 mL. Twenty patients (83%) demonstrated either improvement or stabilization of their symptoms over a mean follow-up of 2.9 years.
2. Systematic Review and Pooled Analysis
Of the 1,369 records identified through databases, 10 were identified for analysis [3,5-13]. Citation cross-referencing yielded an additional 6 articles, bringing the total included for analysis to 16 [14-19]. The article selection process is detailed in Figure 1 as a flow chart.
All included studies presented retrospective cohort data from consecutive IDEM spine tumor patients treated surgically using MIS techniques and tubular retraction. Two series did not provide retractor data, although authors did specify that nonexpandable tubular retractors were used [3,8]. Three of the included reports compared their MIS cohorts, from which patient data was extracted, to similar cohorts treated with open techniques [7,8,13]. Only limited metrics could be extracted from the series of Lee et al. [8]. Despite providing detailed measures from their mixed MIS cohort, authors did not distinguish metrics between open microsurgical cases and those in whom tubular retraction was used. IDEM tumor data was extracted from 4 reports also detailing experience using tubular retraction for extradural and intramedullary pathology [6,11,14,19].
Patient demographics and perioperative details from each of the included studies are summarized in Tables 2 and 3, respectively. Details of reported pre- and postoperative neurologic measures are reported in Table 4. Several case series commonly cited in the MIS literature failed to meet inclusion criteria. Two reports were excluded because of absent data specifying which, of several treated tumors, were IDEM [20,21]. Articles were also excluded for variant techniques. For example, reports from Pompili et al. [22] and Zong et al. [23] describe open microscopic excision through a standard hemilaminectomy. Mende et al. [24] describe a similar mini open technique using a Caspar retractor. Others detail a midline transspinous approach, albeit making use of MIS retraction [25-27]. Combining our case series data with data extracted from the 16 included case series (17 series total) yielded 327 lesions in 323 patients for pooled analysis. Nonmass pathology from our case series (n=2: tethered cord, arteriovenous fistula) was excluded from pooled analysis, as was nonmass IDEM pathology in the series published by Gandhi and German (n=3, tethered cord) [19]. Such pathology was not reported in the other included studies.
3. Pathology
All studies (17 of 17) reported pathologic diagnoses, summarized in Figure 2A. Nerve sheath tumors and meningiomas accounted for over 80% of tumors, similar to distributions previously reported [28-30]. While Wong et al. [13] reported several nerve sheath tumors as a single undifferentiated group (n=16), schwannomas accounted for the vast majority (n=147 of 168) of all pooled nerve sheath tumors. Nearly all studies (16 of 17) reported tumor location (n=321 tumors). Scored according to their most rostral extension, tumors occurred most frequently in the thoracic (40%) and lumbar spine (44%). Isolated sacral tumors were least common (n=5). Location data specific to IDEM pathology, summarized in Figure 2B, was not able to be extracted from 1 series (Lee et al. [8]; n=6), as previously described. Nearly all studies (16 of 17) commented on tumor size (n=321 tumors). Authors reported the number of involved spinal levels in the majority of series (13 of 17, n=286 tumors), with 1 level (n=130, 45%) and 2 level (n=151, 53%) involvement being most common [3,5,6,9-14,17-19]. Three series reported experience with IDEM pathology involving 3 or more levels (n=5, 2%) [11,14,19]. Of those, Balasubramanian et al. [14] reported subtotal resection (STR) of a thoracolumbar (T11–L4) myxopapillary ependymoma while using fixed tubular retraction (18 mm), Gandhi and German [19] reported GTR of a lumbosacral (L3–S1) plasmacytoma using expandable tubular retraction (22–52 mm), and Nzokou et al. [11] reported GTR in each of the following while using fixed tubular retraction (18–24 mm): lumbosacral arachnoid cyst (L5–S2), cervical neuroendocrine tumor (C5–7; GTR locally), and a cervical schwannoma (C5–7) [11,14,19]. Specific tumor dimensions were provided in 7 of 17 series [3,7,8,15,17-19]. Cranio-caudal (CC) diameter was the most frequently reported measure (5 series, n=149), ranging from 5 to 52 mm, with a weighted mean of 19.1 mm [3,7,15,17,18]. Axial diameter was reported in 3 series (n=43), ranging from 3 to 25 mm, with a weighted mean of 10.1 mm [3,8,18]. Sagittal diameter was reported in 4 series (n=63), ranging from 7 to 31 mm, with a weighted mean of 14.9 mm [3,7,8,18]. Eicker et al. [16] (n=6), using MIS tubular retraction specifically to approach ventrally located tumors of the craniocervical junction, reported only that tumors greater than 25 mm were excluded. Gandhi and German [19] reported a single diameter for each patient, from which an overall average was calculated, although the authors did not specify dimension, so this was not included for pooled analysis.
4. Resection
Degree of resection was reported in all studies (n=327 tumors), with most series (11 of 17) reporting 100% GTR. Based on individual series GTR rates/number of tumors treated, GTR was achieved in 303 of 327 cases (93%, pooled series median 100%) (Figure 3A). Six series reported cases of STR [6,10,13,14,17,18]. Thavara et al. [18] reported the lowest resection rate, achieving GTR in only 8 of 12 patients (67%), intentionally leaving tissue adherent to spinal cord and nerve roots in 4 cases. In fact, STR was attributed to characteristics of the targeted pathology (e.g., atypical, infiltrative, adherent to neural structures) in 12 of 24 (50%) cases of STR overall. Formo et al. [17] reported 2 cases of STR during the MIS portion of a staged resection and Wong et al. [13] reported STR in the setting of widely metastatic cancer when palliative STR was deemed an appropriate alternative to GTR. Gandhi and German [19] similarly reported STR in 1 patient with widespread dural disease, although specified that GTR was achieved locally at the surgical level. Because GTR was achieved in all tumor patients for whom resection was performed, the Gandhi and German series [19] was scored as achieving 100% GTR in pooled analysis. Residual tumor was identified only on postoperative imaging in 3 of 24 cases of STR.
5. Blood Loss
Mean EBL was reported in 13 of 17 series (n=219 patients), ranging from 23 mL in our series to 297 mL [9], with an overall weighted mean of 131 mL (pooled series median 150 mL) (Figure 3B) [3,5-7,9-13,15,18,19]. Dahlberg et al. [15] reported only ranges, which were excluded from analysis. Balasubramanian et al. [14] did not report individual patient blood loss but reported an overall mean EBL of 30 mL for their 41 patients’ series, from which we extracted data from 33 patients. This value was included for pooled analysis.
6. Operative Time
Mean operative time was reported in 13 of 17 series (n=195 patients), ranging from 85 minutes [9] to 260 minutes [18], with an overall weighted mean of 169 minutes (series median 160 minutes) (Figure 3C) [3,5-7,9-13,15,18,19]. Balasubramanian et al. [14] and Formo et al. [17] reported ranges, which were not included for pooled analysis. Maduri et al. [9] distinguished between operative time, which was used for pooled analysis, and total time in the operating room, which was not used. Others used these terms interchangeably [11,13], although provided only a single value, which was used for analysis. Maduri et al. [9] reported that the use of image guidance, on average, added 102 minutes to the total time in the operating room. Mannion et al. [10] reported decreasing operative times with sequential cases, highlighting the learning curve associated with MIS tubular retraction (case 1: 4.2 hours; case 8: 1.7 hours).
7. Hospital Stay
Length of hospital stay was reported in 12 of 17 series (n=183 patients), ranging from 2.2 days in our series to 8.7 days [9], with an overall weighted mean of 4.1 days (series median, 3.7 days) (Figure 3D) [5,6,9-15,18,19]. Three patient series (n=43) reported markedly longer than average hospital stays, ranging from 6-8.7 days [5,9,18]. Maduri et al. [9] attributed this, at least in their series, to 1 patient remaining inpatient for 26 days, awaiting placement. Formo et al. [17] (n=83), the largest cohort in our pooled analysis, reported only range and median data for hospital stay (not mean) and was therefore not included in the hospital stay pooled analysis. Interestingly, they reported a similar median hospital stay of 3 days, supporting the likelihood that significantly longer stays reflect outliers.
8. Complications
Nearly all studies (16 of 17) detailed the presence or absence of complications (n=297 patients). The single study omitting such information was written primarily to quantify the expanded visualization attained with the “double tube” technique (43% increase), referred to previously [3]. While the study met minimum inclusion criteria, only limited demographics and perioperative measures were provided. Complications most commonly reported include CSF leak/pseudomeningocele (n=9, 3%), neurologic decline (n=7, 2.4%), and wound complications, including infection, dehiscence, and subcutaneous hematoma (n=6, 2%). Of the 9 patients with CSF leak/pseudomeningocele, 3 required operative wound revision, 3 were treated by oversewing the incision and placing a lumbar drain, and 3 were deemed asymptomatic, for which no intervention was performed. Anticipated sensory abnormalities associated with nerve sheath tumor resection were not considered a complication. Mannion et al. [10] reported wrong level exposure in 2 patients, with 1 requiring conversion from MIS to open, citing poor exposure. A similar case, wherein conversion to open was deemed necessary, was reported by Thavara et al., [18] although excluded from patients reported in the series. While not a surgical complication, per se, Gandhi and German [19] reported a single case of iatrogenic instability associated with facetectomy at the time of surgery, for which interbody fusion was also performed, and adjacent level disease in a second patient with existing spondylosis, who ultimately required reoperation for fusion (mean follow-up of 18 months). Mannion et al. [10] also describe a single case of iatrogenic instability associated with facetectomy, for whom fusion was also performed during the index surgery. While many series acknowledged the potential for instability, fusion was deemed necessary in only 2 patients at the time of surgery, and only 1 patient required reoperation for fusion following their index surgery. In studies reporting surgical morbidity, complications were reported in nearly 8% of patients overall, with mean follow-up ranging from 6-36 months. No studies reported an occurrence of operation-related mortality.
9. Neurologic Outcome
While all studies reported variable improvement in neurologic function with surgery, only 9 of 17 series used standardized measures for myelopathy, pain, and neurologic function. Of those, only 6 of 9 reported both pre- and postoperative measures for pooled analysis, summarized in Table 4 [5,7,9,11,14,18]. Afathi et al. [5] (n=18) classified 22% of patients as Frankel C preoperatively and all patients as Frankel D (89%) or Frankel E (11%) postoperatively, a change reflecting the return of useful motor function with surgery. This was similarly reported by Nzokou et al. [11] (n=4), who classified 2 of 4 patients as ASIA C preoperatively and all patients as ASIA D or ASIA E postoperatively, reflecting improvement to at least antigravity strength with surgery. Balasubramanian et al. [14] (n=33) and Konovalov et al. [7] (n=20) classified patient neurologic function using the modified McCormick classification. Preoperatively, 66% of patients were classified as McCormick 1 or 2, reflecting functional independence without the use of any assistive device, compared to 89% following surgery, with 58% of patients classified as McCormick 1, intact neurologic function with minimal dysesthesia, compared to only 26% before surgery. Only 11% of patients were scored as McCormick 3 postoperatively, modest impairment requiring an assistive device, compared with 34% of patients classified as McCormick 3 or 4 preoperatively, reflecting greater dependency before surgery. This trend was similarly reported by Maduri et al. [9], who classified only 38% of patients as Nurick 0–2 preoperatively, having ambulation and neurologic function conducive to full-time employment, compared to 77% following surgery. Nzokou et al. [11] (n=4) and Thavara et al. [18] (n=12) reported pain using the visual analogue scale, which improved from a weighted average of 7.9 preoperatively to 1.3 following surgery.
10. Dural Closure
Nearly all studies (16 of 17) detailed methods of dural closure, which can be particularly challenging in the setting of tubular retraction (Table 5). Several (5 of 17 series) mentioned the use of dedicated suture knot pushers to facilitate suturing through tubular retractors. Authors also described the use of vascular clips (3 of 17 series) as an alternative to primary dural closure using suture. Closure was frequently bolstered by dural sealant (10 of 17 series), a hemostatic sponge (3 of 17 series), or both (1 of 17 series). Of the 6 series reporting cases of CSF leak/pseudomeningocele, 5 of 6 detailed the use of dural sealant or hemostatic sponge, while only 1 of 6 mentioned the use of a suture knot pusher [6,7,10,13,17,18].
DISCUSSION
Minimally invasive techniques and their application to spinal pathology continue to expand. Our findings summarize robust experience and excellent surgical outcomes achieved with MIS tubular retraction and support its use as safe and effective for the treatment of diverse intradural pathology. GTR was achieved in 100% of patients in our case series and 93% of pooled patients, which is similar to the 89%–100% GTR reported in comparable open cohorts within the included case series [7,8,13]. Mean blood loss was 23 mL in our case series and 131 mL in pooled patients, considerably less than that reported within open cohorts from the included series, which ranged from 297 mL to 559 mL [7,8,13]. Similarly, mean hospital stay for MIS patients, 2.2 days in our case series and 4.1 days in pooled patients, is substantially shorter than open cohorts from the included series, which ranged from 6.1 to 6.8 days [8,13]. Operative time, which is likely surgeon and pathology-dependent, averaged 149 minutes in our case series, 169 minutes in our pooled MIS cohort (decreasing with sequential cases), and 134 minutes to 241 minutes in open cohorts [7,8,13]. In this series, only 1 patient in the pooled cohort required reoperation for postoperative instability. This is consistent with a Cochrane review of posterior decompression techniques, with decreased incidence of postoperative instability following laminotomy compared to laminectomy [31,32]. In the setting of spine tumors, postlaminectomy kyphosis is associated with increasing number of levels involved, decreased age, and preoperative deformity [33,34]. The low rate of postoperative instability within our pooled cohort is unsurprising, considering most lesions extended 1–2 levels and the relative paucity of patients under the age of 18 (n=2). The impact of retractor type (fixed/expandable) and tumor size on outcome is less discrete, especially considering the retrospective nature of our pooled cohort data. Of case series specifying tumor dimensions, the 2 series using fixed tubular retraction reported smaller averages for axial and CC diameters, a trend which was not present for sagittal diameters [3,8]. Greater exposure afforded by an expandable retractor might allow for larger tumors to be resected in axial and CC planes, but not necessarily in the sagittal plane, which would be primarily impacted by retractor length/depth instead of opening diameter. Although large tumors could theoretically be targeted using a multilevel MIS approach, tumors amenable to MIS resection have historically been limited to 1–2 levels. And while tumor size may influence the decision to pursue an MIS approach and guide retractor type, degree of resection in patients deemed appropriate for MIS seems to be largely influenced by features of the pathology. In our pooled cohort, infiltrative/adherent or atypical pathology was cited as the reason for STR in 50% of cases, whereas size seemed to prevent GTR in only a single case, the T11–L4 myxopapillary ependymoma noted previously. A recent meta-analysis assessed the impact of tumor resection on neurologic outcome for spinal tumors resected using MIS techniques, concluding that while both GTR and STR were associated with neurologic improvement, a larger proportion of patients experienced improvement with GTR [29]. Thus, in patients selected for surgery, maximal safe resection confers the greatest benefit, considering features of tumor pathology while reconciling extent of resection and preservation of neurologic function.
Several limitations should be taken into consideration when interpreting our results. Relying entirely on low evidence level case series data, pooled patients had already been selected to undergo minimally invasive surgery (selection bias). In addition, the inclusion criteria and surgical technique were not controlled limiting the consistency of the pooled results. Several factors, including tumor size, which is frequently considered when determining suitability for MIS, may be favorably skewed. Pooled perioperative measures are also likely influenced by reporting bias, particularly with estimated and subjective values. The impact of unpublished observational data, particularly negative data, is difficult to estimate and adds to the likelihood of reporting bias. While we attempted to reduce bias by standardizing the surgical technique of interest, patient-level data was inconsistently available between reports, especially for larger series, from which only group-level descriptors could be used, therefore limiting overall statistical analysis. Moreover, our pooled cohort data is largely derived from single-arm observational series. While 3 series [7,8,13] compared MIS and open cohorts, only Wong et al. [13] provided sufficient data to measure effects for the purpose of metanalysis. Our pooled MIS cohort does provide a means for comparison to a separate pooled open cohort, although this was not performed. Future series should consider prospective enrollment, reporting of patient-level metrics, and comparison of standardized MIS cohorts to open cohorts.
PDF Links
PubReader
ePub Link
XML Download
Full text via DOI
Download Citation 
Print 
