Comparison meta-analysis of intraoperative MRI-guided needle biopsy versus conventional stereotactic needle biopsies

Abstract Background MRI-guided needle biopsy (INB) is an emerging alternative to conventional frame-based or frameless stereotactic needle biopsy (SNB). Studies of INB have been limited to select case series, and comparative studies between INB and SNB remain a missing gap in the literature. We performed a meta-analysis to compare INB and SNB literature in terms of diagnostic yield, surgical morbidity and mortality, tumor size, and procedural time. Methods We identified 36 separate cohorts in 26 studies of SNB (including both frameless and frame-based biopsies, 3374 patients) and 27 studies of INB (977 patients). Meta-regression and meta-analysis by proportions were performed. Results Relative to publications that studied SNB, publications studying INB more likely involved brain tumors located in the eloquent cerebrum (79.4% versus 62.6%, P = 0.004) or are smaller in maximal diameter (2.7 cm in INB group versus 3.6 cm in the SNB group, P = .032). Despite these differences, the pooled estimate of diagnostic yield for INB was higher than SNB (95.4% versus 92.3%, P = .026). The pooled estimate of surgical morbidity was higher in the SNB group (12.0%) relative to the INB group (6.1%) (P = .004). Mortality after the procedure was comparable between INB and SNB (1.7% versus 2.3%, P = .288). Procedural time was statistically comparable at 90.3 min (INB) and 103.7 min (SNB), respectively (P = .526). Conclusions Our meta-analysis indicates that, relative to SNB, INB is more often performed for the challenging, smaller-sized brain tumors located in the eloquent cerebrum. INB is associated with lower surgical morbidity and improved diagnostic yield.

• Lesions biopsied with intraoperative MRI guidance are more likely to be smaller in size.
• INB more likely involved brain tumors located in the eloquent cerebrum.
Molecular and histologic characterization of surgically acquired tumor tissue remains the foundation of modern neurooncology, 1 and stereotactic needle biopsy (SNB) is the mainstay surgical technique for achieving such diagnosis when risks outweigh the benefits of an open surgical procedure. 2The diagnostic yield for SNB is generally favorable, ranging 84-100%. 3,4wever, the likelihood of definitive tissue diagnosis decreases with smaller tumors or tumors with significant regional histologic heterogeneity. 5In cases where tissue diagnosis is not achieved in the first procedure, a second SNB achieved definitive diagnosis in 90% of the cases, 6 suggesting technical challenges associated with SNB rather than the intrinsic properties

Comparison meta-analysis of intraoperative MRI-guided needle biopsy versus conventional stereotactic needle biopsies
of the lesion as the underlying cause for diagnostic failures.While the likelihood of procedural complication is low, with reported postoperative morbidity generally <10%, 7,8 the likelihood of complication increases with the number of samples taken, 9 with the number of trajectories to the lesion, 10 and for lesions in the eloquent cerebrum or deep gray matter/brainstem. 10Though most procedural complications are tolerated or self-resolving, devastating sequelae, including mortality, do occur in 1-4% of patients after SNB. 7,11onventional SNB is performed using frame-based or frameless stereotaxy.The diagnostic yield and safety profile of both techniques is comparable for tumors > 1 cm in maximal diameter. 12,13For both procedures, target localization is based on triangulation of the target relative to fiducials of known spatial coordinate, 14 without direct visualization of the target during the procedure.While the mathematics underlying this methodology is elegant, 15 the accuracy of the procedure can be affected by several elements, including technical errors, 16 unintended movements of the fiducials, 16,17 CSF egression during the procedure, [18][19][20] and poor equipment maintenance/calibration. 18 Unfortunately, real-time visual confirmation that the actual biopsy site coincides with the intended target site is not possible with conventional SNB. 21In cases where the frozen pathology of the SNB-acquired samples revealed nondiagnostic tissue, the surgeon is left in the precarious position of making surgical adjustments without understanding the root cause.
The technological convergence of intraoperative magnetic resonance imaging (iMRI) and MRI-compatible stereotactic systems has led to the development of Intraoperative MRI-guided needle biopsy (INB).Key benefits of INB involve the near-real-time visualization of the actual trajectory relative to the planned trajectory as well as opportunities for intraoperative trajectory adjustment. 21or instance, if the location of the Burr hole is suboptimal to support the trajectory, the Burr hole and/or the planned trajectory can be modified in real-time. 22Similarly, if the lesion has shifted as a result of pneumocephalus or CSF egression, the planned trajectory can be modified accordingly. 20,22Moreover, imaging between biopsies affords assessment of biopsy-related complications such as hematoma formation and allows the surgeon to adjust surgical decisions accordingly. 23here has been a growing literature describing the effectiveness and safety profiles of INBs. 9,13Here, we reviewed the available INB literature and performed a comparative meta-analysis of this literature relative to our previously published results for SNB.

Search Algorithm
This systematic review and meta-analysis were conducted in accordance with the PRISMA guideline. 24A comprehensive PubMed database search for articles focusing on intraoperative MRI guidance for needle biopsies primarily for brain tumor indications was performed. 9he following search strategy was used: )) NOT case reports. 9,13The protocol for this meta-analysis titled "Supplemental tools assisting intracranial stereotactic biopsy" was registered with PROSPERO (#CRD42020152735) on 28 April, 2020. 25he studies included in the meta-analysis followed the following inclusion criteria: (1) written in English (or English language translation available); (2) involved human subjects; (3) fully reported peer-reviewed clinical studies; (4) studies reporting diagnostic yield, morbidity, mortality, operative time, and site of lesion biopsied with intraoperative MRI guidance for needle biopsies.
Two authors independently extracted the following data from the included articles (S.D. and C.C.C.).Pertinent data for the size of the lesion was also extracted from the studies.For studies reporting the volume of the lesions, V (cm 3 , or cc) or area of lesions, A (cm 2 ), approximate lesion size was calculated using the formula: r = 2x ∛ V. (3/4π) or r = √A/ π, respectively.Data for frame-based and frameless needle biopsy groups was used from our previous publications. 9,13Overall, the variables used in data compilation and analysis were: first author, year of publication, Q score (Newcastle-Ottawa quality assessment score), 26 methods of needle biopsy (FB/FL, intraoperative MRI), diagnostic yield, morbidity, mortality, operative time, and size of the biopsied lesion.
Quality (Q) score was assigned to each study using the Newcastle-Ottawa scale in a similar manner as our previously published analyses. 9,13,26The studies included in this analysis were assessed in a similar manner.A score of 1(versus 0) was assigned for satisfactory fulfillment of each criterion.Studies with a NOS ≥ 5 were classified as highquality studies and those with NOS < 5 were categorized as low-quality studies. 9,13

Statistical Analysis
DerSimonian and Laird random effects model was applied to pool the meta-analysis results. 27Meta-analysis by proportions and meta-regression analyses were performed.The event rates for diagnostic yield, morbidity, and mortality were calculated using meta-analysis by proportions.The effect size was reported in terms of odds ratio (OR) for diagnostic yield, morbidity, and mortality, and in standard difference in means (SDM) for surgical time and maximum tumor diameter, with a 95% confidence interval.Metaregression analysis was used to compare event rates for DY, morbidity and mortality; SDM for surgical time, and mean size of the lesion between the INB and SNB groups.Subgroup analysis was used to compare the proportion of patients across INB and SNB groups who underwent biopsy in the eloquent area of the cerebrum.Eloquent areas included left frontal, left temporal, bilateral parietal and occipital lobes, corpus callosum, pineal region, and deep gray matter-basal ganglia, thalamus and brainstem.
Similar to our previous analysis, heterogeneity across the studies was gauged using the Higgins inconsistency Dhawan and Chen: Intraoperative MRI-guided needle biopsy for brain tumors index (I 2 ) and Cochran's Q χ² test. 28High, moderate, and absence of heterogeneity was reported if I 2 was >50%, 25-50% or <25%, respectively. 28unnel plots, Egger's regression intercept test, and Duval and Tweedie's trim and fill test were used to assess the publication bias. 29,30The overall stability of our analysis was determined using a cumulative meta-analysis, performed after arranging the studies from largest to smallest w.r.t sample size (and from most to least precise) in a similar fashion as performed in our previous publications. 9,13The influence of individual studies was ruled out by performing sensitivity analysis by excluding one study at a time. 31omprehensive meta-analysis (CMA) software, version 3.3070 Biostat, Englewood, New Jersey, USA was used for our analysis.P-value < 0.05 was considered statistically significant.

Study Selection
Out of the total 5801 studies that were identified from the PubMed database in our previous study, 9,13 540 pertinent studies focused on supplemental tools for needle biopsy.Intraoperative MRI was selected for manual screening by title and abstract.Sixteen studies added from references mentioned in other studies were additionally screened.Ninety-four studies were reviewed as full text after excluding 430 studies not directly related to intraoperative MRI or frame-based/frameless needle biopsies.Fifty-three out of 94 studies focused on intraoperative MRI guidance for needle biopsies.After further screening, 27 studies were included in the meta-analysis (Figure 1).

Study Characteristics
Twenty-seven studies included in our meta-analysis were published from 1999 to 2023 and were either prospective observational studies or retrospective analyses. 4,21,22, Demoaphics of individual studies are shown in Table 1.
Using the Newcastle-Ottawa scale, 23 studies were categorized as high-quality, and 4 were graded as low-quality studies.In total, the 26 studies yielded results for 3374 patients in the FB/FL group, and 27 studies yielded results for 977 patients in the INB group.The clinical characteristics of each study in the INB group are shown in Table 2.The distributions of these clinical variables are comparable to those previously published for SNB. 13 The mean age of patients in the study cohorts ranged from 19 to 57.5 years.The predominant pathology was glioblastoma.The main difference between the INB and SNB studies were: (1) the mean maximum diameter of INB biopsied lesions ranged from 0.8 to 5.3 cm and SNB biopsied lesions 3.1 to 5.1 cm in FB/ FL group (Table 2), (2) 79.4% INB biopsied patients harbored lesions in eloquent cortex whereas 62.6% in the SNB group harbored lesions in the eloquent cortex (Table 2).

Diagnostic Yield, Morbidity, Mortality, Mean Lesion Size, and Procedural Time
Pooled estimates for diagnostic yield for studies in the INB group were 95.4% (93.3-96.8%), in comparison to 92.3% (89.1-94.7%) in the FB/FL group, P = .026(Figure 2).Data for morbidity (including the rate of postoperative hemorrhage and neurological deficit) were available in 26 studies in the INB group; and 20 studies (33 cohorts) in FB/FL group.We found the rate of morbidity to be 6.1 % (4.2-9.0%) in the INB group, which was significantly lower (P = .004)than the morbidity observed in the FB/FL group [12.0%(9.1-15.8%)](Figure 3).No significant difference in mortality was found between the 2 groups: INB (1.7%) and FB/FL (2.3%), P = .288(Figure 3).The mean maximal diameter of lesions in patients in the INB group (2.7 cm) was significantly smaller than the lesion size in patients in the SNB group (3.6 cm, P = .032)(Figure 4).The pooled estimate of procedural time for INB and SNB was comparable at 90.3 min and 103.7 min, respectively (P = .526)(Figure 4).
We performed a continuous meta-regression analysis with the lesion size as the moderator variable.We do not

Neuro-Oncology Advances 5
Dhawan and Chen: Intraoperative MRI-guided needle biopsy for brain tumors see a significant effect of lesion size on morbidity (P = .724)(Supplementary Figure S1).

Heterogeneity and Publication Bias Analysis
The Cochran's Qχ 2 and I 2 tests were used to gauge the heterogeneity of results in the INB literature (Supplementary Table 1).Sensitivity and cumulative analysis were performed if I 2 was >50% to determine any outlier study.The funnel plots showed no evidence of gross asymmetry for diagnostic yield, morbidity, mortality, duration of procedure, and size of lesion for the INB studies (Supplementary Figure S2), suggesting no detectable publication bias.Egger's regression intercept test was also performed for quantitative evidence of bias (Supplementary Table 1).

Discussion
To our knowledge, this study is the first to compare INB and SNB through a pooled meta-analysis.Our analysis indicates that, relative to SNB, INB studies more likely involved brain tumors generally deemed more difficult to biopsy or more likely to be associated with increased risk for procedural morbidity, including smaller brain tumors and tumors located in the eloquent cerebrum.Despite these differences, the pooled estimates of diagnostic yield for INB were higher than those associated with SNB (95.4% versus 92.3%, P = .026),and the pooled estimate of surgical morbidity for INB was half that associated with SNB (6.1% versus 12.0%, P = .004).The pooled estimates for procedural mortality and time were comparable between SNB and INB groups.These results support consideration for INB as a surgical technique for achieving tissue diagnosis in select brain tumors.
There are genuine advantages to real-time visualization of the biopsy trajectory relative to the target during a needle biopsy.For instance, surgical maneuvers can be adopted to adjust to technical deviations from the planned trajectory or to procedural complications, such as entry/biopsy site hematoma. 23These benefits highlight challenges associated with SNB, which is fundamentally a "blind" procedure that relies on mathematical triangulation of the target lesion relative to fiducials.Challenges associated with such "blind" procedures underlie the impetus for the development of INB.In this context, our meta-analysis supports the future development and adoption of INB for select lesions.Moreover, our analysis indicates that the benefits associated with INB were not associated with an overall increase in procedural time relative to SNB.Nevertheless, the benefits of INB should be weighed against the cost associated with intraoperative MRI use, the cost of the MRIcompatible stereotactic systems, 56 as well as the learning curve associated with INB. 57With a pooled diagnostic yield estimate of 92.3% and procedural morbidity of 12.0%, SNB is an excellent choice for most brain tumors.For brain tumors that the surgeon considered difficult to biopsy because of size or location, or for higher-risk patients, our meta-analysis suggests consideration for INB is warranted.As a meta-analysis, the results provided here should be considered in the context of the limitations intrinsically associated with a meta-analysis design. 58To address these limitations, we followed the best practiced in the published guidelines for literature review and meta-analysis. 24The standard metric of assessment for our meta-analysis revealed no significant heterogeneity or publication bias in the INB literature.That said, as a newer technology, the INB literature is small relative to that available for SNB.
The important causes for reported peri-procedural morbidity and mortality were an expanding intraparenchymal hematoma or worsening edema postbiopsy for which the patients were taken back to the operating room.In this context, one must carefully plan the biopsy trajectory to avoid entrance through the sulci, or any blood vessel.It is also important to monitor neurological exams in the postoperative period for any clinical signs of worsening edema.One reported mortality occurred before the biopsy was performed, hence, securing a preoperative clearance from our anesthesia colleagues cannot be underestimated.
Although the cost-effectiveness of intraoperative MRI for the treatment of high-grade gliomas has been well established in the literature, 56 but the same still needs to be established for the purposes of biopsy.One can argue that on one hand with intraoperative MRI the higher diagnostic yield, lower postoperative complications and the ability to correct the trajectory in real time for brain shifts make this modality cost-effective, while on the other hand, the cost of MRI with relatively longer anesthesia duration and cumulative operating room time adds up to the total cost of the procedure.Future studies should focus toward analyzing the cost of intraoperative MRI-guided biopsies versus conventional stereotactic needle biopsies.

Conclusion
Our meta-analysis indicates that the INB literature more likely involves smaller brain tumors or tumors located in the eloquent cerebrum.Despite these differences, INB is associated with higher diagnostic yield and lower procedural morbidity.

Figure 1 .
Figure 1.Preferred reporting items for systematic reviews and meta-analysis (PRISMA) flow diagram for article search and study selection.

Figure 2 .
Figure 2. (A) Forest plot showing pooled estimates comparing the diagnostic yield of INB and SNB groups.(B) Meta-regression bubble plot for diagnostic yield with the modality of biopsy as the moderator variable.

Figure 3 .
Figure 3. Forest plot showing pooled estimates comparing morbidity (A) and mortality (C) of INB and SNB groups.Meta-regression bubble plots for morbidity (B), and mortality (D) with modality of biopsy as the moderator variable.

Figure 4 .
Figure 4. Forest plot showing pooled estimated comparing mean lesion size (A), lesion in the eloquent area (B) and procedural time (C) in INB and SNB groups.Meta-regression bubble plots for mean lesion size (D), lesion in the eloquent area (E) and procedural time (F) with the modality of biopsy as the moderator variable.

Table 1 .
Demographics of studies included in the INB group.Q: Newcastle-Ottawa Score

Table 2 .
Clinical characteristics of studies included in the INB group.* The median size of the lesion.¶ Distribution of tumor location not clearly mentioned