Introduction

Brain metastases (BM) are the most common intracranial neoplasms, with an estimated incidence of 9% in adults with systemic malignancies [1, 2]. Surgical resection and radiotherapy remain the mainstay of treatment, but expose patients to potential adverse events [3, 4]. Radiation necrosis (RN) is a known complication, occurring approximately 3–12 months after completion of radiotherapy. Incidence ranges between 5 and 25% based on modality of treatment, total dose, and fractionation [5,6,7]. Symptoms are non-specific and stem from necrotic foci mass effect, which may mimic tumor recurrence. The diagnosis is supported by characteristic “Swiss-cheese” or “soap-bubble” enhancement, and may be confirmed with biopsy [7, 8]. Steroids provide temporary symptomatic relief, but long-term use correlates with serious complications; similarly, surgical debulking is not risk-free [7, 9]. Bevacizumab has proven effective in treating RN by counteracting the upregulation of VEGF [10, 11]. Likewise, laser interstitial thermal therapy (LITT) can resolve necrotic foci by generating thermal thrombosis of abnormal surrounding vessels [12, 13].

Bevacizumab and LITT are both viable treatments for RN not amenable to surgical excision, but direct comparative data is scarce [14]. In this review, we assess the differences in clinical outcomes, radiological responses, and survival rates between bevacizumab and LITT in patients with RN from BM.

Methods

Literature search

A systematic review was conducted following the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines [15]. PubMed, EMBASE, Cochrane and Scopus databases were screened for eligible articles from inception to April 3, 2021 operating the Boolean full-text search [(radiation necrosis OR radionecrosis) AND (anti-VEGF OR bevacizumab OR laser interstitial thermal therapy OR LITT)]. Eligible studies were collected and exported to Mendeley; duplicates were removed.

Study selection

Inclusion and exclusion criteria were set a priori. Articles were included if they met the following criteria, in line with the PICOS format: (1) retrospective or prospective studies (Study design) including a minimum of 5 patients confirming the radiological or histological diagnosis of RN following radiotherapy for BM (Population); (2) treatment with bevacizumab or LITT (Intervention, Comparison); (3) available data on radiological response and clinical improvement (Outcome). We excluded: (1) systematic reviews, meta-analyses, case series with less than 5 patients, animal, cadaver, and laboratory studies; (2) studies lacking adequate reports on clinical/radiological outcomes; (3) studies with unclear radiological or histological distinction between patients with RN from BM, BM recurrences or other pathology.

Two authors (C.D.N. and P.P.) independently screened titles/abstracts of all identified articles and reviewed full-texts of studies that met the inclusion criteria. Disagreements were settled by a third author (A.S.H.). References of included articles were also searched to retrieve additional papers.

Data extraction

Data were extracted by one reviewer (P.P.) and independently verified by two additional reviewers (A.S.H. and C.D.N.). Patient-level data were extracted directly or calculated from raw data. Data included: authors, year, study-design, sample-size, age, gender, primary tumor, radiation type, symptoms, imaging findings, steroids, bevacizumab dosage/cycles, hospital-stay, adverse-events, clinical outcome, radiological response, recurrence, progression-free survival (PFS), and overall survival (OS) [16, 17]. Clinical symptoms, steroid wean-off, and radiological responses were evaluated at 6-months after treatment or at the last available follow-up (at least > 1 month). Radiological response was assessed using the modified RANO criteria for BM: complete response (CR) = resolution, partial response (PR) = reduced volumes, stable disease (SD) = same volumes, progression (PD) = increased volumes [16, 17].

Data synthesis and quality assessment

The primary outcomes of interest were clinical and radiographic outcomes in patients with RN treated with bevacizumab or LITT. These included post-treatment symptomatic improvement, weaning off steroids, RN recurrence, radiological responses, and survival. Treatment-related adverse events were also evaluated. For each study, level of evidence was assessed using the 2011 Oxford Centre For Evidence-Based Medicine guidelines, and risk of bias evaluated with the Joanna Briggs Institute (JBI) checklists for case series and randomized controlled trials [18,19,20].

Statistical analysis

Continuous variables are presented as medians and ranges, and categorical variables as percentages. Two-sample weighted means t-test was performed to assess differences in the number of lesions per patient and volumes of treated lesions between bevacizumab and LITT cohorts. The time intervals between RN treatment and RN recurrence (PFS curve) or death (OS curve) were estimated with the Kaplan–Meier method. The survival analyses were conducted with the log-rank test. Indirect meta-analyses were performed for post-treatment symptomatic improvement, weaning off steroids, RN recurrence, radiological responses, and OS rates at 3–6-12–18 months. Outcomes were summarized with pooled proportions of events (effect size—ES), and confidence intervals (CI) were calculated with the Wilson score method, both graphically displayed with forest plots [21]. The Freeman-Tukey transformation was performed to include studies with 0 or 1 event rate and stabilize variance, and the DerSimonian-Laird approach for random effect models was used to account for high-variability between studies [22, 23]. Heterogeneity was assessed with the Higgins I-square (I2) and considered significant for I2 > 75% [24]. All analyses were bilateral and P-values < 0.05 were considered statistically significant. Statistical analyses were conducted using SPSS V.25 (IBM Corp, Armonk, NY) and STATA 16.1 (StataCorp LLC, College Station, TX).

Results

Study selection and quality assessment

Figure 1 illustrates the flow diagram of the literature search and study selection. The search strategy yielded 477 citations (PubMed: 289, EMBASE: 117, Scopus: 40, Cochrane: 31), of which 18 were included in the qualitative and quantitative synthesis accordingly to the pre-specified criteria (Supplementary File 1). Nine studies reported the use of bevacizumab [25,26,27,28,29,30,31,32,33]. Eight studies described patients treated with LITT [13, 34,35,36,37,38,39,40]. One study compared bevacizumab and LITT [14]. Critical appraisal based on JBI criteria returned high quality (i.e., low risk of bias) for all included articles (Supplementary File 2).

Fig. 1
figure 1

PRISMA 2020 flow diagram

Full size image

Patient demographics, clinical and management characteristics

In total, 291 patients diagnosed with RN from BM were analyzed. Patients were divided in two treatment cohorts: 143 (49.1%) received bevacizumab and 148 (50.9%) underwent LITT (Table 1). Of note, 14 patients included in the LITT cohort received late bevacizumab courses for refractory lesions, but their outcome data referred to the period before receiving bevacizumab [13, 14, 38]. Median ages were 58 (range 27–79) and 60 (range 29–83) in the bevacizumab and LITT cohorts, respectively, with a male proportion of 53.1% and 33.3%. Lung cancers represented the prevalent primary tumors, followed by melanoma and breast cancer. BM were most treated with stereotactic radiotherapy (SRT) (95.1% in bevacizumab, 100% in LITT)—including intensity-modulated radiotherapy and stereotactic radiosurgery –, and less with whole brain radiotherapy (WBRT) (39.1% in bevacizumab, 14.6% in LITT), concomitant or without SRT.

Table 1 Summary of demographics and clinical characteristics of all pooled patients grouped in treatment cohorts
Full size table

Rates of symptomatic RN were 94.4% in bevacizumab cohort and 74.7% in LITT. Lesions were mostly diagnosed with imaging in the bevacizumab cohort (97.9%), and with biopsy prior to LITT (82.4%). The mean number of lesions per patient was 1.13 in the bevacizumab cohort and 1.05 in LITT, showing no significant difference (P = 0.624). Similarly, the mean contrast-enhancing pre-treatment volumes were not statistically different (P = 0.582), namely 30cm3 in bevacizumab cohort (n = 91) and 5cm3 in LITT (n = 44). Palliative steroids were administered in 86% and 46% patients before starting bevacizumab or LITT. In the bevacizumab cohort, patients completed a median of 4 treatment cycles (range 1–31) at dosages of 5 mg/kg q2w (38.5%), 7.5 mg/kg q3w (20%), 10 mg/kg q2w (21.5%) and 15 mg/kg q4-6w (3%). Zhuang et al.[31] also reported the use of low dose bevacizumab (1 mg/kg q3w) in 21 patients (16.1%). In LITT cohort, NeuroBlate (Monteris Medical Inc., Minneapolis, MN) and Visualase (Medtronic Inc., Dublin, Ireland) systems were used, and median post-treatment hospital-stay was 1.5 days (range 0.5–6).

Outcomes, adverse events, and survival analysis

Table 2 summarizes pooled treatment outcomes. Post-treatment symptomatic improvement—i.e., reduction or resolution of RN symptoms—occurred in 73.3% patients treated with bevacizumab and 60.8% with LITT, while 66.7% and 44.1% patients achieved post-treatment steroid wean-off. Follow-up radiological assessment returned higher rates of PR in bevacizumab cohort (79.6%) and SD in LITT (49.2%), with lower rates of CR (3.6% and 8.2%) and PD (10.2% and 13.1%) in both [16, 17].

Table 2 Summary of treatment outcomes, adverse events and survival of all pooled patients grouped in treatment cohorts
Full size table

Treatment-related adverse events, reported using the “Common Terminology Criteria for Adverse Events, v5.0”, showed rates of 14.7% and 12.2% respectively in bevacizumab and LITT cohorts [41]. In the bevacizumab cohort, the most common were grade 1 and 2 bleeding and proteinuria, and grade 3 hypertension. In LITT cohort, the most frequent were grade 1 headache, and grade 3 limb weakness and seizure. Sporadic cases of thromboembolic events (4.8%) and intracerebral hemorrhage (5.6%) have been associated with bevacizumab and LITT, respectively.

RN recurrence rates were 17.2% and 22.4% in the bevacizumab and LITT cohorts, and death occurred in 56.8% and 59.6% of patients, respectively. Median PFS and OS were 3.5 (range 0–22) and 6.5 (0–38.4) months in bevacizumab cohort, and 6 (0–64.6) and 11 (1–89) months in LITT. OS rates for bevacizumab and LITT cohorts were: 90.7% and 94.7% at 3 months, 80% and 91.6% at 6 months, 39.4% and 69.3% at 12 months, and 25% and 46.4% at 18 months. As shown in Fig. 2, PFS (P = 0.209) and OS (P = 0.484) were not significantly impacted by treatment strategy.

Fig. 2
figure 2

Kaplan–Meier survival curves of patients (no.) with available individual data: A PFS (n = 22) and B OS (n = 64) of the total pooled cohort; C PFS (n = 22) and D OS (n = 64) based on treatment strategies—bevacizumab versus LITT. PFS progression free survival, OS overall survival, LITT laser interstitial thermal therapy

Full size image

Meta-analysis: comparison of clinical, radiological and survival outcomes rates

Table 2 summarizes the results of all indirect comparisons between bevacizumab and LITT, displayed as forest plots in Fig. 3. There were no significant differences in post-treatment symptomatic improvement (P = 0.187, I2 = 54.8%), steroid wean-off (P = 0.614, I2 = 25.5%), and RN recurrence (P = 0.5, I2 = 0%), between bevacizumab and LITT (Supplementary File 3).

Fig. 3
figure 3

Forest plots for indirect comparisons between bevacizumab and LITT treatments of radiation necrosis: post-treatment symptomatic improvement; post-treatment weaning off steroids; radiation necrosis recurrence; complete response, partial response, stable disease, and progression at 6-month radiological follow-up; overall survival rates at 3-month, 6-month, 12-month, 18-month. LITT laser interstitial thermal therapy, CI confidence interval, Effect effect size

Full size image

Rates of CR (P = 0.29, I2 = 44.5%) and PD (P = 0.645, I2 = 61.6%) were comparable between the two cohorts. Rates of PR were significantly higher (P = 0.001, I2 = 88.9%) in bevacizumab (ES:0.77; 95% CI 0.51–0.96) compared to LITT (ES:0.28; 95% CI 0.17–0.41), while rates of SD were significantly higher (P = 0.002, I2 = 81.9%) in LITT (ES:0.42; 95% CI 0.17–0.69) compared to bevacizumab (ES:0.04; 95% CI 0–0.12); but both with significant heterogeneity (Supplementary File 4).

OS rates at 3 months (P = 0.427, I2 = 61.9%), 6 months (P = 0.289, I2 = 75%), and 12 months (P = 0.183, I2 = 72.1%) were comparable between the two cohorts. OS rates at 18 months were significantly higher (P = 0.038, I2 = 73.7%) in LITT (ES:0.59; 95% CI 0.35–0.82) as compared to bevacizumab (ES:0.15; 95% CI 0–0.48) (Supplementary File 5).

Discussion

RN represents an inflammatory response of the brain parenchyma occurring months to years after radiotherapy. Our pooled patients most frequently received SRT for BM [11, 40, 42]. Stereotactic protocols pose higher risk of RN by delivering focused high-doses of radiation to specific targets, triggering endothelial injury, hypoxia, and local necrosis [5, 7]. Due to similarities between tumor recurrence and progression, the diagnosis of RN is challenging, mostly relying on the multidisciplinary review of clinical findings and advanced MRI scans [26, 43]. When feasible, a biopsy of the enhancing tissue may be pursued, but may render false negative or positive results due to intermingling necrosis and tumor cells [42]. We found that patients in the LITT group had higher rates of biopsy (82.4%) as compared to bevacizumab (5.6%). While pre-ablative biopsy and LITT ablation can be performed during the same session, patients receiving bevacizumab rarely undergo biopsy so as to avoid related surgical risks [13, 31, 33, 37]. Since tumor recurrence may be misinterpreted as RN on imaging but also on biopsy, the different diagnostic strategies remain a potential confounder in the analysis of treatment outcomes [14]. Based on our findings, LITT may be recommended in reasonable surgical candidates, while bevacizumab may be offered to patients with lower functional status.

Treatments may provide symptomatic relief and/or hamper lesion progression [5]. Early management with steroids suppresses inflammation and reduces brain edema, easing mass effect-related symptoms [9]. The majority of our pooled patients in both treatment groups were symptomatic upon clinical presentation (94.7% and 74.7%), mostly undergoing palliative therapy with steroids (86% and 46%). In the long-term, steroids may result in severe systemic toxicities and impaired quality of life; thus, adjunct treatments should be offered to prevent steroid dependance. Bevacizumab proved to be superior to steroids in treating RN from BM, nasopharyngeal carcinomas, and gliomas [10, 11, 44]. By neutralizing VEGF, bevacizumab counters vessel permeability and restores blood–brain-barrier function, improving short-interval clinical and radiological outcomes [5, 10]. In this review, treatment protocols were heterogeneous, ranging from 1 to 15 mg/kg cycles, corroborating the theory that the effectiveness of bevacizumab derives from its anti-angiogenic action rather than its dose [31]. The documented versatile dosage profile may increase worldwide accessibility to bevacizumab treatments by mitigating costs and dose-related adverse events [45].

Surgical excision of the necrotic foci may be pursued to resolve mass effect-related symptoms of aggressive RN lesions. However, the surgical risk and the poor baseline clinical status of patients with systemic malignancies make less invasive options more appealing [5, 9, 46]. LITT is a minimally invasive surgical ablative technique, which, by targeting peri-necrotic zones, induces the thermocoagulative necrosis of dysfunctional endothelial cells and removes the primary source of active VEGF [47]. LITT has been shown to improve functional and cognitive statuses, achieving prolonged lesion control and survival comparable to surgical resection [13, 37, 40]. The median hospital-stay of patients treated with LITT (1.5 days) was noted to be less than half that of patients undergoing craniotomy (3.9 days in previous cohorts) [13, 40]. Late bevacizumab courses were seen in 14 patients presenting with LITT-refractory RN lesions, but failed to improve clinical outcomes [13, 14, 38]. In these cases, therapeutic failures were likely related to the underlying poor clinical statuses of affected patients.

Bevacizumab and LITT had a positive impact on clinical and radiological outcomes of patients with RN. Both treatments showed favorable rates of symptomatic improvement and ability to wean off steroids, 73.3% and 66.7% in the bevacizumab group as compared to 60.8% and 44.1% in the LITT group (P-values 0.187 and 0.614, respectively) [14]. Bevacizumab may exhibit modest clinical advantages due to its direct effects on vessel and blood–brain-barrier permeability. However, our findings may be ascribed to the diversified clinical assessments amongst studies and to the complex multidisciplinary decision to modulate steroids therapy in patients with systemic malignancies. The introduction of standardized assessments and reporting of performance status scores may produce more accurate clinical data. Similarly, the pooled low rates of radiological RN recurrence were comparable between the two groups, but the absence of consistent histopathology reports may have failed to exclude cases of tumor progression [25, 30, 34, 35].

RN lesions treated with bevacizumab and LITT followed different patterns of radiological responses based on the modified RANO criteria for BM. At 6 months, while rates of complete response and disease progression were low and comparable between the two groups, rates of partial response were statistically higher in the bevacizumab cohort (P = 0.001, I2 = 88.9%) and rates of stable disease higher in LITT (P = 0.002, I2 = 81.9%); however, our findings are limited by their significant heterogeneity. This is likely related to the different mechanism-of-action of each treatment modality [14]. Bevacizumab may promptly restore blood–brain-barrier function, reducing brain edema and post-contrast lesion volumes upfront; but the lack of permanent anti-inflammatory effects may lead to increased volumes after prolonged cessation of treatment [14, 32, 33, 48]. In contrast, LITT ablation may directly and permanently inactivate inflammatory cells, resulting in apparent early disease progression due to transient increased lesion size; but, upon follow-up, the lesions decrease in volume and stabilize for a longer duration [14, 36, 37, 40]. Despite the rapid clinical effect of LITT, it is common knowledge that LITT-treated lesions expand radiologically during the first 4–6 months post-treatment due to the enlarging necrotic cores, but shrink and/or disappear upon late follow-up (12–15 months) [49].

In our pooled data, the median OS was longer in patients treated with LITT (11 months) as compared to bevacizumab (6.5 months), but the difference was not statistically significant. The low median OS in our pooled-data may reflect the limited patient-level survival data reported amongst studies, most of which restricted their follow-up times to 6 and 12 months [25, 27, 32,33,34, 37, 38, 40]. In regards to OS rates, we found longer survival rates in the LITT group, reaching statistical significance at 18 months (P = 0.038) [14, 26, 28, 31, 37, 38, 40]. The likely uneven distribution of patients between the two treatment groups may explain the observed differences. No significant difference was seen in the number-of-lesions per patient and pre-treatment volumes between the bevacizumab and LITT cohorts, although the mean volume of RN was much higher in the bevacizumab cohort (30 cm3) as compared to the LITT cohort (5 cm3). The low number of patients with data related to pre-treatment volumes did not allow statistical significance. However, it seems that patients who received bevacizumab had larger pre-treatment volumes to start with as compared to patients who received LITT, as most surgeons likely avoid treating larger lesions with LITT, which may provide a false impression that LITT is associated with better outcomes. This finding needs to be further studied in a controlled/randomized fashion, as a higher number of patients is probably needed to detect difference. Further, the intracranial extent of BM and RN lesions and their anatomical location in “eloquent” or “non-eloquent” areas were poorly defined in both groups. We speculate that patients with multiple lesions, or located in “eloquent” cortex, and with poor baseline functional status were probably not ideal candidates for LITT but were eligible for bevacizumab.

Both treatment modalities showed favorable toxicity profiles and proved to be safer than long-term steroids and, in some cases, surgical resection of RN [10, 40, 44]. The adverse events are listed in Table 2. In the bevacizumab cohort, bleeding and proteinuria were often self-limited after temporary treatment interruption [14, 25, 29, 31]. In the LITT cohort, headache and mild motor impairments were mostly transitory or managed with short-term steroids [36,37,38]. In rare cases, severe thromboembolic events and intracerebral hemorrhage were linked to bevacizumab and LITT, but no life-threatening complications were described [14, 27, 35]. Patient receiving bevacizumab may require up to 4-weeks of “wash-out” before qualifying for surgery or clinical trials [50]. These limitations may strongly influence the inclusion criteria among the two treatment strategies [11, 14].

Limitations

Except for three prospective studies, most included studies were retrospective with class IIIb-IV evidence, prone to selection and recall biases. Patient-level data was limited, with most data collected at study-level from heterogeneous populations and treatment centers. Possible clinical confounders could not be investigated due to the limited data on performance statuses in both cohorts and on pathology reports in the bevacizumab group. Data on lesion volume and per-patient number-of-lesions were also limited and not equally distributed between the two cohorts, thus likely responsible for the lack of statistical difference despite the numerical difference in averages. Finally, follow-up intervals varied amongst included studies and treatment groups, which made it hard to draw robust conclusions about survival outcomes.

Conclusion

This meta-analysis compared the role of bevacizumab and LITT in the treatment of RN in BM. Both strategies showed good safe toxicity-profiles and equal efficacy in relieving symptoms, weaning off steroids, and achieving local lesion control. Patterns of radiological responses were different and LITT resulted in longer overall survival likely related to its use in patients with smaller lesions and better baseline functional-status.