Percutaneous image-guided thermal ablation of bone metastases: a retrospective propensity study comparing the safety profile of radio-frequency ablation and cryo-ablation

Abstract Objectives To retrospectively compare the safety profile of percutaneous image-guided radiofrequency ablation (RFA) and cryoablation (CA) of bone metastases (BM) with and without a propensity score analysis. Methods Between January 2008 and April 2018, 274 consecutive patients (mean age 61.6 ± 12.1 years) with BM were treated at our Institution with RFA (53 patients; 66 BM) or CA (221 patients; 301 BM) and included in this study. Complications were assessed according to the type of ablation modality before and after applying a 1:1 propensity score method taking into account patient’s demographics, BM features, procedural details and follow-up findings. Results In the whole 9 BM (2.5%) reported major complications without significant difference between RFA (1/66; 1.5%) and CA (8/301; 2.7%; p = 1); 40 BM (10.9%) showed minor complications, which were more common with RFA (22/66; 33.3%) than with CA (18/301; 6.0%, p<.001) mainly due to post-procedural pain occurring more frequently with RFA than CA (20/66; 30.3% vs. 7/301; 2.3%, p<.001). Following 1:1 matching, similar results were obtained, since there were similar rates of major complications with RFA and CA (1/66 [1.5%] and 0/66 [0.0%], respectively; p = 1); and higher rates of minor complications with RFA compared to CA [33.3% (22/66) vs. 2/66 (3%); p<.001] due to preponderant postprocedural pain (90.9% [20/22] minor complications with RFA). Conclusions Similar low rates of major complications are expected with RFA and CA of BM. In the post-operative period, RFA appears more painful than CA, thus warranting for adoption of dedicated analgesic protocols for patients undergoing RFA.


Introduction
Bone metastases (BM) represent a frequently encountered complex clinical scenario [1], often requiring a multidisciplinary approach to management, involving orthopedic surgeons, oncologists, histopathologists, as well as diagnostic and interventional radiologists. Treatment plans agreed by the tumor board are often conferred based on the patient's demographics, histological grade and stage as well as on the symptomatic status of BM resulting from numerous skeletalrelated events (SREs) such as pain, fractures, nerve root compression and hypercalcemia, warranting palliative treatment [2]. Less often treatment is proposed with a curative intent of achieving effective locoregional tumor control in selected patients presenting with oligometastatic or oligoprogressive disease [3].
In the last few decades, the therapeutic armamentarium for BM has broadened to include percutaneous thermal ablation (TA) techniques, available to the multidisciplinary tumor board to provide palliative or curative treatment. Radiofrequency ablation (RFA) and cryoablation (CA) [4][5][6][7] are the two most commonly used and investigated techniques, and share comparable effective clinical results in the palliative [4,5,8,9], and curative [9,10] setting. Moreover, the safety of RFA and CA has been proven, with series reporting rates of overall/major complication of 16%/5% and 9.1%/ 2.5%, respectively [11][12][13][14][15]. Nevertheless, there are no formal explicit analyses directly comparing RFA and CA, specifically in terms of safety; and two large recently published series specifically investigating the safety of bone tumor RFA and CA could not formally provide a definitive appraisal due to the lack of direct comparison between these two techniques; heterogeneous type of targeted tumors including primary/ secondary, and benign/malignant tumors with variable histological grading; as well as due to the bias related to the selection of the TA technique, which was mainly operator's driven, generally according to the radiological category of the index tumor (e.g., RFA not offered for sclerotic tumors; CA applied to sclerotic/mixed tumors, etc.) [12,15]. Therefore, we have hypothesized that a direct comparison among RFA and CA in 1:1 matched populations including BM with comparable features could provide more robust evidence regarding the safety profile of these two common ablation techniques. Therefore, the aim of this study was to retrospectively compare the safety profile of percutaneous image-guided RFA and CA of BM before and after applying a propensity score analysis with the intent to lessen the influence of patient's, BM, procedural and follow-up factors. Our study specifically focused on BM since these definitely represent the most common type of bone tumors undergoing TA.

Materials and methods
This retrospective study was approved by our institutional review board with permission to perform chart review, with a waiver to written informed consent. All research was performed in accordance with relevant regulations.

Study population
All consecutive patients with BM undergoing percutaneous RFA and CA in a single tertiary university hospital between January 2008 and April 2018 were identified by research performed in the institutional radiological information system. Three different key-words were simultaneously entered ('bone metastases', 'cryoablation' and 'radiofrequency ablation'); and a total of 320 consecutive unpaired patients were identified: 81 had received RFA, and 239 had received CA; 28 patients treated with RFA and 18 with CA were excluded due to loss of follow-up data subsequent to hospital discharge.  [12,15].
The decision to perform TA was established following discussion in a multidisciplinary tumor board meeting including oncologists, orthopedic surgeons, anesthesiologists and interventional radiologists. Patients were referred for consideration of percutaneous TA if they required either a palliative treatment due to painful BM not previously treated with, or refractory to standard treatments (including analgesics and radiotherapy [RT]); either a curative treatment due to an oligometastatic (<3 metastases with tumors size 3 cm) or oligoprogressive disease (1-2 metastases not responding to systemic therapy despite an overall stable systemic disease) [9,10].
Patients were not offered TA if they presented with an irreversible coagulopathy, signs of active infection, life expectancy less than 1 month, mechanically unstable spinal lesions, focal neurological deficit related to the BM or tumors for which TA was deemed technically unfeasible due to the high risk of damage to adjacent structures (nerves, vessels, solid/hollow organs, cartilage or skin) in spite of extensive application of thermo-protective measures [16].

Thermal ablation procedures
Procedures were performed by seven interventional radiologists with variable (1-13) years of experience.
Patients underwent either CA ( Figure 2) or RFA ( Figure 3) at the discretion of the interventional radiologist performing the procedure. All interventions were performed under conscious sedation, spinal or general anesthesia in a strict sterile environment under computed tomography (CT) or conebeam CT (CBCT)-guidance. Additional fluoroscopic guidance was applied when required.
One or more cryoprobes (IceSphereVR, IceRodVR or IceForceVR and Galil medical) and RFA electrodes (Cool-tip or Osteo-Cool, Medtronic) were used for each BM according to its size and goal of the procedure. When multiple probes/ electrodes were applied, simultaneous activation was performed to achieve a synergistic effect.
For curative treatments, the ablation area was planned to cover the entire BM volume with an additional 5-10 mm safety margin. For palliative treatments, the minimal goal set was to achieve effective ablation of the BM/normal bone interface and, whenever possible, complete ablation of the entire BM.
Whenever necessary, adjuvant thermo-protective measures including carbo-or hydro-dissection (Figures 2 and 3), nerve root electrostimulation or thermal monitoring through thermocouples were applied to avoid iatrogenic damage to nearby non-target critical structures located 1 cm within the BM [16]. Subsequent mechanical stabilization was offered during the same interventional session with osteoplasty and/ or osteosynthesis (depending on the location of the tumor Patients who received thermoablation for bone metastases between January 2008 and April 2018 were identified through RIS using the following keywords ("bone metastases" AND "cryoablation" OR "radiofrequency ablation") N=320 (239 CA; 81 RFA) Figure 1. Study flow-chart. and the regional biomechanics [17]) for BM considered at high risk of secondary fracture.

Data collection
Chart review was performed by PA and PDM (interventional radiologists with 2 and 1 years of experience, respectively), who were blinded to procedural, clinical and BM details during data collection. Chart review was always performed by single authors. Intra-and inter-observer variability was not specifically assessed.  . Axial CT image (D) shows the maximal iceball (arrow heads); in this case, the supraspinatus muscle was hydrodissected (black arrow). A hydrodissection of the suprascapular nerve was also performed with a 22 G needle placed in the suprascapular notch (not shown). In this patient, a symptomatic post-ablation fracture occurred 1 month after the procedure as shown on coronal CT reconstructions (arrows in E). This fracture was treated by percutaneous screw fixation (arrows in F). [3][4][5], and minor (CTCA grade 1-2) [18]. Among the assessed complications, postprocedural pain was recorded in our study when an increase of 3 or more points on the 0-10 visual analog scale (VAS) was noted at the treated site compared to the patient's preprocedural status. The complication rate was reported as the number of complications divided by the number of BM.

Statistical analysis
Descriptive statistics were used to present the results. Mean and standard deviation or proportions were reported for continuous or categorical variables, respectively.
Complication rate analysis was performed per tumor according to the intention-to-treat principle.
The Student t-(for continuous variables) and the Chisquared (for categorical variables) or Fischer's tests (when the number of events was too low) were used to compare the two unmatched RFA and CA populations to highlight BM and procedural characteristics, as well as safety outcomes. Nevertheless, to confirm or reject the findings achieved by this analysis derived from such heterogeneous non-comparable populations, we have matched 1:1 BM treated with RFA and CA according to a propensity score method taking into account patient's demographics (sex, age, ECOG-PS, previous RT and goal of the treatment), BM features (size, location, radiological features, cortical disruption and presence of nearby critical structures), procedural details (n. probes, thermo-protective measures and subsequent bone consolidation) and follow-up findings. The nearest neighbor approach was used. A p value <.05 was set to be statistically significant. Statistical analysis was performed using R statistical software version 3.6.3 (R Foundation for Statistical Computing, Vienna, Austria).

Baseline characteristic in the unmatched population
Baseline characteristics of the study population are summarized in Table 1

Discussion
In this study, the overall complication rate for TA applied for the curative or palliative treatment of BM was 13.4%, with minor complications being largely more common (10.9%) than major ones (2.5%). In the unmatched population, the incidence of major complications was similar with RFA and  CA (1.5 vs. 2.7%); which was not the case for minor complications, that were more likely to occur with RFA when compared to CA (33.3 vs. 6.0%) primarily due to the excessive incidence of immediate post-procedural pain with RFA compared to CA (30.3 vs. 2.3%). Such trend remained substantially unchanged in the RFA and CA populations resulting from the propensity matching that we have applied to lessen the influence of several confounding factors dealing with patients, BM, procedural and follow-up features. In fact, after matching the two populations, the major and minor complication rates with RFA and CA were 1.5% and 0.0%, and 33.3 and 3.0%, respectively; with post-operative pain accounting for 90.9% of the minor complications noted in the RFA group. Except for the high rate of postprocedure pain, which in our series was considered as a complications since it required an increased level of care, these results of ours are in line with literature on TA of BM reporting rates of minor and major complications ranging between 16 and 24%, and 7 and 8%, respectively [11,12,14,15,[19][20][21].
Coming to the analysis of the type of complications encountered in our study, pathological fracture was the most common major complication occurring in the CA group, with fractures occurring (in two acetabulae, one iliac bone and one scapula) approximately 2 months after the treatment, and requiring percutaneous fixation. This is recognized following bone tumor TA and results from loss of structural and mechanical integrity of the bone due to tumor and post-TA necrosis, which necessitates immediate bone consolidation during the same interventional session of TA to prevent such sequela [13][14][15]17]. This represents a major advantage of percutaneous TA when compared to RT [13,14,22], which is also affected from a non-negligible risk of secondary fracture particularly when stereotactic techniques are used. Indications to perform bone consolidation should be based on Mirel's score for long bone, SINS score for spine and should also take into account the size of the target tumor and the cortical involvement [17,23,24]. Cementoplasty is the most adapted consolidation technique for bones mainly undergoing compressive stresses (e.g., vertebral body). On the other hand, screw-mediated osteosynthesis should be applied in bones mainly undergoing shearing stress, or whenever there is a need to bridge a fracture [17,25]. Other consolidation techniques have been proposed; nevertheless, many of them lack long-term biomechanical assessment [26][27][28][29]. Accordingly, their use should be evaluated case by case within the multidisciplinary tumor board. Thus, interventional radiologists may promote TA of BM in the multidisciplinary tumor board by levering on the potential to integrate TA with percutaneous consolidation to prevent secondary pathological fractures, which carries high morbidity and mortality risks independently of the tumor type/stage.
On the other hand, the sole major complication experienced in the RFA population was a lethal septic shock developing at an RFA site. Therefore, strict sterile intra and postprocedural conditions and avoidance of any percutaneous procedure in case of suspected local or systemic infection are strongly recommended throughout literature [5,15,16,30].
Concerning minor complications, postprocedural pain was the salient adverse event after RFA, which was not the case with CA. This may be partially attributed to the analgesic properties of the ice-ball, which may significantly contribute to lowering postprocedural pain after bone tumor CA; and conversely by the pro-inflammatory properties of the necrosis associated with RFA [31][32][33][34]. Aside from the pathophysiology of such event, there are practical concerns of poorly managed post-RFA pain, which has driven us to consider pain as a possible post-operative complication. In fact, we share the view from Thacker et al. who have reported that the intense post-RFA pain developing within the first few post-operative hours and often requiring active analgesia, may negatively impact the length of the in-hospital stay [35]. Moreover, intense postoperative pain represents a major element considered whilst grading complications in the most common complication scales used in the field of interventional radiology (SIR, CIRSE and CTCAE scales) [18,36,37]. To overcome such consequence, we advocate the adoption of dedicated institutional analgesic protocols for patients undergoing RFA, and we present in Table 5 our recently adopted protocol, which is based on a standard systemic intraprocedural multimodal analgesia administered by the anesthesiologist, which is further reinforced by a central or peripheral nerve block performed by the interventionalist immediately following the RFA. Although we do not dispose of large prospective data on the presented analgesic protocol of ours due to its recent introduction, preliminary results from the first five consecutive patients treated with it revealed interesting results, with 24-h post-ablation pain never reaching levels >2/10 on the 0-10 VAS compared to baseline.
Other common minor complications were almost equally distributed in RFA and CA populations and included postablation neural damage, which needed conservative management with steroids with subsequent remission. Many of these minor complications were attributed to damages to adjacent non-target structures, which occurred despite the large use (71.4%) of thermo-protective measures, often in the form of hydrodissection (79%), thermal monitoring (40%) or a combination of both of them.
Limitations of this study include its retrospective design, which did not allow prospective randomization of the population according to the TA technique resulting in unmatched groups of treatment that were mainly selected on a per tumor basis by each operator. To overcome this limitation, we have applied the propensity score method to match the RFA and CA populations; and although similar safety profiles were obtained in the unmatched and matched populations, Table 5. Analgesic protocol recently adopted at authors' institution.

Type of intervention
Bone tumor location Management Systemic multimodal analgesia All tumors -Non-opioid drugs (paracetamol, NSAID, nefopam) -Opioids: -Tramadol for moderate pain -Morphine for severe pain Central nerve block From T10 to the feet -Epidural injection of 5-10 ml Naropin 2%

Peripheral nerve block
Limbs -Local injection 10-20 ml Naropin 2-7.5% NSAID: non-steroidal anti-inflammatory drugs results obtained with the latter allowed more robust appraisal of the safety profiles of these two commonly used TA techniques. Moreover, we did not report the in hospital stay; however, since post-procedural pain was systematically graded as minor, it is unlikely that it had resulted in a significant prolongation of the hospital stay [18,36,37]. Furthermore, some patients referred to our institution for TA of BM have only benefited from 1-month follow-up consultation in our department. Therefore, we could not entirely rule out the potential for further complication detected in further follow-up reviews. Lastly, we compared only the most established ablation techniques (RFA and CA) despite there are emerging techniques such as microwave ablation and high-focused ultrasound. Nevertheless, experience with these new techniques is limited, and mainly focus on palliative treatments, often without large agreement on ablation protocols [38].
In conclusion, CA and RFA of BM performed for curative or palliative reasons are both safe given the very low rate of major complications. Nevertheless, minor complications are more common after RFA due to high rates of post-operative pain, which should encourage the adoption of dedicated institutional analgesic protocols for patients receiving RFA.

Disclosure statement
No potential conflict of interest was reported by the author(s).