The value of using a brain laser interstitial thermal therapy (LITT) system in patients presenting with high grade gliomas where maximal safe resection may not be feasible

Background The objective of this analysis was to determine the value (incremental cost/increment benefit) of a brain LITT system versus employing current surgical options recommended by NCCN guidelines, specifically open resection (i.e. craniotomy) methods or biopsy (collectively termed CURRENT TREATMENTS) in patients where maximal safe resection may not be feasible. As has been demonstrated in the literature, extent of resection/ablation with minimal complications are independently related to overall survival. Methods A cost effectiveness analysis from a societal perspective was employed using TreeAge Pro 2014 software. Direct costs (using national average Medicare reimbursement amounts), outcomes (overall survival), and value [defined as increment cost/incremental survival—evaluated as cost/life year gained (LYG)] were evaluated. Sensitivity analysis was also performed to determine which variables had the largest effect on incremental costs and outcomes. Results In the base case, the overall survival was improved with brain LITT versus CURRENT TREATMENTS by 3.07 months at an additional cost of $7508 (or $29,340/LYG). This amount was significantly less than the current international threshold value for $32,575/LYG and considerably less than the US threshold value of $50,000/LYG. This incremental cost may also qualify under NICE criteria for end of life therapies. In sensitivity analysis: As percent local recurrence GBM increased; cost of DRG25/26 increased; percent GTR increased; and gliadel use increased—the value of brain LITT improved. Additionally, in those patients where a biopsy is the only option, brain LITT extended life by 7 months. Conclusions Brain LITT should be considered a viable option for treatment of high grade gliomas as it improves survival at a cost which appears to be of good value to society. This incremental cost is less than the international and US thresholds for good value.


Background
According the Central Brain Tumor Registry of the US (CBTRUS), there are over 138,000 people in the US living with primary brain and central nervous system malignant tumors primary in the United States [1]. The incidence of primary malignant brain tumors is expected to be over 23,000 in 2015 [1]. Of these types of tumors, >50 % [2][3][4] (approximately 11,500) are classified as being at high risk [for complications] for resection. This is mainly due to the grade of tumor and these tumors residing in or near areas of eloquence or being deep seated in nature (e.g. tumor residing in the brain stem [also referred to as complex anatomy]). Extent of resection (EOR) with the aim of maximal cytoreduction of the tumor is strongly correlated with outcomes (i.e. survival and function as classified under the Karnofsky performance scale) [5] and the effectiveness of other treatment modalities such as radiation or chemotherapy [6,7]. One of the main issues with tumors that are in or near areas of eloquence or, that are deep seated in nature, is an inability for neurosurgeons to adequately resect the tumor without causing longer term neurological complications from surgery (i.e. open resection or biopsy or CURRENT TREATMENTS). Craniotomy procedures that have been performed on high-grade gliomas in or near areas of eloquence have historically resulted in neurological complications (i.e. functional and/or cognitive deficits on a neurological basis) that are permanent in nature and result in suboptimal resection. These major complication rates range from 4.5 to 13 % in large cohorts of patients [8][9][10][11][12][13][14][15][16][17] and also result in suboptimal EOR of 78-<95 % [5,8,[12][13][14] . Further, these acquired neurological complications resulting from surgery result in decreased median survival rates [18]. While the National Comprehensive Cancer Network (NCCN) includes the use of craniotomy or biopsy [19] [referred to as CURRENT TREATMENTS moving forward] for subtotal resection in its treatment algorithms for primary or recurrent glioblastoma (where maximal resection is not safe or feasible), it does not make evidentiary or consensus recommendations on their use [19].
Recently new MRI guided laser interstitial thermal therapy (LITT) systems for ablating neurological softtissue have been FDA cleared and; covered and paid for by Medicare (via a new technology add on payment) in treating primary and recurrent gliomas [20,21]. These technologies, have been reported on extensively in the literature, including in two systematic reviews [22,23]. These technologies apply focused laser energy which a surgeon uses to ablate tissue such as tumors from the inside of the brain (using a bur or twist drill hole for brain access). Real time MRI thermometry is also used so that surrounding healthy tissue damage can be minimized. These systems allow surgeons to selectively ablate tumors and lesions in the brain that may have been previously deemed inoperable, difficult to access, or unsafe to resect based on their location in or near areas of eloquence.
In examining the use of brain LITT in the peer review literature in the subset of patients whose high grade glioma resides in or near areas of eloquence or; that are deep seated in nature, it was found that the major complication rates directly resulting from the brain LITT surgery were in the 0-6 % range; average of 2.7 % (Table 1), which is lower than 4.5-13 % seen in CURRENT TREATMENTS (Table 2). (Note: The analysis as found in Table 1 shows all types of brain LITT used in high-grade gliomas located in areas of eloquence over the years, with and without the use of MRI guidance.) As well, it has been found that physicians who are experienced in using brain LITT technology in complex anatomy experience an EOR approaching 98 % [24] and that those who are treated with brain LITT experience a length of stay (LOS) that averages 3 days [24]. This is in contrast to patients who are treated with open craniotomy with ICD9CM Diagnosis codes 191.0-191.8, whose LOS average 6.55 ± 1.77 days under diagnostic related groups [DRG] 25-27 (craniotomy with and without comorbidity/complication) [25].
In establishing the value of a new treatment, the new option is compared to the weighted costs and outcomes of the combined existing treatments [CURRENT TREATMENTS] for the same patient population. In the United States cost effectiveness ratios of <$50,000/life year gained [LYG] are considered attractive [26]. From an international perspective, <30,000€ (or $32,575 in current US dollars)/LYG [or at $2714/month survival gained] is considered a good value [27].
It is with these facts in mind that an analysis was undertaken to evaluate the direct costs and overall survival of treating complex high-grade gliomas utilizing either brain LITT or CURRENT TREATMENTS (per the NCCN CNS practice guidelines [19]). The costs and overall survival (OS) were evaluated via a cost effectiveness analysis (including sensitivity analysis) as described below. The hypothesis being tested is that the use of brain LITT in these types of patients would be considered cost effective (i.e. of value) at a willingness to pay (WTP of <$32,575/LYG or incremental cost $2714/incremental month of survival) in patients with complex brain anatomy (which included brain tumors in or near areas of eloquence or; in deep seated tumors which are difficult to access via surgery). This analysis examines this incremental cost/incremental survival benefit (termed ICER).

Methods
A decision tree was developed to evaluate the cost-effectiveness of using brain LITT versus CURRENT TREAT-MENTS (collectively craniotomy ± gliadel wafer, plus biopsy) in patients with complex anatomy. Additionally, brain LITT was compared to the separate procedures of craniotomy without gliadel wafer, craniotomy w/gliadel wafer, and biopsy only (which collectively make up CURRENT TREATMENTS) in these patients. The software program used was TreeAge Pro 2014, a decision tree/Markov modeling software program widely used in health care for evaluating cost effectiveness.
The decision tree evaluated the initial procedure and the resultant outcome (i.e. gross total resection [GTR]; subtotal resection [STR])-using probabilities as identified in the peer-reviewed literature and as found in Tables 3, 4. Further it was assumed that patients received adjunct care (e.g. chemotherapy, external beam radiation therapy (EBRT) where appropriate based on the EOR and; as per the NCCN guidelines [19], and evidencebased recommendations [28,29]. Patients were followed through the treatment decision tree until they died. Gross total resection was defined as an EOR of ≥98 % and a subtotal resection (STR) was classified as less than <98 %. The outcome of the surgery was also evaluated based on the resultant Karnofsky performance scale (KPS). The progression free survival (PFS) of the initial procedure was based off of the EOR as found in the literature and as outlined in Tables 3, 4 below. Progression free survival times and KPS were determinants of when/whether a second procedure was performed. Further, patients whose recurring tumor was local in nature were treated with a second surgery and follow on adjunctive treatment where appropriate, based on the clinical guidelines (e.g. ± gliadel wafer [based on evidence from databases and the literature this occurred 10-30 % of the time] [30,31], systemic chemotherapy). Patients whose tumor was diffuse in nature were treated with either palliative care or EBRT (external beam radiation therapy) ± chemotherapy, depending upon their KPS. As well, patients whose resultant KPS was <70 after surgery were treated with palliative care for the remainder of their lives. Thus as an example: if a primary procedure under brain LITT resulted in a GTR and the outcome was favorable (e.g. KPS >70 post procedure) (note: the primary procedure would include adjunctive EBRT plus chemotherapy)the PFS time as identified in the literature was used for determining a second procedure. If the tumor recurred locally, a second brain LITT procedure was performed. If the outcome of the second procedure was favorable (e.g. KPS >70) the patient was treated with follow on therapy (e.g. chemotherapy) and then followed for the remainder of their life. The decision tree for treatment followed the clinical guidelines as found in the NCCN CNS clinical practice guidelines [19] . Neurocognitive complication rates as identified above were also used in the model and affected treatment options and downstream costs such as rehabilitation post procedure.
Direct societal costs used in the model were derived from 2015 Medicare national averages for surgery (including hospital in-patient and physician services rendered during the inpatient stay) and follow on care (e.g. radiation therapy, chemotherapy, rehabilitation [for a complication] or palliative/home hospice care). These costs can be found in Tables 3, 4. Table 5 shows what the overall costs for an acute inpatient stay for tumor removal and; is be based on a weighted average use and cost for DRG's 25-26 (using 2012 Medicare data on the incidence of each procedure and 2015 National average reimbursement rates for Medicare) [32] plus the physician services rendered for each treatment type      (brain LITT, craniotomy, biopsy). Table 5 also shows the national average weighted costs for DRG 23 (when craniotomy with gliadel wafer placement was performed). Effectiveness was evaluated as overall survival (OS) of the patient. These values were derived from the literature and based on the outcome of the first and second (if indicated and based on the clinical guidelines) surgical procedures. Costs and OS were discounted at 3 % annually-which is the most commonly used discount rate for medical therapies [33]. Costs and OS used normal distributions along with confidence intervals and standard deviations for probabilistic analysis. Cost effectiveness analysis (incremental cost of using brain LITT plus other interventions (includes such interventions such as: adjunctive therapies, treatment for complications, hospice care) over the life of the patient/LYG; termed incremental cost effectiveness ratio (ICER) moving forward) was analyzed to determine whether the incremental cost/incremental survival was under internationally accepted cost/LYG thresholds. Sensitivity threshold analysis was performed to determine which variables had the greatest effect on the ICER. Lastly, Monte Carlo simulation (expected value for 10,000 simulated trials) was also run. These analyses were performed using TreeAge Pro 2014 (TreeAge Software, Inc., Williamstown, MA). Figure 1 depicts a section of the decision tree related to treatment with brain LITT for gross total resection (≥98 % tumor ablated). Table 6 compares the overall costs and survival in employing either brain LITT or CURRENT TREAT-MENTS (again which is comprised of either open resection ± gliadel wafer or biopsy) in cases where high-grade gliomas reside in or near eloquent areas of the brain or are deep seated. As can be seen in the base case, the additional costs (over the lifetime of the patient) with LITT vs. CURRENT TREATMENTS is $7508 and the overall improved survival with brain LITT vs. CURRENT TREATMENTS is 3.07 months. More specifically, as it relates to cost effectiveness, for every month in survival gained, it would cost an additional $2445 in using brain LITT versus CURRENT TREATMENTS. This translates into an incremental cost per LYG of $29,340 when using brain LITT. If one examines brain LITT compared to each option separately (that is the procedures contained within CURRENT TREATMENTS) the increment costs per LYG are: $8458/LYG compared to craniotomy (which consisted of a combination of craniotomy with and without gliadel wafer) and $48,552/LYG when compared to biopsy (Table 7). Table 8 shows similar findings with a Monte Carlo simulation.

Results and discussion
Sensitivity analysis performed via a tornado plot (Fig. 2) showed that with a willingness to pay (WTP) of $2714/ month of survival (same as the international threshold of $32,572/LYG) the following variables had the greatest effect on the model: Percent local recurrence of the GBM (Fig. 3)-with the higher the occurrence of local GBM recurrence (vs. diffuse recurrence) the more likely brain LITT was to be cost effective; the higher the cost of a craniotomy procedure (i.e. DRG 25/26)-the more cost effective brain LITT became (Fig. 4); the higher the likelihood (or probability) of a subtotal resection (versus GTR); the less cost effective brain LITT became (Fig. 5) and; the higher the probability of use of gliadel wafers, the less cost effective brain LITT became (Fig. 6) (Note that with Fig. 6, the higher use of gliadel wafers resulted in a negative incremental cost/negative OS (resulting in a positive ICER for brain LITT-which in reality shows that the ICER is reflective of the additional cost and additional survival with gliadel versus brain LITT). What Fig. 7 further clarifies is that brain LITT dominates craniotomy plus carmustine use in that it is both less expensive and produces improved overall survival (Note: Strategies that dominate are depicted in the lower right hand corner of cost effectiveness graphs and; strategies that are dominated are shown in the upper left hand corner of the same graph). Lastly Fig. 7 shows that at a WTP of $2714/additional month of survival, the favored strategy is brain LITT (as the WTP intersects with the brain LITT data point). Table 9 shows the likelihoods of a "good performance status (Karnofsky score ≥70) post surgery with and without GTR and; the costs and OS survival associated with each (as calculated by TreeAge 2014). As mentioned above, GTR is defined as an EOR of ≥98 % (with subtotal resection <98 % EOR). What Table 9 also shows is that for the "ideal outcome" for brain cancer surgery (i.e. a GTR with good performance status): this occurs 36 % of the time in brain LITT surgery (with an OS of 22.58 months); 9 % with craniotomy without gliadel wafer (with an OS of 21.75 months) and; 8 % of the time in craniotomy w/gliadel wafer (with an OS of 25.05 months). (NOTE: Biopsy was not factored in this analysis since the result is <98 % EOR). In other words there appears to be 4× higher likelihood of having a good functional outcome along with GTR using brain LITT than with the other options available (Fig. 8).
Based on current US (<$50,000/LYG) and International (<30,000€/LYG or $32,575/LYG) threshold values for value, brain LITT should be considered to be "of value"   as; its incremental cost/LYG gained (compared to CUR-RENT TREATMENTS) of $29,340/LYG is less than the thresholds accepted as good value [26,27]. If one were to examine a comparison in resection of the tumor (i.e. brain LITT compared to craniotomy-the most reasonable side by side comparison to make), the value is significantly improved at $8458/LYG; again an amount considered to be of good value and; well below accepted thresholds internationally and in the US. As well, compared specifically to other cancer therapies, brain LITT represents a better value in money spent in extending survival [27,34] . Additionally, these findings have important implications for providers, payers, and patients. For providers, the use of brain LITT may extend the overall survival in these types of patients and, at a stable or possibly a better KPS (based on reduced neurological complication rates which in turn would lower KPS scores). Secondly, for payers, this represents good value based on accepted value thresholds. For patients, it appears that OS may be improved based on improved EOR; with less of a likelihood of ending up with surgical complications-which in turn can compromise cognitive and physical functioning. There are currently no evidence based recommendations for resecting high-grade gliomas when maximal safe resection is not feasible as per the NCCN guidelines [19]. While both subtotal resection and biopsy are mentioned in the NCCN clinical practice guidelines as surgical options, they have significant limitations. Both biopsy and open resection generally result in no or suboptimal resection (<98 %) [5,8,[12][13][14], respectively, with a high rate of surgically related complications in the resection group (4.5-13 % [8, 10-14]) when used in these types of patients. As well, patients who had perioperative complications with open resection are less likely to receive radiotherapy or chemotherapy-thus affecting their survivability [27].
The ideal outcome in these types of patients is to achieve GTR without postoperative neurologic complications as; cytoreduction (EOR) without complications plays a very important role in overall survival [12,27]. Recent advances in less invasive brain LITT (under real time MRI guidance) have produced promising results, with lower complication rates (Table 1). What this decision model demonstrates is that by improving the EOR and lowering procedure related major complications via brain LITT, the overall costs for treating these type of patients over their remaining lives increase minimally (with the additional costs being incurred post procedure via adjunct therapies used to improve OS-based mainly on the ability of clinicians to more frequently/effectively use adjunct therapies with brain LITT). These adjunct therapies can be used due to a greater volume of tumor removal with lower complication rates [6,7,35]. The main reason for this is the ability of the brain laser under MRI guidance to selectively ablate cancerous lesions in and around areas of eloquence with less perioperative complications than CURRENT TREATMENTSs. Perioperative complications (including surgically acquired motor, sensory and cognitive deficits) have been found to be an  independent risk factor associated with overall survival [35]. An added benefit of the use of brain LITT is a decreased length of hospital stay which in turn also reduces overall costs (i.e. lower physician related costs for inpatient care) [20]. Additionally, a potential benefit of brain LITT versus open resection may be the ability of patients to ambulate more quickly, potentially reducing the incidence of venous thromboembolism (VTE), which can be high in brain cancer and whose risk is further increased post craniotomy [36].
The use of gliadel wafers in high grade gliomas is controversial [37] and their use appears to be practiced judiciously in the US despite being recommended by the NCCN guidelines [19]. The model attempted to account for this and used ranges found in the literature and in publicly available datasets [30,31].
While the incremental cost/LYG using brain LITT versus biopsy only ($48,552/LYG) exceeds the international threshold of $32,575/LYG, it is lower than the US threshold of <$50,000/LYG and thus would be considered acceptable in the US. Additionally, incremental cost effectiveness ratio (ICER) threshold values established by countries such as the UK (via NICE) for end of life therapies (with the criteria for consideration under this being:  life extension of >3 months, small patient population, and prognosis of <24 months) may be more flexible than the threshold set of £20,000-£30,000 for all other therapies/diagnostics/interventions [38]. Thus it may meet the UK threshold. More importantly however, brain LITT should be compared to CURRENT TREATMENTSS as; brain LITT would take place of the other therapeutic options (craniotomy ± gliadel and biopsy) listed under CURRENT TREATMENTS in this patient population. Lastly in this analysis, incremental cost/LYG is likely the more appropriate analysis than incremental cost/quality adjusted life year (QALY) for this condition as; mortality effects are likely to have a more significant impact relative to Quality of Life (QoL analysis); along with the fact that time and resource use are especially constrained in this condition. In these type of circumstances, analysts typically have chosen LYG versus QALY [39].
As it relates to limitations of this analysis the following should be noted: It was not possible to examine KPS as an outcome based on the small number of patients where this was evaluated. However, and as mentioned above, postsurgery KPS is reflective of neurological complications resulting from surgery. Thus in aggregate, KPS would likely have been higher in the brain LITT arm of the decision tree.
Early studies with brain LITT demonstrate a learning curve and as clinicians gain more experience, it appears that the outcomes improve [24,40]. Additionally, the outcomes/data used for "CURRENT TREATMENTS (open resection or biopsy)" are well established. It is possible that as time progresses and brain LITT becomes better established (especially for use in these types of patients) that the overall outcomes would improve for brain LITT.
Complications resulting in repeat surgery from chemotherapy implants (e.g. gliadel wafer); craniotomy, brain LITT and biopsy were not evaluated for cost and their effect on overall survival. These complications resulted in approximately a 3 % repeat surgery rate and are highest in the gliadel wafer group [29,37]. It was assumed based on level II evidence and in the literature reviewed that gliadel wafers were implanted in approximately 10-33 % of patients in the open resection arm of CURRENT TREATMENTS [28,29]. This may not be the case in all situations. If not used in all instances, the overall costs for the open resection arm (of CURRENT TREATMENTS's performed) would be less (however to the detriment of overall survival). It was also assumed in the decision tree model that the application of a chemotherapy implant conferred a 3.3 month increment in overall survival [29].
Fluorescent guided surgery using 5-ALA may not be entirely accurate in identifying complete resection and is not FDA-approved in the United States [41]. Some of the studies where EOR was evaluated for craniotomy used this technology [3,8]. Thus the EOR for craniotomy may have been overestimated-which in turn affected the overall survival numbers in this arm of the decision tree.
Extent of ablation (EOA) from the recent Mohammadi study was used for the brain LITT arm of the trial as a proxy for EOR [24]. In this study the EOA was defined by thermal damage threshold (TDT) lines. Since there is no data in the literature regarding the EOA of thermal damage by brain LITT, it was assumed that the TDT lines defined EOR and, because of this, the 98 % figure was used for EOR in the brain LITT arm of the decision tree model for both PFS and OS. In a prior review of the literature, EOA and EOR, were considered equivalent for PFS [24]. Lastly, since OS had not been followed out long enough for brain LITT, it was also assumed that the 98 % value for EOA with brain LITT assumed a similar OS trajectory as craniotomy.

Conclusions
The use of brain LITT under MRI guidance in complex craniotomies where high-grade gliomas reside in or near areas of eloquence (or where these types of tumors are deep seated) appears to be cost effective-providing value based on it being lower than "value" thresholds established by policy makers. The implications are that brain LITT should be considered a treatment option in these types of high-risk patients.