Auditory brainstem response prior to MRI compared to standalone MRI in the detection of vestibular schwannoma: A modelling study

Abstract Objectives To determine the cost‐effectiveness of auditory brainstem response prior to MRI (ABR‐MRI) compared to standalone MRI to diagnose vestibular schwannoma. Design A state transition model was developed to simulate costs and effects (quality‐adjusted life years [QALY]) for both diagnostic strategies for patients suspected of a vestibular schwannoma. Model input was derived from literature, hospital databases and expert opinions. Scenario and sensitivity analyses addressed model uncertainty. Results Over a lifetime horizon, ABR‐MRI resulted in a limited cost‐saving of €68 or €98 per patient (dependent on MRI sequence) and a health loss of 0.005 QALYs over standalone MRI. ABR‐MRI, however, did miss patients with other important pathology (2% of the population) that would have been detected when using standalone MRI. In total, €14 203 or €19 550 could be saved per lost QALY if ABR‐MRI was used instead of standalone MRI. The results were sensitive to the detection rate of vestibular schwannoma and health‐related quality of life of missed patients. Conclusion The cost‐saving with ABR‐MRI does not seem to outweigh the number of missed patients with VS and other important pathologies that would have been detected when using standalone MRI.


| INTRODUC TI ON
Vestibular schwannoma (VS) is a benign intracranial tumour growing on the myelin-forming cells of the hearing and balance nerves of the inner ear (8th cranial nerve). 1 The overall incidence is about two per 100.000 person-years but is slowly increasing due to the general use of diagnostic imaging techniques such as magnetic resonance imaging (MRI). 1 Patients with symptoms of VS are usually screened with MRI (contrast-enhanced T1 weighted [T1W] and/or T2 weighted [T2W]) of the brain and cerebellopontine angle. 2,3 Consequently, VS is confirmed in approximately 3% of patients with symptoms. 4 Using MRI, other pathologies aside from VS-like cysts, aneurysms, infarctions and malignancies can be diagnosed. However, approximately 84% of the scans do not reveal any pathology. 4 Due to this low incidence of VS, this diagnostic pathway is costly. To reduce the costs, hospitals frequently use auditory brainstem response (ABR) as a preliminary diagnostic method to select high-risk VS patients prior to MRI. [5][6][7] In a survey performed in European ear-nose-throat physicians, nearly 50% indicated that they still use ABR as a preliminary diagnostic method prior to MRI for VS (ABR-MRI). 8 ABR, however, has a lower sensitivity and specificity compared to MRI, especially for the detection of small VSs (<1 cm) and will miss patients with VS and other pathology.
There seems to be a paradigm shift from active treatment to prolonged periods of W &S even when (small) tumour growth is present. This is mainly based on the improved understanding of limited VS growth and small differences in health-related quality of life (HRQoL) of the different treatment strategies. 9 Therefore, the impact of a missed case of VS with ABR is arguably reduced, increasing the likelihood that ABR-MRI is cost-effective compared to standalone MRI. Several studies have been performed to measure the (cost-) effectiveness of ABR as a diagnostic tool.
However, most studies were outdated, 7,10,11 did not include an effect measure, 6,12 or did not include the consequence of a missed case when ABR would be used as a screening tool prior to MRI. Therefore, we aimed to evaluate the cost-effectiveness of ABR screening prior to MRI compared to standalone MRI for the detection of VS.

| Ethical considerations
This modelling study was based on the published literature and did not involve human subjects, and therefore, ethical approval or informed consent was not required.

| Decision-analytic model
The differences between the ABR-MRI strategy and the current standalone MRI strategy were mapped using a decision-analytic

| Overview of the model
We developed the model to determine the costs and effects of both strategies to diagnose patients suspected of VS based on an existing model for VS. 14 Patients were defined as suspected of VS if they reported asymmetrical audiovestibular symptoms including asymmetrical sensorineural hearing loss, tinnitus or vertigo. A hypothetical cohort of 10 000 patients with symptoms of VS went through the model for each strategy (Figure 1 and Appendix S1).
The cohort started at age 50, the mean age of the population. 15 The model started by dividing the patient population according to the disease group: patients with VS, patients with other important pathology (OIP) that require treatment (like aneurysms and intracranial malignancies), and patients (NP) that do not require treatment (like atrophy of the brain or infarctions). 4 Previous studies have shown that T2W MRI sequence is cheaper compared to T1W sequence but will miss some patients due to lower sensitivity. 3 ABR's sensitivity mainly depends on the size of the tumour. 5,6,10 Therefore, both MRI sequences were compared to ABR, and all VS patients were classified according to tumour size using the Koos classification system. 16 The long-term consequences were simulated using a Markov chain simulation. Using cycles to simulate time, patients transitioned between so-called health states. Every cycle of the model simulated

| Transition probabilities
All probabilities of the decision model were based on the literature or, if unavailable, on consensus expert opinion (Table 1).
Parameters from the literature were collected using a detailed search of Pubmed and EMBASE. The consensus expert opinion was constructed by two clinical experts. The patient population was divided according to the disease group using incidences. 4

Key points
• A state transition model was developed to simulate costs and effects for ABR prior to MRI and standalone MRI for patients suspected of a vestibular schwannoma.
• The cost-saving with ABR-MRI does not seem to outweigh the missed patients with vestibular schwannoma and other important pathologies.
• ABR-MRI missed patients with other important pathology that would have been detected when using standalone MRI.
• The cost-saving with ABR-MRI was only worthwhile in scenarios with no negative consequences for missed patients.
• Sensitivity analyses showed that the cost-effectiveness of ABR-MRI is sensitive to the MRI sequence, the probability that a missed patient is detected during follow-up and the quality of life of a missed patient. year after diagnosis (10%) and gradually decreased every year (5%, 2.5%, 1.5%). 1 After four years the annual probability of transitioning to the next Koos stage was assumed to be 0.18%, until 30 years after initial diagnostics, when growth was assumed not to occur. 1 When a patient with VS was missed, we assumed that they had a yearly 10% chance to be detected based on the assumption that the majority of all missed patients with VS will receive additional diagnostics in the upcoming years (after 10 years, 65% is detected).

| Costs
All costs were drawn from cost-effectiveness guidelines, hospital tariffs or by reaching consensus with experts (Table 2). 20,21 To calculate the cost of an MRI T1W sequence, the extra cost of gadolinium was added to the cost of a normal MRI brain, including a margin to cover extra setup and sequence times, estimated at €94. Patients with detected VS in the W&S strategy were annually monitored with MRI until death. All costs were indexed to 2020.

| Effects
To measure the effectiveness of both strategies, the number of missed cases were calculated and the differences in HRQoL between the patient groups using quality-adjusted life years (QALYs).
The QALY is a preferred health outcome including both quality and quantity of life. To calculate a QALY, the HRQoL is transformed into a utility score. A utility score is a numeric value ranging from death (0) to perfect health (1). Utility scores of the different health states were derived from literature and expert opinion (Table 3). HRQoL scores derived from the 36-Item Short Form Survey (SF-36) were converted to utilities. 22 The utility scores of missed VS patients were assumed to be equal to the detected patients, while the disutility of missed patients with OIP were assumed to be 0.01.

| Assumptions
Every cost-effectiveness model is restricted by assumptions in order to be functional and to improve comprehensibility. We assumed that all patients were eligible for both ABR and MRI and accepted the prescribed diagnostic test. In the case of a positive ABR result, we assumed that MRI would subsequently identify it as false-positive,

| Analysis
We compared the number of missed cases, expected costs and effects over a lifetime for the ABR-MRI strategy and the standalone MRI strategy for both MRI sequences. According to guidelines, all costs and effects were discounted with a 4% rate for costs and 1.5% rate for effects. 21 To determine the cost-effectiveness of the two diagnostic strategies, the incremental cost-effectiveness ratio (ICER) was calculated by dividing the incremental cost by the incremental effect.

| Scenario and sensitivity analyses
In the base case analysis, we assumed that the HRQoL of missed VS patients was equal to detected patients. In the scenario analyses, these utilities were varied to include a scenario in which we assumed that the lack of a diagnosis reduces the utility score (−0.05) or improves the utility score (+0.05) compared to detected patients.
To explore the influence of uncertainty on the parameter estimates, sensitivity analyses were performed. For the univariate sensitivity analyses, we varied (1) the probability of detecting a TA B L E 1 Transition probabilities used in the cost-effectiveness model, including the 95% confidence interval missed VS patient after one cycle and (2) the disutility score of a patient with OIP that receives a false-negative diagnosis, to determine the impact of a missed case in need of treatment. We varied the probability of detecting a missed VS patient between 0% and 100% using five intervals, and the disutility score of a patient with OIP was varied from 0.00 to 0.05. The probabilistic sensitivity analysis consisted of 10 000 (Monte Carlo) simulations to reflect the sampling uncertainty, drawn from beta distributions (Tables 1 and 3).

| Scenario analysis
Assuming that all missed patients had a lower HRQoL compared to detected patients resulted in an ICER of €3198 and €2322 per QALY lost for T1W and T2W scenario, respectively, indicating that standalone MRI is cost-effective over ABR-MRI. Assuming that all missed patients had a higher HRQoL compared to detected patients resulted in the domination of ABR-MRI over standalone MRI.

| Univariate sensitivity analysis
The probability of detecting a missed VS was assumed to be 10% per year, but varying this probability did not result in substantial cost savings.

| Probabilistic sensitivity analysis
The scatterplot of the probabilistic sensitivity analysis showed that in nearly all simulations ABR-MRI was not cost-effective because of the limited cost-saving and health loss (Figure 3). At a hypotheti- In conclusion, the cost-saving with ABR prior to MRI does not seem to outweigh the number of missed patients with VS and other important pathologies that would have been detected when using standalone MRI for the diagnosis of patients suspected of a vestibular schwannoma.

ACK N OWLED G EM ENTS
None.

CO N FLI C T S O F I NTE R E S T
None to declare.