Included studies

Five studies (number of participants, N=51) were included in the review: 1 randomised trial (Pozin 2014) and 4 before-and-after observational studies (Koy 2014, Marks 2011, Romito 2015, and Vidailhet 2009).

The clinical studies included in this evidence review are summarised in Table 2.and evidence from these is summarised in the clinical evidence profiles (Table 3 and Table 4).

Pozin 2014 compared levodopa with placebo. The remaining studies (Koy 2014, Marks 2011, Romito 2015, and Vidailhet 2009) compared pre and post-operative outcomes in people receiving bilateral pallidal deep brain stimulation.

See also the literature search strategy in appendix B, study selection flow chart in appendix C, forest plots in appendix E and study evidence tables in appendix D.

Excluded studies

Studies excluded from this systematic review, with reasons for their exclusion, are provided in appendix K.

Summary of clinical studies included in the evidence review are provided in Table 2.

Table 2

Summary of included studies.

See appendix D for the full evidence tables.

Included studies

A systematic review of the economic literature was conducted but no studies were identified which were applicable to this review question.

Excluded studies

No studies were identified which were applicable to this review question.

Results of the economic model

Base case results

When Romito 2015 was used to inform the model, DBS was more costly and more effective than usual care, with an ICER of £20,169 per QALY (Table 8). DBS was also more costly than usual care when Vidailhet 2009 was used, but less effective than Romito 2015. As a result, the ICER was higher at £77,181 per QALY.

Table 8

Base case results (deterministic).

Figure 2

Cost-effectiveness plane (base case).

Sensitivity analysis results

The total QALYs increased for DBS when utility decrements were removed and when the risk of complications were removed. This reduced the ICER for Vidailhet 2009 and Romito 2015, but the ICER for Vidailhet 2009 remained above NICE’s upper threshold of £30,000 per QALY.

Reducing the time horizon reduced the number of QALYs that could be accrued and amplified the cost of the DBS procedure. This analysis increased the ICER above NICE’s upper threshold in both studies.

When usual care consisted of botulinum toxin (a more costly treatment than trihexyphenidyl) the incremental cost reduced. This reduced the ICER for Vidailhet 2009 and Romito 2015, but the ICER for Vidailhet 2009 remained above NICE’s upper threshold of £30,000 per QALY.

The results of each analysis are provided in Table 9 for Romito 2015 and Table 10 for Vidailhet 2009.

Table 9

Results of sensitivity analysis (Romito 2015).

Table 10

Results of sensitivity analysis (Vidailhet 2009).

Using Romito 2015, the worst case scenario, raising the cost of the total procedure by 50%, increased the ICER to £23,918. In the best case scenario, lowering the total cost of the procedure by 50% reduced the ICER to £16,420. The most influential parameters were related to the replacement of the IPG. When the cost to replace the IPG was varied by 50% the ICER ranged from £16,342 to £23,750. When the frequency of replacements was changed from every 5 years to every 2 or 8 years, the ICER ranged from £16,761 to £33,351. (Figure 3, Figure 4) For Vidailhet 2009 all ICERs remained above a cost-effectiveness threshold of £30,000 per QALY when parameters were varied (Figure 5, Figure 6). Similarly to Romito 2015, the most influential parameters included the total cost of the procedure (namely stimulation equipment) and IPG replacements.

Figure 3

Tornado diagram of the costs associated with the procedure and monitoring (Romito 2015).

Figure 4

Tornado diagram of the costs to treat the complications of DBS (Romito 2015)).

Figure 5

Tornado diagram of the costs associated with the procedure and monitoring (Viadilhet 2009).

Figure 6

Tornado diagram of the costs to treat the complications of DBS (Vidailhet 2009).

Probabilistic results

For Romito 2015, all simulations found DBS to be more effective and more expensive than usual care with a mean probabilistic ICER of £20,077. Furthermore, 739 of 1,000 simulations had ICER’s below £20,000 and 927 below £30,000. This is illustrated in Figure 7Figure 3 where simulations cross the £20,000 threshold in the north-east quadrant. The cost effectiveness acceptability curve (CEAC) also illustrated that DBS would be considered as the most optimal treatment for any threshold over £17,000 per QALY (Figure 8). In Figure 7Figure 3 the simulations do not fall below an incremental cost of £60,000 the cost to provide DBS.

Figure 7

PSA simulations (Romito 2015).

Figure 8

CEAC (Romito 2015).

When Vidailhet 2009 was used to inform the model, the mean ICER was £72,323 with almost all simulations (996 of 1,000) in the north-east quadrant above NICE’s threshold (Figure 9). The CEAC also illustrated that usual care would be considered as the most optimal treatment for thresholds up to £65,000 per QALY (Figure 8).

Figure 9

PSA simulations (Vidailhet 2009).

Conclusions

DBS is more effective but also more costly than usual care when either study is used to inform the model. When the ICER is considered, DBS could be considered cost effective according to Romito 2015, who produce an ICER just above NICE’s advisory threshold of £20,000 per QALY in the base case with many iterations of the PSA being below it. The opposite was the case when Vidailhet 2009 was used to inform the model with both the base case and nearly all iterations of the PSA, DBS was above NICE’s conventionally held threshold of £20,000 per QALY and therefore not considered cost effective.

Given the large uncertainty inherent in the clinical evidence it is difficult to make strong conclusions. Greater certainty around cost effectiveness would be obtained through further research. Given the evidence around cost effectiveness DBS should only be considered for use when all other medical and surgical interventions have been considered and exhausted – in line with the current NHS commissioning policy on DBS for dystonia.

Pharmacological treatments

According to NHS Reference Costs 2015/16 the first attendance for a pre-assessment in neurology would cost £217 (currency code WF01B, service code 400, non-admitted face-to-face attendance, first, neurology), but the committee advised that pharmacological treatments for dystonia could be initiated by a specialist clinic neurologist, rehabilitation medicine consultant, specialist nurse or specialist prescribing physiotherapist.

Drug acquisition costs for all pharmacological interventions for which evidence was searched, were taken from the NHS Electronic Drug Tariff May 2017 and dosages from the BNF August 2017. (Not presented) Dosages were verified with the committee to ensure they were appropriate for this patient group.

Often, oral treatments for dystonia do not incur administration costs as they are administered at home, without health care professional assistance. However, if families or carers administer oral treatments via PEG, they will require additional training and equipment. Oral treatments may be monitored by the patient’s GP and community team at routine visits, but advice from a rehabilitation medicine or neurologist on increasing or decreasing medication would be sought if they were not directly responsible for monitoring the treatment. Furthermore, for levodopa, an additional review with the patient’s GP or neurologist after the initial 3 months of treatment would be incurred to assess efficacy.

Botulinum toxin involves a day–case admission performed by a neurologist, rehabilitation medicine doctor, or a specially trained physiotherapist or nurse in a specialist clinic. Adults with cerebral palsy are unlikely to be sedated, but ultrasound or electromyography (EMG) may be used for guidance.

The appointment for the injection of botulinum has a NHS reference cost assigned – Torsion dystonia and other involuntary movements drugs band 1 (code XD09Z). This reference cost (£324) will include all costs related to the procedure, the day case admission, drug costs and staff costs.

Following the injections, patients would be monitored every 3 to 4 months by the specialist clinic at a cost of £161 (NHS Reference Costs 2015/16, currency code WF01A, service code 400, non-admitted face-to-face attendance follow-up, neurology) to assess their response and need for repeat injections.

Dynamic orthotics

Healthcare Improvement Scotland identified no published cost-effectiveness evidence on dynamic lycra splinting. For completeness they provided the cost of body suits currently available from personal communications. Those costs are presented alongside costs converted to 2015/16 using the hospital and community health services pay and prices index uplift (Curtis 2015) in Table 11.

Table 11

Cost of dynamic orthotic equipment.

Dynamic orthotic equipment would be offered after an assessment with an orthotist (NHS Reference Costs 2015/16 WF01B 658, £77) following a referral from an occupational therapist or physiotherapist. Orthotic equipment should be reviewed annually by an occupational therapist (regarding upper limb and hand orthotics), physiotherapist (regarding body suits or legs and feet orthotics) or orthotist. If there is a ‘change’ or ‘problem’ 2 or 3 of those healthcare professional may complete a joint review (Table 12).

Table 12

Follow-up costs, orthotics.

The committee advised that orthotic equipment would typically last between 6 to 24 months before it needs to be replaced, but reiterated that the lifespan would depend on how much it is used and during which activities.

Intrathecal baclofen (ITB)

Sampson 2002 published a study on ITB in which detailed cost estimates were derived from 3 centres in the UK where the procedure was being performed. The costs included in the study were obtained in 1999 and have been converted to 2015/16 costs using the hospital and community health services pay and prices index uplift (Curtis 2015) in Table 13.

Table 13

Cost of intrathecal baclofen reproduced from Sampson 2002.

The East Midlands Specialised Commissioning Group also produced detailed paediatric and adult costs for ITB treatment in 2009. They assumed the admission for the test dose usually takes 2 days whilst the admission for the implant usually takes an additional 5 days. The test dose, implant and refills were worked out using the contract code AB05Z (for intermediate pain procedures), at 2009/2010 prices. Those prices are presented alongside 2015/16 costs in Table 14.

Table 14

Cost of ITB treatment based on East Midlands commissioning policy 2009.

The total costs over 5 years are similar in the Sampson 2002 study and in the East Midlands Commissioning Policy; however, it is likely that the costs from the latter source are more accurate as costs were based on an HRG code, reflecting more recent UK practice.

Comparison 1. Levodopa versus placebo

Comparison 2. bilateral pallidal deep brain stimulation (DBS) – pre versus post-operative

Interpreting the evidence

Cost effectiveness and resource use

The committee noted that no relevant published economic evaluations had been identified for this topic.

Dystonia is aggravated by factors such as pain and anxiety which if not identified and managed appropriately, can negatively impact on the patients’ health-related quality of life. Therefore, knowing what factors can aggravate dystonia may lead to increased vigilance and thus more timely management with potential cost savings. Estimating the costs to manage those factors would go beyond the scope of the guideline although they were likely to offset the cost of the recommendations.

The committee discussed the evidence that levodopa provided no additional benefit compared to placebo and agreed that relatively cheap treatments should not be recommended if they are ineffective or have the potential to incur adverse effects. For this reason, the committee made a recommendation to not routinely prescribe levodopa to stop potentially cost-ineffective practices. However, other pharmacological treatments such as gabapentin and trihexyphendyl are currently available in practice and could be considered before more costly and more invasive options. Given that no evidence was identified on those alternatives, the committee made a recommendation to refer adults with problematic dystonia to a specialist movement disorder or spasticity service to consider treatment in line with their experience and expertise. The committee noted that a recommendation to refer adults to a specialist tone management team would not increase current resource use as it would be beyond the remit of GPs to initiate treatments for dystonia in primary care. The committee also noted it would be cost-ineffective to refer people with asymptomatic or tolerable dystonia as treatment would not provide any additional benefits to justify the cost, burden, or potential adverse effects of treatment.

The committee noted that no one should remain on cheap, ineffective treatments as the burden of treatment and long-term cost, including the cost to manage their adverse events could be substantial. However, treatments for dystonia should be discontinued gradually, to minimise withdrawal symptoms such as anxiety and distress, as those symptoms would offset the cost of immediate discontinuation.

Some centres would consider splinting (including dynamic Lycra) following an assessment with occupational therapy, before more invasive treatments such as botulinum toxin or intrathecal baclofen are considered. However, the committee acknowledged the high cost to provide orthotic equipment and agreed there was no clinical evidence it provided a cost effective use of resources to implement its wider use. For this reason, the committee made a research recommendation to assess the clinical and cost effectiveness of splinting options.

The committee stated botulinum toxin was frequently used in current practice to manage focal dystonia when it is causing discomfort or affecting care or function and cannot be managed effectively using cheaper, less invasive treatments. The committee agreed it was important to state those criteria in their recommendation to prevent practises which were unlikely to be cost effective. However, in the absence of comparative high quality evidence, the committee did not make a strong recommendation.

The committee agreed intrathecal baclofen therapy was an expensive but successful option provided by specialists to manage dystonia when other options have been exhausted. The committee also added that additional research on its effectiveness in a population of adults with cerebral palsy would not change current practice unless it was shown to be harmful as the benefits have been shown to outweigh the costs in other populations. In the absence of published evidence on intrathecal baclofen therapy in adults with cerebral palsy, the committee made a recommendation based on their clinical experience and expertise which they considered to reinforce best practice.

Deep brain stimulation (DBS) is a relatively new and expensive treatment used to manage dystonia in England and Wales and is commissioned in the NHS for patients with generalised dystonia, status dystonicus, laryngeal dystonia and cervical dystonia if the criteria set out in the commissioning policy are met. However, the cost effectiveness of DBS had not been assessed in adults with cerebral palsy when that policy was produced.

Two studies included in the clinical evidence review provided SF-36 data, before and after deep brain stimulation treatment which enabled a cost utility analysis to be developed. The Committee stated that it was crucial the complications of deep brain stimulation were taken into consideration when making their recommendations, as they may outweigh the benefits deep brain stimulation can provide. As a result, the economic modelling was used by the Committee as one of many ways to assess those trade-offs.

The results of the model were sensitive to the study used to inform the improvement in utility as one study included participants with a much lower utility pre-DBS who saw a much greater improvement in their utility post-DBS. The committee agreed that DBS would only be considered when all pharmacological treatments had failed and for this reason placed more weight on the study who included participants with a lower utility value pre-DBS. The committee however noted that the patient group in Vidailhet 2009, for which DBS was not cost effective, was of people with dystonia which had become unresponsive to pharmacological treatment. Whilst this more closely reflected the patient group in the recommendation, Romito 2015 (individuals with persistent dystonia) was more recent better reflecting current technology and was a larger study with longer follow-up. The larger increases in utility values from Romito 2015 also better reflected what the committee experienced in clinical practise.

Uncertain parameters were varied in deterministic one-way sensitivity analyses to assess the robustness of the results. Those parameters included the cost of the procedure, inclusion and consequences of complications and frequency of battery replacements. The results of those analyses provided ICERs below or within NICE’s advisory threshold for cost-effectiveness when parameters set out in Romito 2015 were used to inform the model, providing evidence that DBS could be a cost-effective option. Those results were also reiterated in probabilistic analysis with 739 of 1,000 simulations below an ICER of £20,000.

Based on the economic evidence and their clinical expertise, the committee agreed that deep brain stimulation should be recommended in line with the current NHS commissioning policy as a cost effective option. As a result, the committee made a recommendation to consider a referral to a specialised tone management service with experience in providing deep brain stimulation for adults with intractable dystonia that is severe and painful.

Other factors the committee took into account

The committee also took into account the recommendations made in the NICE interventional procedure guideline Deep brain stimulation for tremor and dystonia (excluding Parkinson’s disease) IPG188 (2006). It is recommended in IPG188 that deep brain stimulation can be used provided that the normal arrangements are in place for consent, audit and clinical governance and only in the context of a multidisciplinary team specialising in the long-term care of patients with movement disorders. The committee therefore believed that the recommendation that they made aligns with IPG188 and made a cross-reference to it.

Probability of DBS-related complications

DBS-related complications were included in the model as they can have important cost and QALY implications. The trials included in the clinical evidence review were small and unrepresentative of the adverse effects seen in practice, so alternative papers that analysed DBS were sought to inform the probability of complications in the model.

Boviatsis 2010 and Voges 2006 reviewed the complications of DBS experienced by their departments; from 2003 to 2010 in 106 patients and from 1996 to 2003 in 262 patients, respectively. Both also compared their own results to others reported in the literature.

Both of those papers considered ICHs to be important and serious adverse events associated with DBS, reporting probabilities in their own departments of 1.9% (2 of 106 patients) and 0.4% (1 of 262 patients) and higher results in the literature they reviewed (Beric 20013.3%; Kondziolka 2002, 1.5%; Oh 2002, 3.6%; Umemura 2003, 3.6%; Limousin 1999, 2.7%; Lyons 2004, 1.2%). However, details on the event, such as the severity, were not reported. As a result, the committee sought the paper by Binder 2005 who examined symptomatic and asymptomatic haemorrhages across all 280 DBS procedures performed for movement disorders between June 1998 and May 2004.

Skin infection may be caused by both DBS surgery and implanted hardware components. Consequently, the definition of infection in the literature was not unanimous; some restricted the definition to hardware-involving infections with positive only cultures, whereas others also included superficial infections over the implanted hardware. As a result, the model considered early infections in the first cycle, later infections in the second cycle and antibiotic treatment to remain on DBS, or removal of the system.

Table 23 below presents the probability of perioperative DBS-related complications used in the model.

Table 23. Probability of perioperative DBS-related complications

According to the committee, hardware-related failures can occur at any time during or after the procedure. Bovistis 2010 defined hardware failures as an electrode breakage, lead or extension fracture or migration or misplacement and found those to be experienced by 4 of 106 patients (3.8%) in their department. Voges 2006 reviewed the literature and found lead fractures to range from 1.7% (Voges 2006) to 15.2% (Kondziolka 2002), lead migrations from 1.5% (Kondziolka 2002) to 6.3% (Lyons 2004) and extension wires from 1.1% (Voges 2006) to 3.5% (Beric 2001). However, they also found zero cases reported for each type of hardware failure. In their own study, Voges 2006 reported hardware-related problems in 25 of 180 (13.9%) of patients during their long-term observation. In the model an annual probability of 4.0% was used to reflect a weighted average of those papers. The methods and data used to obtain this value is provided in Table 24.

Table 24. Probability of hardware-related complications

Health-related quality of life

The QALY is NICE’s preferred measure of benefit for economic evaluation. This is because it can be seen as a generic measure of health which allows a comparison across treatments which affect different dimensions of health.

The QALY reflects the 2 principle objectives of health care:

• increase longevity;
• increase quality of life.

Estimating a QALY involves placing a quality of life weight on a particular health state. This quality weight lies between 0 and 1, where 1 denotes full or ‘perfect health’ and 0 denotes death. Based on a need for consistency across appraisals and guidelines, NICE favours the EQ-5D to value health states - a generic, preference based measure which comes with preexisting utility values obtained from a representative sample of the UK general population, although others measures and value sets are available.

Clinical effectiveness data (specifically health-related quality of life data) was taken from 2 before and after type studies (Vidailhet 2009 and Romito 2015) that reported the results for each of the 8 domains of the SF-36, pre- and post- DBS treatment. To allow for subsequent use in the health economic analyses, the SF-36 was mapped on to the EQ-5D using the mapping regression coefficients produced by Ara and Brazier 2008 (Table 25).

Table 25. EQ-5D regression coefficients

Table 26. Vidailhet 2009

Table 27. Romito 2015

Given that no comparative data was identified, it was assumed the utility pre-DBS is equivalent to the utility associated with “usual care”. It was also assumed that the utility post-DBS holds when patients remain on DBS care.

When Romito 2015 was chosen to inform the model, the values 1-year and 2-years post-DBS were applied in the first and second year, whilst the value associated with the last visit was carried to a lifetime horizon.

There is a clear difference in pre-treatment utility between Vidailhet 2009 and Romito 2015, with participants in Romito 2015 entering the study with a much lower quality of life (0.35 vs. 0.61) (Table 27) most likely due to Romito 2015 only including patients with acquired dystonia (who may be less accustomed or have adapted to their condition) compared to Vidailhet which only included patients with idiopathic or inherited dystonia. For example, Vidailhet 2009 required optimum pharmacological treatments to be ineffective, whereas Romito 2015 did not specify this. In addition, it has been shown that mapping functions tend to overestimate utilities associated with severe health states and underestimate utilities associated with good health (Rowen 2009). For these reasons, the studies were not pooled and used separately in the model. However, it is evident that if DBS is not found to be cost-effective when Romito 2015 is used to inform the model, it will not be cost-effective according to Vidailhet 2009 who provides lower incremental QALY gains post- vs. pre- DBS.

People who undergo DBS experience some level of disutility due to the length and intensity of their inpatient stay, as DBS is an invasive and complex procedure. Despite this, no utility values in relation to the procedure were identified from the literature. Instead, the disutility was imputed using the EQ-5D health state valuation equation for the UK reported by Dolan 1997 which allows estimation of a person’s utility based on their responses to the EQ-5D classification system. The system has 5 dimensions (mobility, self-care, ability to perform usual activities, pain/discomfort, and anxiety/depression) and in the version used by Dolan 1997, each dimension had 3 levels of response (no problems, moderate problems, and severe problems).

Only the utility decrement due to usual activities was applied as this was considered to be the most dependable dimension on the neurosurgical procedure. This disutility is expressed by the following equation:

Where:

• α = 0.081 (constant applied to any level of disutility in any of the 5 EQ-5D dimensions)
• UA = −0.036 (for each level of disutility associated with usual activities)
• U2 = −0.022 (for being unable to perform usual activities)
• N3 = −0.269 (when any of the 5 dimensions of EQ-5D is severe)

As the baseline utility for people with cerebral palsy in the model is less than 1 (perfect health) for both Romito 2015 (0.35) and Vidailheit 2009 (0.61) the α value was not applied at the estimation of the utility decrement. and they moved from a state of moderate problems to being unable to perform them. Also assuming that at least one other dimension was severe, the N3 value is not added again, resulting in a disutility of −0.094 (−0.036–0.036–0.022).

To reflect the length of the procedure, the disutility was applied for 2 weeks in the model - a QALY loss of −0.004 (−0.094*(2/52)).

Hardware-related failures also require surgery to correct. For this reason, a 1 week QALY loss −0.002 (−0.094*(1/52)) was applied to patients receiving surgical treatment to correct hardware-related failures.

Cerebral palsy-related

The committee considered Brook 2014 to provide up-to-date survival estimates for people with cerebral palsy living in California that would be generalisable to adults living in England and Wales.

Brook 2014 reported survival estimates for 5 levels of severity which enabled the model to select those levels appropriate for people with dyskinetic cerebral palsy. To select the appropriate levels, the committee agreed it would be reasonable to assume that GMFCS is stable and can be informed by paediatric data that assess GMFCS and learning disability in older children with dyskinetic cerebral palsy. Himmelmann 2007 described 48 participants with dyskinetic cerebral palsy in Western Sweden. Their gross motor function was classified according to the GMFCS and was subsequently transformed into the limitations by Brook 2014 to create a weighted average (Table 29).

Table 29. GMFCS levels used to inform dystonic limitations

The probability a child with cerebral palsy will survive is reported up to the ages of 10, 15, 20, 25, and 30 years by Brook 2014. However, given that people with cerebral palsy are expected to live up to 70 years, the data beyond 30 years was extrapolated using a polynomial trend (Figure 17).

Figure 17. Survival

The committee agreed there was no evidence to suggest DBS treatment impacts survival following the procedure; hence, the same trend was applied to both treatment arms in the model.

DBS procedure

DBS is a risky and invasive procedure. The committee agreed that procedure-related mortalities reported in the literature were low, but concluded procedure-related mortality was an important possibility to capture in the model.

Boviastis 2010 reported a perioperative mortality of 0.94% in their study, whilst the literature review by Voges 2006 identified 1 study that reported mortality (Umemura 2003, 1.8%) with a similar rate. However, the remaining studies reviewed by Voges 2006 did not report mortality.

Major ICH

Gonzalez 2013 investigated short-term case fatality and long-term mortality after ICH using data from The Health Improvement Network (THIN) database over the years 2000 to 2008. A total of 1,733 individuals with an ICH and 9,583 controls were available with follow-up data.

Using logistic regression, event fatalities were stratified by age. For people aged 20 to 49 years Gonzalez 2013 estimated a 30-day case fatality of 29.7% for an ICH.

Cox proportional hazards regression analyses were used to determine whether patients were at increased risk of death in the first year (excluding the first 30 days immediately after the event) and after 1 year compared with the general population (controls) in THIN.

They found that the risk of death was significantly higher among stroke patients during the first year of follow-up compared with controls (HR 2.60, 95% CI 2.09–3.24) and remained elevated among survivors at 1 year (HR 2.02, 95% CI 1.75–2.32).

Deep brain stimulation (DBS)

Yianni 2005 provided a detailed cost-analysis of DBS surgery, including the preoperative assessment, surgery, equipment, postoperative management/follow-up and complications when they estimated cost-effectiveness. Costs were examined over a period of 2 years on 26 patients with primary dystonia. The effectiveness of DBS between primary dystonia and dystonic cerebral palsy will differ; hence, their estimate of cost-effectiveness was not considered to be relevant for this guideline.

However, the Committee agreed that the resources reported by Yianni 2005 would be very similar to those for dystonic cerebral palsy. Those costs are reproduced here in 2002/3 prices and 2015/16 prices using the hospital and community health services pay and prices index uplift (Curtis 2015) (Table 30).

Table 30. Cost of DBS reproduced from Yianni 2005

The committee suspected that Yianni 2005 may not reflect the latest innovations in equipment, particularly with regards to the type of rechargeable battery now available. To account for this uncertainty, a tornado diagram was presented, varying the cost inputs by +/−50%.

DBS-related complications

Usual care

In the base case, patients received trihexyphenidyl (5mg daily) to align with the type of pre-treatment participants received in Vidailhet 2009 and Romito 2015 and the type of pharmacological treatment used in clinical practice today (Table 31). Patients receiving trihexyphenidyl visit a neurologist each year to monitor their response at a cost of £161 (NHS Reference Costs 2015/16, currency code WF01A, service code 400, non-admitted face-to-face attendance follow-up, neurology).

Table 31. Cost of usual care (trihexyphenidyl)

A sensitivity analysis administering botulinum toxin every 6 months was also explored as a sensitivity analysis. Botulinum toxin involves a day–case admission performed by a neurologist, rehabilitation medicine doctor, or a specially trained physiotherapist or nurse in a specialist clinic. Adults with cerebral palsy are unlikely to be sedated, but ultrasound or electromyography may be used for guidance. However, given that recommendations were not included on ultrasound or electromyography guidance, they are not added here for consistency.

The appointment for the injection of botulinum has a NHS reference cost assigned – Torsion dystonia and other involuntary movements drugs band 1 (code XD09Z). This reference cost (£324) will include all costs related to the procedure, the day case admission, drug costs and staff costs.

Following the injections, patients would be monitored every 3 to 4 months by the specialist clinic at a cost of £161 (NHS Reference Costs 2015/16, currency code WF01A, service code 400, non-admitted face-to-face attendance follow-up, neurology) to assess their response and need for repeat injections.

Base case results

When Romito 2015 was used to inform the model, DBS was more costly and more effective than usual care, with an ICER on NICE’s lower threshold (Table 34). This is illustrated in Figure 18 with an ICER in the north-east quadrant.

DBS was also more costly than usual care according to Vidailhet 2009, but relatively less effective than Romito 2015. As a result, the ICER was higher when Vidailhet 2009 was used to inform the model.

It is important to note that Vidailhet 2009 produced more QALYs than Romito 2015 because Vidailhet 2009 reported higher utility values pre- and post-DBS treatment (Table 34).

However, as stated above, the range is greater for Romito 2015. The total costs do not differ between the studies as they only differ in utility values.

Table 34. Base case results (deterministic)

Figure 18. Cost-effectiveness plane (base case)

Sensitivity analysis results

The total QALYs increased for DBS when utility decrements were removed and when the risk of complications were removed. This reduced the ICER for Vidailhet 2009 and Romito 2015, but the ICER for Vidailhet 2009 remained above NICE’s upper threshold of £30,000 per QALY.

Reducing the time horizon reduced the number of QALYs that could be accrued and amplified the cost of the DBS procedure. This analysis increased the ICER above NICE’s upper threshold in both studies.

When usual care consisted of botulinum toxin (a more costly treatment than trihexyphenidyl) the incremental cost reduced. This reduced the ICER for Vidailhet 2009 and Romito 2015, but the ICER for Vidailhet 2009 remained above NICE’s upper threshold of £30,000 per QALY.

The results of each analysis are provided in Table 35 for Romito 2015 and Table 36 for Vidailhet 2009.

Table 35. Results of sensitivity analysis (Romito 2015)

Table 36. Results of sensitivity analysis (Vidailhet 2009)

Figure 19 illustrates the ICERs for Romito 2015 when each component of the procedure was varied using half the cost of the base case (−50%) and 150% of the base case (+50%). In the worst case scenario, increasing the cost of the total procedure by 50% increased the ICER to £23,918. In the best case scenario, reducing the total cost of the procedure by 50% reduced the ICER to £16,420.

Figure 20 also used this method to show the variability in ICERs to treat the complications of DBS for Romito 2015. The most influential parameters were related to the replacement of the IPG. When the cost to replace the IPG was varied by 50% the ICER ranged from £16,456 to £23,873. When the frequency of replacements was changed from every 5 years to every 2 or 8 years, the ICER ranged from £16,884 to £33,474.

Figure 19. Tornado diagram of the costs associated with the procedure and monitoring (Romito 2015)

Figure 20. Tornado diagram of the costs to treat the complications of DBS (Romito 2015)

Figure 21 illustrates the ICERs for Vidailhet 2009 when each component of the procedure was varied using half the cost of the base case (−50%) and 150% of the base case (+50%). Figure 22 also used this method to show the variability in ICERs to treat the complications of DBS for Vidailhet 2009. All ICERs remained above NICE’s upper threshold when those parameters were varied. Similarly to Romito 2015, the most influential parameters included the total cost of the procedure (namely stimulation equipment) and IPG replacements.

Figure 21. Tornado diagram of the costs associated with the procedure and monitoring (Viadilhet 2009)

Figure 22. Tornado diagram of the costs to treat the complications of DBS (Vidailhet 2009)

Probabilistic results

For Romito 2015, all simulations found DBS to be more effective and more expensive than usual care with a mean probabilistic ICER of £20,077. Furthermore, 739 of 1,000 simulations had ICER’s below £20,000 and 927 below £30,000. This is illustrated in Figure 23 where simulations cross the WTP threshold in the north-east quadrant. The simulations do not fall below an incremental cost of £60,000 as the cost inputs were not sampled. For this reason, the incremental cost cannot fall below the cost to provide DBS.

The cost-effectiveness acceptability curve (CEAC) also illustrated that DBS would be considered as the most optimal treatment for WTP thresholds over £17,000 (Figure 24).

Figure 23. PSA simulations (Romito 2015)

Figure 24. CEAC (Romito 2015)

When Vidailhet 2009 was used to inform the model, the mean ICER was £72,323 with almost all simulations (996 of 1,000) in the north-east quadrant above NICE’s threshold (Figure 25). The CEAC also illustrated that usual care would be considered as the most optimal treatment for WTP thresholds up to £65,000 per QALY (Figure 26).

Figure 25. PSA simulations (Vidailhet 2009)

Figure 26. CEAC (Botulinum toxin)