Spinal cord stimulation for chronic pain of neuropathic or ischaemic origin: systematic review and economic evaluation

Intervention

• spinal cord stimulator devices.

This included spinal cord stimulators with implantable pulse generator systems (non-rechargeable and rechargeable) and spinal cord stimulators with radiofrequency receiver systems.

Population

• adults with chronic neuropathic or ischaemic pain who have had an inadequate response to medical or surgical treatment (appropriate to condition) other than SCS.

Comparator

• medical and/or surgical treatment (appropriate to condition) that does not include SCS.

Outcomes

• pain

• health-related quality of life

• physical and functional abilities

• anxiety and depression

• medication use

• complications and adverse effects (e.g. procedural complications and technical failures).

Pain outcomes

Both trials used VAS to measure pain. In the PROCESS trial, leg pain was reduced by 50% or more in significantly more patients in the SCS group than in the CMM group at 6 months (p < 0.001) and 12 months (p = 0.005). A similar outcome in the North trial, patient satisfaction plus 50% or more pain relief, was achieved by significantly more patients in the SCS group than in the reoperation group (p = 0.04). Patient satisfaction was also assessed in the PROCESS trial, with significantly more SCS (66%) than CMM (18%) patients satisfied with pain relief at 6 months (p < 0.001). Table 3 shows ITT/worst-case analyses. The PROCESS trial per treatment analysis at 12 months also showed a significant difference between groups (p = 0.03), as did the North trial analysis of patients available for long-term follow-up (p = 0.01).

Patient self-reported pain related to daily activities did not differ between the SCS and reoperation groups (North).

Medication outcomes

As shown in Table 4, there was no difference between the SCS and CMM groups in opioid use, morphine equivalent dose or non-steroidal antiinflammatories, or antidepressants (borderline significance p = 0.06; PROCESS). Significantly fewer SCS than CMM patients were taking anticonvulsants at 6 months (p = 0.02) (because of a change in CMM group; PROCESS). The reoperation group required an increase in opiate analgesics significantly more often than the SCS group (p = 0.025; North), which may indicate that the difference between groups in pain as measured by the VAS in this trial could have been more pronounced if analgesic use had remained at baseline values.

Functional outcomes

Functional ability at 6 months, as measured by the Oswestry Disability Index (Table 5), improved significantly from baseline for the SCS group (p < 0.001), but not for the CMM group, with the difference between groups being significant (p < 0.001) (PROCESS).

Both trials reported no difference between groups in employment status.

Patient self-report of neurological function (lower extremity strength and co-ordination, sensation, bladder/bowel function) did not differ between SCS and reoperation groups (North).

North reported that patients randomised to reoperation were more likely to cross over to SCS (n = 14 out of 26) than vice versa (n = 5 out of 24) (p = 0.02). The authors noted that not all patients whose treatment was classified as not successful opt to cross over. In the PROCESS study five SCS and 32 CMM patients requested crossover.

HRQoL outcomes

The PROCESS trial assessed HRQoL using Short Form 36 (SF-36; Table 6). At 6 months, the SCS group improved significantly in seven out of eight domains measured but not in the domain ‘Role–emotional’, whereas the control group only showed significant improvement in the domain ‘General health’. There was a significant difference between groups in seven out of eight domains p ≤ 0.02, but not in ‘Role–physical’.

Summary

Evidence from FBSS trials suggested that SCS was more successful than CMM or reoperation in terms of pain relief. SCS resulted in more reduction in use of opiates than reoperation. SCS was more effective than CMM in improving functional ability and HRQoL.

There was no difference between SCS and reoperation in pain related to daily activities or neurological function. Medication use was similar for SCS and CMM groups. Employment status was not improved by SCS, CMM or reoperation.

Pain outcomes

The Kemler trial (Table 7) reported that the SCS group showed significantly more reduction in pain as measured by VAS than the PT group at 6 months (p < 0.001) and 2 years (p = 0.001) but not at 5 years (p = 0.25). The change in significance was partly the result of a lower pain reduction in the SCS group and partly of a reduction in pain in the PT group at longer follow-up.

The Kemler trial also measured Global Perceived Effect (GPE), a seven-point scale, finding thatsignificantly more SCS patients than PT patients considered that they were ‘much improved’ at 6 months (p = 0.01), and at 2 years (p = 0.001). This difference was also significant in a per treatment analysis at 6 months and 2 years (p < 0.001). A review by Grabow et al. 50 calculated the number needed to treat to obtain at least one patient with a GPE rating of ‘much improved’ as 3.0 (95% CI 1.9–7.0), which was comparable to that for medications for chronic pain. 50 When the Kemler trial measured ‘success’ as either ‘much improved’ on GPE or a 50% or more decrease in pain measured by VAS, 20 of 35 SCS patients achieved success at 2 years. 67

Functional outcomes

Functional outcome was measured using the Jebsen test for hand function and a standardised test devised by the authors for foot function, testing speed to perform tasks (Table 8), strength and function (Appendix 5.2). There was no clinically important improvement in function in either of the treatment groups at 6 months or 2 years. Apart from ankle range of motion reaching borderline significance (p = 0.04) favouring the PT group at 2 years (based on n = 5 in control group), none of the function tests differed between groups at 6 months or 2 years.

HRQoL outcomes

The HRQoL outcome measures cited by Kemler were Nottingham Health Profile, EuroQol 5D, the short version of the Sickness Impact Profile and Self-rating Depression Scale. There were no differences in HRQoL between groups in any ITT analysis (Table 9). Per treatment analyses at 6 months and 24 months suggested that the SCS group (n = 24) had significantly more improvement than the PT group as measured on the pain component of the Nottingham Health Profile, for patients with either an affected hand (p = 0.02) or an affected foot (p = 0.008).

Summary

Evidence from the CRPS trial suggests that SCS was more effective than PT in reducing pain at 6 months and 2 years, but not at 5 years, and was more successful in terms of patients’ GPE of treatment.

SCS and PT were similar in effectiveness for functional ability of the affected hand or foot, and for HRQoL.

Pain outcomes

Two of the four included trials reported pain outcomes. The ESES trial (Table 10) measured pain on VAS at 1, 6, 12 and 18 months and found no difference between SCS and CMM groups. ESES also found that the pain-rating index of the McGill Pain Questionnaire showed that for both the SCS and CMM groups pain was decreased significantly at 1 month and 3 months (p < 0.001), remaining stable up to 18 months, with no difference between groups. In the Jivegard trial the SCS group had significant long-term pain relief throughout 18 months of follow-up (p < 0.01), and the analgesics group had significant pain relief at 2 months follow-up (p < 0.05), but no significant pain relief at 6 months or 12 months follow-up. Skin temperature in the ischaemic area, measured by VAS, did not differ between the SCS and analgesics groups and neither group differed significantly from baseline (Jivegard).

When considering only non-amputated patients, ESES reported more pain relief in the SCS group than the CMM group, whereas in the case of amputation pain relief slightly favoured CMM.

Medication outcomes

The ESES found a reduction in the numbers of patients taking narcotics in both SCS and CMM groups (Table 11). In a different measure of medication use, ESES used a Medication Quantification Scale to evaluate the use of analgesics, and found a significant difference between groups at 1 month and 3 months (p < 0.001) and 6 months (p = 0.002), with SCS on a lower dose than CMM. This difference was borderline significant at 12 months (p = 0.055) and non-significant at 18 months (p = 0.70). The direct pain measurement outcomes of this trial showed no difference between groups, but the lower medication use in the SCS group up to 6 months may have affected the pain measures.

Functional outcomes

All four trials reported limb survival or amputation rates (Table 12), and none of the trials found a significant difference between SCS and control groups. The Jivegard trial reported a borderline significant difference between groups when categorising amputations by none/moderate/major, with fewer major amputations in the SCS than in the analgesics group.

Despite differences in trial comparators, two meta-analyses have been published. A meta-analysis by Klomp et al. 99 including the studies ESES, ESES pilot, Suy, Jivegard and Claeys, produced a non-significant relative risk of amputation at 18 months of 0.80 (95% CI 0.60 to 1.06) [risk difference – 0.07 (95% CI – 0.17 to 0.03) for SCS with reference to control]. The systematic review by Ubbink et al. 87 included a non-randomised trial (Amann101) in a meta-analysis of limb salvage at 12 months which indicated significantly greater limb salvage with SCS compared with control; however, by excluding the non-randomised trial a non-significant difference was found between SCS and control RR 0.78 (95% CI 0.58 to 1.04), risk difference 0.09 (95%CI – 0.01 to 0.19).

The systolic toe to brachial pressure index did not differ between the SCS and analgesics groups in the Jivegard trial, with values for both groups significantly increased from baseline at 2 months but not at 6 months. Jivegard found no difference between SCS and analgesics groups in the ankle to brachial pressure index, with neither group differing from baseline. For the ankle to brachial pressure index Claeys found that the mean change for SCS patients was significantly different (p < 0.02) from the mean change for patients taking prostaglandin E1 at 12 months, although the mean ankle to brachial pressure index of the SCS patients was not significantly increased. Transcutaneous oxygen pressure (TcPo₂) did not differ between the SCS and CMM groups (ESES), but was higher (p < 0.05) in SCS patients than in the prostaglandin E1 group at 12 months.

A priori subgroup analysis of the ESES trial found that patients with intermediate skin microcirculation before treatment showed a non-significant trend for the SCS group to have a lower amputation rate at 18 months follow-up (Appendix 5.3). Success within subgroups can suggest that selection criteria be employed to decide which patients are more likely to benefit from SCS. Ubbink and Vermeulen102 suggested that SCS may be more effective for CLI patients if they have a TcPo₂ between 10 and 30 mmHg. The systematic review by Ubbink et al. 87 included a non-randomised trial101 that suggested patients with adequate TcPo₂, pain relief and paraesthesia coverage in response to test stimulation, benefited significantly more from SCS than from conventional treatment. Subgroup analysis for the Jivegard trial, in surviving patients without arterial hypertension, found a significantly lower (p = 0.045) amputation rate in the SCS group than the analgesics group. On a different outcome, the Claeys trial suggested a better response with SCS in patients with TcPo₂ > 10 mmHg in terms of ulcer healing.

HRQoL outcomes

One of the trials, ESES, assessed HRQoL (Table 13). There was no significant difference between SCS and CMM on Nottingham Health Profile (NHP; significant reduction in NHP pain score for both groups), EuroQol 5D, or the mobility index of the Sickness Impact Profile.

Subgroup analysis in ESES found that non-amputated patients had better mobility and energy scores on NHP in the SCS compared with the control group.

Summary

Evidence from CLI trials suggests that SCS was more effective than CMM in reducing the use of analgesics up to 6 months, but not at 18 months.

Although there was significant pain relief achieved, there was no significant difference between groups in terms of pain relief, for SCS versus CMM or analgesics treatment. SCS had similar limb survival rates to CMM, or analgesics treatment, or prostaglandin E1. SCS and CMM were similarly effective in improving HRQoL.

Pain outcomes

One of the trials (Hautvast) reported pain as measured by VAS (Table 14). There was no significant difference between SCS and inactive stimulator groups, despite the SCS group having a significant reduction in mean pain score at 6 weeks (p = 0.03).

Medication outcomes

Three trials (DeJongste, ESBY, Hautvast) investigated nitrate consumption and all found significant differences between SCS and the control group (Table 15). DeJongste found a greater reduction (p < 0.05) in glyceryl trinitrate consumption for SCS than for the No SCS group at 6–8 weeks. The ESBY trial found significantly more reduction for CABG, than for SCS group, for long-acting nitrates (p < 0.0001) at 6 months, although there was no significant difference in short-acting nitrates with both groups having a significant reduction (p < 0.0001) in consumption from baseline. Hautvast found a significant reduction (p = 0.01) in nitrate consumption in the SCS group at 6 weeks, which differed significantly from the Inactive stimulator group (p = 0.03).

Angina – functional outcomes

Three of the trials (DeJongste, ESBY, Hautvast) assessed the frequency of angina attacks (Table 16). There was a significantly reduced frequency of angina attacks in the SCS group compared with the No SCS group (p < 0.05) at 6–8 weeks (DeJongste), and the SCS compared with Inactive stimulator at 6 weeks (p = 0.01) (Hautvast). The ESBY trial found no difference between treatment groups, with a significant reduction in angina attacks for both the SCS and CABG groups at 6 months.

The SPiRiT trial assessed change in angina class as measured by the Canadian Cardiovascular Society angina scale. No difference was found at 12 months between the SCS and PMR groups in an analysis treating deaths and dropouts as failures, although an analysis excluding patients without follow-up indicated that the SCS group had greater improvement in the Canadian Cardiovascular Society class (p = 0.042).

Three of the trials had the SCS device switched on during exercise testing (DeJongste, SPiRiT, Hautvast). Total exercise duration (Table 17) was significantly more improved in the SCS than the No SCS group (p < 0.03) (DeJongste), and in SCS than in the Inactive stimulator (p = 0.03) (Hautvast), but there was no difference between SCS and PMR (SPiRiT). Exercise testing of time to angina was significantly more improved in the SCS than the No SCS group (p < 0.05) (DeJongste), and in SCS than Inactive stimulator (p = 0.01) (Hautvast), and in SCS than PMR at 3 months (p = 0.028) although it was not significantly different at 12 months (SPiRiT).

In the ESBY trial, the SCS patients had the device switched off during exercise testing, which would be expected to diminish its effectiveness (ESBY authors had previously reported on a case series of angina patients in which SCS when switched on could improve exercise training103). The exercise test in the ESBY trial found that at 6 months CABG had a significantly greater increase in maximum workload capacity than SCS (p = 0.02).

HRQoL outcomes

All four trials evaluated HRQoL, all using different outcome measures. DeJongste assessed Daily activity and Social activity scores and showed a significantly greater improvement for both measures (p < 0.05) for SCS compared with the No SCS group at 6–8 weeks (Table 18). The ESBY trial found no differences between the CABG and SCS groups, at 6 months and 58 months, in any subcategory of NHP, with both groups significantly improving from baseline (p < 0.001). Both groups had significant improvements in ‘energy’ and ‘pain’ scores, and the magnitude of improvement in NHP total score for both groups was > 30%, with both groups reaching a level comparable to that of a healthy population. There was no difference between SCS and PMR as measured by the Short Form 36 at 3 and 12 months (SPiRiT). Hautvast found no difference between SCS and Inactive stimulator groups at 6 weeks when measured using the Linear Analogue Self-Assessment scale, although the SCS group showed a significant improvement (p = 0.01) (Table 18).

Two trials assessed disease-specific quality of life. The ESBY trial employed the Questionnaire Angina Pectoris QLQ-AP, and found no difference between SCS and CABG groups at 6 months and 58 months, with both groups showing significant improvements at 6 months (p < 0.001) and the results remaining consistent after 4.8 years. The SPiRiT trial found no difference between SCS and PMR groups on the Seattle Angina Questionnaire, with both groups improved at 3 and 12 months.

Summary

Evidence from Angina trials suggested that SCS was more effective than No SCS or Inactive stimulator for nitrate consumption, frequency of angina attacks, exercise duration and time to angina at 6–8 weeks. SCS was also more effective than PMR (at 3 months, not at 12 months) for time to angina. HRQoL was more improved by SCS than No SCS at 6–8 weeks.

There was no difference between SCS and Inactive stimulator in terms of pain relief. SCS and CABG had similar results for short-acting nitrates and frequency of angina attacks. There was no difference in effectiveness of SCS and PMR for change in angina class or exercise duration. SCS did not differ from CABG or PMR or Inactive stimulator in terms of HRQoL.

The SCS was less effective than CABG in reducing consumption of long-acting nitrates. SCS was less effective than CABG in increasing maximum workload capacity, although the SCS device was switched off during this comparison.

Comparator 1: conventional medical management

The CMM comprises drug therapy and non-drug therapy. The drug therapy consists of opioids, non-steroidal anti-inflammatory drugs (NSAIDS), antidepressants and antiepileptics. On the other hand, non-drug therapy comprises physical rehabilitation, psychological rehabilitation, acupuncture, blocks, massage, chiropractic sessions, acupressure, etc.

Comparator 2: reoperation

Reoperation is defined as lumbosacral spine surgery. Reoperation patients also receive CMM.

Costs of health states, monitoring and treatments in the model

The costs of CMM are taken from the PROCESS study,59 which reported data based on a follow-up of 6 months (Table 21). It is assumed that the annual cost of CMM in year 2 is reduced by 13.5% compared to the cost of year 1. This assumption was taken from a 5-year analysis of cost for CMM in Canada. 111

Patients that undergo SCS have costs additional to CMM including screening, device implant, device reimplant, etc. (Table 22).

For FBSS patients that undergo revisional spinal surgery, it is assumed that the cost of CMM is the same as for SCS patients if they achieve optimal pain reduction. For those patients that do not achieve optimal pain reduction, it is assumed that the CMM cost is the same as for the patients that undergo CMM alone. The cost of revisional surgery of £4252 is taken from the NHS National Tariff R09. 112

For CRPS patients, it is assumed that the costs of drug and non-drug treatments are similar to those for FBSS.

Utilities used in the model

As per NICE recommendations,110 the health-state quality of life utilities are based on the EuroQol 5D (EQ-5D) administered within the PROCESS trial. 59 The baseline utility value for all patients is 0.168 (Table 23).

FBSS: SCS + CMM versus CMM

The results of the probabilistic analysis using a 15-year horizon suggest that SCS + CMM compared to CMM alone produce more QALYs. The cost-effectiveness acceptability curve (ABHI report; Appendix 12 p.117) shows that when using a threshold of £20,000 per QALY the probability of SCS + CMM being cost-effective is around 80%. Additionally, at a £30,000 per QALY threshold, this probability is over 95%.

FBSS: SCS + CMM versus reoperation

The results found in the probabilistic analysis using a 15-year horizon suggest that SCS + CMM compared to reoperation produce more QALYs. The cost-effectiveness acceptability curve (ABHI report; Appendix 13 p.121) shows that when using a threshold of £20,000 per QALY the probability of SCS + CMM being cost-effective is higher than 90%. Additionally, at a £30,000 per QALY threshold, this probability is around 98%.

CRPS: SCS + CMM versus CMM alone

Using a threshold of £20,000 per QALY, the results of the probabilistic analysis using a 15-year horizon suggest that the probability of SCS + CMM being cost-effective is over 40% whereas the probability at a £30,000 per QALY threshold is higher than 60% (ABHI report; Appendix 14 p.124).

Population considered in the ScHARR economic evaluation

The model evaluates the cost-effectiveness of treatment in the three following populations:

• Adult patients (> 18 years old) with FBSS suffering from neuropathic pain of radicular origin predominantly in the legs for at least 6 months after one or more surgeries for a herniated disc (anatomically successful), as per the PROCESS trial59 (SCS versus CMM). Their pain intensity is at least 50 mm on VAS (where 0 mm represents no pain and 100 mm represents the worst pain possible). Some patients had undergone other procedures, for instance spinal fusion, laminectomies or repeat lumbar disc operations.

• Adult patients (> 18 years old) with FBSS suffering from persistent or recurrent radicular pain, after one or more lumbosacral spine surgeries. All patients meet the criteria for surgical intervention (pain refractory to conservative care, with concordant neurological tension, and imaging finding of neural compression). Patients receive a second opinion from a neurosurgeon. Patients are excluded if they have a disabling neurological deficit in the distribution of a nerve root caused by surgically remediable compression or critical cauda equina compression. This patient population represents that of the North trial62 (SCS versus reoperation).

• Patients with CRPS are based on the Kemler trial65 (SCS versus CMM). Patients are adults (> 18 years old) who have suffered the indication for at least 6 months with impaired function and symptoms beyond the area of trauma. The patients’ pain is restricted to one hand or foot and affects the entire hand or foot. Patients have not had a good level of response to standard treatment and have a pain intensity of at least 50 mm on a VAS (where 0 mm represents no pain and 100 mm represents very severe pain). Patients are excluded if they suffer from Raynaud’s disease, neurological abnormalities not related to CRPS, other conditions affecting the function of the qualifying extremity, a blood-clotting disorder or use of a pacemaker.

Treatment/comparator

Guidelines from the European Federation of Neurological Societies (EFNS) make an evidence-based recommendation for the use of SCS in the treatment of FBSS and CRPS type I. 21 The British Pain Society suggests that SCS may be considered when first-line therapies for chronic pain have failed. These therapies can include drug therapies, physical therapies (non-drug therapies) and surgical interventions. 35

Structure of the model

A decision tree model is used to explore the clinical pathway of individuals FBSS or CRPS in the short term. A Markov model is used to explore the clinical pathway of individuals suffering from FBSS or CRPS in the long term. The pathway is divided into a finite number of mutually exclusive health states. The proportion of patients in each health state is determined by the probabilities of achieving different levels of pain relief.

Time horizon

The model explores the costs and benefits accrued through pain relief over a 15-year period. This timeframe is taken from an observational clinical study that assesses clinical predictors of outcomes (e.g. age, sex, aetiology of pain, duration of pain, duration of treatment, employment status and quality of life) in patients who received SCS for the treatment of chronic pain. The study presents a Kaplan–Meier survival curve that illustrates subsequent gradual loss of pain control during a 15-year period. It was decided not to extrapolate beyond the 15-year period because of the increased uncertainty this would cause. 113

Decision tree health states modelled

The first stage of the model (first 6 months) is defined with four possible health states:

• optimal pain relief with no complications

• optimal pain relief with complications

• suboptimal pain relief with no complications

• suboptimal pain relief with complications.

It is assumed that the patients do not change therapy during the first 6 months of treatment. The decision tree is populated with data from the Kumar et al. (PROCESS), North et al. and Kemler et al. RCTs and it replicates their randomisation phase, where the proportions of patients experiencing ‘no pain relief’ were not reported. 59,62,65 For the decision tree model all patients commence suffering from FBSS or CRPS and enter either the SCS trial or CMM (Figures 3 and 4).

FIGURE 3.

Six-month decision tree for SCS + CMM versus CMM in FBSS and CRPS.

FIGURE 4.

Six-month decision tree for SCS + CMM versus reoperation in FBSS.

Markov health states modelled

The second stage of the model (Markov process) is defined according to the indication. For FBSS, there are five possible health states:

• optimal pain relief (includes patients with or without complications)

• suboptimal pain relief (includes patients with or without complications)

• no pain relief (SCS)

• no pain relief (surgery)

• dead (all causes).

In this model, patients that have no pain relief would change treatment: patients on SCS will move to CMM and patients on CMM will undergo surgery (Figure 5). For CRPS, there are four possible health states:

• optimal pain relief (includes patients with or without complications)

• suboptimal pain relief (includes patients with or without complications)

• no pain relief (SCS)

• dead (all causes).

FIGURE 5.

Schematic of the long-term Markov model for FBSS.

Surgery is not an option for CRPS patients (Figure 6). It is assumed that all patients are in the same health state that they were at the time of the decision tree when entering the Markov model. During each 3-month cycle of the model a proportion enter one of the health states defined in Figure 5 and Figure 6.

FIGURE 6.

Schematic of the long-term Markov model for CRPS.

Optimal pain relief is defined as having at least 50% pain reduction from baseline, measured on a VAS. Suboptimal pain relief is defined as having less than 50% pain reduction from baseline, measured on a VAS.

Perspective

A UK NHS perspective is used, therefore productivity lost through illness or costs incurred directly by patients are not included. Discount rates of 3.5% are applied to both costs and health benefits, according to current NICE guidelines. 110 Costs are at 2007 prices.

Probabilities of levels of pain relief

Costs and resources used

CRPS

It is assumed that the drug and non-drug costs for CMM in CRPS are equivalent to those costs for CMM in FBSS. Clinical experts reported that CMM for both indications (FBSS and CRPS) includes opioids, antineuropathic pain drugs, physiotherapy, psychology and occasional nerve block. Therefore, it was agreed that the costs of drug treatments in CRPS were similar to those of FBSS.

Health-related quality of life utility by health state

A literature review was carried out to obtain the most appropriate and recent published evidence on utility measures for the health states modelled (Appendix 2).

The criteria used to evaluate the identified studies are as follows:

• use of a preference-based utility instrument (EQ-5D, in the UK)110

• UK setting studies are preferred to non-UK studies

• patients suffering from neuropathic pain.

There is a dearth of published evidence reporting quality of life measurements for individuals with chronic neuropathic pain. Utility values for FBSS are based on those reported in the PROCESS trial. 59 The utility for no pain relief health state is assumed to be equal to the baseline utility across all patients. It is found that having a complication reduced the utility values by 0.07 (Table 29).

A study by McDermott et al. 16 investigated the burden of neuropathic pain in a cross-sectional survey of 602 patients recruited from general practitioners in six European countries: France, Germany, Italy, the Netherlands, Spain and the United Kingdom. The population comprised adult patients (> 18 years old) with at least a 1-month history of the condition who had experienced symptoms in the week before the survey. The patient questionnaire included the Brief Pain Inventory, the EQ-5D, and questioned productivity, non-drug treatment and frequency of physician visits. Most patients reported moderate (54%) or severe (25%) pain. There was a significant association (p < 0.001) between pain severity and EQ-5D scores. The scores for mild, moderate and severe pain severity were 0.67, 0.46 and 0.16, respectively. In this ScHARR economic evaluation, it is assumed that in CRPS for optimal pain relief, the utility value is 0.67, for suboptimal pain relief the utility value is 0.46 and no pain relief has a utility value of 0.16. These figures suggest that the benefit achieved from having a pain reduction of at least 50% is approximately 0.5 utility units, showing that the prevailing factor in utility values is level of pain.

Taylor and Taylor108 reported a utility loss associated with SCS complication (e.g. infection, electrode or lead problems) as – 0.05 utility units. This was applied to both optimal and suboptimal pain relief health states. Table 29 presents the utility values used in this economic assessment.

Mortality

National statistics were accessed online to obtain the proportion of patients dying from all causes. 36 The death rate per annum is 0.94%.

Key modelling assumptions

A summary of the key modelling assumptions is provided below.

• Optimal pain relief is defined as achieving at least 50% of pain relief from baseline, measured by VAS.

• Suboptimal pain relief is defined as achieving less than 50% of pain relief from baseline, measured by VAS.

• No patient dies within the first 6 months (short-term decision tree).

• Patients, when entering the Markov process remain in the same health state (optimal or suboptimal pain relief) as they were at the end of the first 6 months (short-term decision tree model).

• It is assumed that patients on CMM do not experience either short-term or long-term complications. 114

• It is assumed that after 6 months complications in SCS occur at a rate of 18% per annum. 49

• It is assumed that the cost of device explant is the same as the cost of failed trial stimulation.

• It is assumed that the cost of acupuncture is the same as the cost of massage. 119

• It is assumed that the annual cost of CMM in year 2 is reduced by 17.8% compared to the cost of year 1. 111

• After year 2 the cost of CMM remains constant.

• It is assumed that the drug and non-drug costs for CMM in CRPS are equivalent to those costs for CMM in FBSS.

• Annual SCS withdrawal rate is 3.24%. 113

• The model explores the cost and benefits accrued through pain relief over a 15-year period. 113

• In FBSS, the utility for the no pain relief health state was assumed to be equal to the baseline utility across all patients (0.168). 59

• In CRPS, the utility values were taken from a cross-sectional survey that investigates the burden of neuropathic pain. 16

• The transitions from health states are presented in Table 30 and are based on the assumptions discussed above.

Cost-effectiveness ratios

The ICERs measure the additional cost per QALY gained of Treatment A versus Treatment B such that ICER = (Cost of Treatment A – Cost of Treatment B) / (Utility of Treatment A – Utility of Treatment B).

Population considered in the ScHARR economic evaluation

The model is based on a prospective observational study that compares the cost-effectiveness of CABG, PCI or CMM. 120 Consecutive, unselected patients who had undergone coronary angiography between April 1996 and April 1997 at three hospitals of one NHS Trust in London were recruited. A total of 4121 patients were identified and followed for 6 years. From these patients, a subgroup of 1740 patients was rated appropriate to have CABG (n = 815), PCI (n = 385) or both revascularisation procedures (n = 520). Twenty patients were excluded because they died before undergoing revascularisation. Clinical judgement and available evidence were used to define appropriateness using a nine-member Delphi panel. 121 Approximately 70% of the 1720 had a Canadian Cardiovascular Society (CCS) score III–IV (severe angina). Hence, it could be assumed that the population of this study was representative of patients with refractory angina. 120 Three different scenarios based on clinical appropriateness were defined:

• scenario 1: patients clinically appropriate to receive CABG

• scenario 2: patients clinically appropriate to receive PCI

• scenario 3: patients clinically appropriate to receive both revascularisation procedures (CABG and PCI).

Treatment/comparator

Time horizon

The model explores the costs and benefits accrued through pain relief over a 6-year period. This timeframe is taken from an observational clinical study that assesses clinical predictors of outcomes in patients who received CABG, PCI or both revascularisation procedures in the treatment of angina pectoris. 120

Perspective

A UK NHS perspective is used, therefore productivity lost through illness or costs incurred directly by patients are not included. Discount rates of 3.5% are applied to both costs and health benefits, according to current NICE guidelines. 110 Costs are at 2007 prices.

Costs and resources used

Results for 15-year time horizon and 4-year device longevity

Table 35 shows the discounted cost and QALYs for each indication based on a 4-year device longevity and a 15-year time horizon. The results range from £7043 per QALY for FBSS (SCS + CMM versus reoperation) to £25,095 per QALY for CRPS (SCS + CMM versus CMM).

The results presented in Table 35 suggest that SCS is expected to be more effective for FBSS than for CRPS. This analysis suggests that although SCS and CMM for CRPS are slightly less expensive than SCS and CMM for FBSS, the small difference between the effectiveness of SCS and CMM increases the incremental cost-effectiveness ratios (£25,095 per QALY).

Table 36 shows the discounted incremental costs and ICERs for each indication, at 6 months and every subsequent year until 15 years, based on a 4-year device longevity. The results suggest that for FBSS (CMM and reoperation) the ICERs are below £20,000 per QALY gained in year 3 onwards, although in year 5 the ICERs increase because of the costs incurred from having a battery replacement. Appendix 10 presents more detailed tables with discounted and undiscounted costs, QALYs and ICERs for each indication (FBSS and CRPS).

Another parameter that can impact the results is the cost of the SCS device. Table 37 shows the ICERs for FBSS (SCS + CMM versus CMM) using a 4-year device longevity and a device costs range from £7000 to £14,000.

At any device cost in the range from £5000 to £14,000 and a device longevity of 4 years, the ICERs for the FBSS indications (CMM and reoperation) are below £20,000 per QALY. In the CRPS indication, when the device cost is £8000 the ICER is £26,555. When the device cost ranges from £9000 to £15,000, the ICERs are above £30,000 per QALY.

Figure 8 shows the trend of the incremental cost-effectiveness ratios for different SCS device costs. The cost-effectiveness estimates are more sensitive to the device cost with CRPS than with FBSS. The expected device cost to obtain ICERs below £30,000 per QALY is £8000.

FIGURE 8.

Incremental cost-effectiveness ratios versus device cost.

Results for 15-year time horizon and variable device longevity and device cost

The most sensitive parameters are device longevity and device cost. Table 38 presents the results when the parameters device longevity and device average price are varied simultaneously, for the FBSS indication (SCS + CMM versus CMM). The tables for FBSS (SCS + CMM versus reoperation) and CRPS are presented in Appendix 10.

Results for 15-year time horizon, 4-year device longevity and variable proportion of patients failing the test stimulation

According to RCTs the proportion of patients that are successful in the trial stimulation varies from 67% to 83%. The discounted ICER for FBSS (SCS + CMM versus CMM) with a success rate of 67% is £8190 per QALY while the ICER with a success rate of 82.7% is £7996 per QALY, suggesting that the impact of the proportion of patients failing the test stimulation on the overall ICERs is very small.

Probabilistic sensitivity analysis results

Comprehensive sensitivity analyses were undertaken to explore the joint uncertainty in model parameters on the cost-effectiveness of each indication (Appendix 9). Monte Carlo sampling techniques (10,000 samples) were used to generate information on the probability that each indication (FBSS: SCS versus CMM, FBSS: SCS versus reoperation, and CRPS: SCS versus CMM) is optimal in terms of amount of net benefit. The results of the probabilistic sensitivity analyses are presented as incremental cost-effectiveness acceptability curves (CEACs). Table 39 below is a summary of the mean net benefit at thresholds of £20,000 per QALY gained and £30,000 per QALY gained for the base-case analysis (device price of £7745 and a 15-year time horizon). The 95% confidence interval indicates the uncertainty around the mean benefit.

FBSS: SCS + CMM versus CMM

The results of the probabilistic analysis using 15-year horizon and a base case using a 4-year device longevity and a device price of £7745 suggest that SCS + CMM compared to CMM alone produce more QALYs. The cost-effectiveness acceptability curve (Figure 9) shows that when using a threshold of £20,000 per QALY the probability of SCS + CMM being cost-effective is around 99.02%. Additionally, at a £30,000 per QALY threshold this probability is around 99.96%.

FIGURE 9.

Cost-effectiveness acceptability curve for FBSS: SCS + CMM versus CMM.

FBSS: SCS + CMM versus reoperation

The results found in the probabilistic analysis using the base case, suggest that SCS + CMM compared to reoperation produce more QALYs. The cost-effectiveness acceptability curve (Figure 10) shows that when using a threshold of £20,000 per QALY the probability of SCS + CMM being cost-effective is 100%.

FIGURE 10.

Cost-effectiveness acceptability curve for FBSS: SCS + CMM versus reoperation.

CRPS: SCS + CMM versus CMM alone

The results of the probabilistic analysis, using a 15-year horizon, a 4-year device longevity and a device price of £7745, suggest that the probability of SCS + CMM being cost-effective at a £20,000 per QALY threshold is around than 78.36% (Figure 11). Additionally, at a £30,000 per QALY threshold this probability is around 97.38%.

FIGURE 11.

Cost-effectiveness acceptability curve for CRPS: SCS + CMM versus CMM.

Scenario 1: Patients clinically appropriate to receive CABG

Figure 12 presents the incremental difference of SCS + CMM compared with CABG, PCI and CMM. The vertical axis represents the incremental survival benefit due to SCS + CMM versus revascularisation (CABG or PCI) or CMM and the horizontal axis shows the incremental cost per LYG.

FIGURE 12.

Threshold analysis in terms of incremental cost per LYG.

Figure 12 shows that for patients who are clinically appropriate to receive CABG, SCS + CMM must provide an additional 0.0235 life-years when compared with CABG to achieve £20,000 per LYG and 0.0155 additional life-years to achieve £30,000 per LYG. SCS + CMM must provide an additional 0.185 life-years when compared with PCI to achieve an incremental cost per LYG of £20,000 and at least 0.125 additional life-years to achieve incremental costs per LYG below £30,000. The model suggests that SCS + CMM must provide at least an additional 0.35 life-years when compared with CMM to achieve incremental costs per LYG below £20,000. Figure 12 shows that SCS + CMM should provide an additional 0.23 additional life-years to achieve an incremental cost per LYG of £30,000.

Figure 13 presents the incremental cost-effectiveness ratios of SCS + CMM compared with CABG, PCI and CMM. The horizontal axis represents the incremental QALYs due to SCS + CMM versus revascularisation (CABG or PCI) or CMM and the vertical axis shows the ICERs.

FIGURE 13.

Threshold analysis in terms of incremental cost per QALYs.

Table 40 shows that for patients who are clinically appropriate to receive CABG, SCS + CMM must provide at least an additional 0.0231 and 0.0154 QALYs when compared to CABG to achieve ICERs of £20,000 and £30,000 per QALY gained, respectively. Therefore, the SCS utility value to achieve an ICER of £20,000 per QALY is 0.6218 whereas the utility value should be 0.6203 to achieve £30,000 per QALY gained. SCS + CMM must provide at least an additional 0.1877 and 0.1251 QALYs when compared to PCI to achieve ICERs of £20,000 and £30,000 per QALY gained, respectively. Therefore, the SCS utility value to achieve an ICER of £20,000 per QALY is 0.6001 whereas the utility value is 0.5884 to achieve £30,000 per QALY gained. Table 40 also shows that SCS + CMM must provide at least an additional 0.3480 and 0.2320 QALYs when compared with CMM to achieve ICER of £20,000 and £30,000 per QALY gained, respectively. The SCS utility value to achieve an ICER of £20,000 per QALY is 0.6321 whereas the utility value is 0.6103 to achieve £30,000 per QALY gained.

Scenario 2: Patients clinically appropriate to receive PCI

For patients who are clinically appropriate to receive PCI, SCS + CMM dominates in terms of cost per LYG when compared with CABG. This means that SCS cost less and accrued more survival benefits. The model suggests that in terms of incremental cost-effectiveness ratios (£/QALY), SCS + CMM is dominant when the incremental QALYs are in the range from 2.25 to 0.12.

Figure 14 presents the incremental difference of SCS + CMM compared with PCI and CMM. The vertical axis represents the incremental survival benefit due to SCS + CMM versus revascularisation (PCI) or CMM and the horizontal axis shows the incremental cost per LYG.

FIGURE 14.

Threshold analysis in terms of incremental cost per LYG.

The model suggests that SCS + CMM must provide an additional 0.235 life-years when compared with PCI to achieve an incremental cost per LYG of £20,000 and at least 0.155 additional life-years to achieve incremental costs per LYG below £30,000. SCS + CMM must provide at least an additional 0.38 life-years when compared with CMM to achieve incremental costs per LYG below £20,000. Figure 14 shows that SCS + CMM should provide an additional 0.25 to achieve an incremental cost per LYG of £30,000.

Figure 15 presents the incremental cost-effectiveness ratios of SCS + CMM compared with PCI and CMM. The vertical axis represents the incremental QALYs due to SCS + CMM versus revascularisation (CABG or PCI) or CMM and the horizontal axis shows the ICERs.

FIGURE 15.

Threshold analysis in terms of incremental cost per QALYs.

Table 41 shows that for patients who are clinically appropriate to receive PCI, SCS + CMM must provide at least an additional 0.2337 and 0.1558 QALYs when compared with PCI to achieve ICERs of £20,000 and £30,000 per QALY gained, respectively. Therefore, the SCS utility value to achieve an ICER of £20,000 per QALY is 0.6650 whereas the utility value is 0.6504 to achieve £30,000 per QALY gained. SCS + CMM must provide at least an additional 0.3777 and 0.2518 QALYs when compared to CMM to achieve ICERs of £20,000 and £30,000 per QALY gained, respectively. The SCS utility value to achieve an ICER of £20,000 per QALY is 0.6620 whereas the utility value is 0.6384 to achieve £30,000 per QALY gained.

Scenario 3: Patients clinically appropriate to receive both revascularisation procedures

For patients who are clinically appropriate to receive CABG and PCI, SCS + CMM dominates in terms of cost per LYG when compared with CABG. This means that SCS cost less and accrued more survival benefits. The model suggests that in terms of incremental cost-effectiveness ratios, SCS + CMM is dominant when the incremental QALYs are in a range from 2.20 to 0.07.

Figure 16 presents the incremental difference of SCS + CMM compared with PCI and CMM. The vertical axis represents the incremental survival benefit due to SCS + CMM versus revascularisation (PCI) or CMM and the horizontal axis shows the incremental cost per LYG.

FIGURE 16.

Threshold analysis in terms of incremental cost per LYG.

The model suggests that SCS + CMM must provide an additional 0.1 life-years when compared with PCI to achieve an incremental cost per LYG of £20,000 and at least 0.067 additional life-years to achieve incremental costs per LYG below £30,000. SCS + CMM must provide at least an additional 0.275 life-years when compared with CMM to achieve incremental costs per LYG below £20,000. Figure 16 shows that SCS + CMM should provide an additional 0.185 to achieve an incremental cost per LYG of £30,000.

Figure 17 presents the incremental cost-effectiveness ratios of SCS + CMM compared with PCI and CMM. The vertical axis represents the incremental QALYs due to SCS + CMM versus revascularisation (CABG or PCI) or CMM and the horizontal axis shows the ICERs.

FIGURE 17.

Threshold analysis in terms of incremental cost per QALYs.

Table 42 shows that for patients who are clinically appropriate to receive CABG and PCI, SCS + CMM must provide at least an additional 0.1004 and 0.0669 QALYs when compared with PCI to achieve ICERs of £20,000 and £30,000 per QALY gained, respectively. Therefore, the SCS utility value to achieve an ICER of £20,000 per QALY is 0.5687 whereas the utility value is 0.5624 to achieve £30,000 per QALY gained. SCS + CMM must provide at least an additional 0.2762 and 0.1842 QALYs when compared with CMM to achieve ICERs of £20,000 and £30,000 per QALY gained, respectively. The SCS utility value to achieve an ICER of £20,000 per QALY is 0.5829 whereas the utility value is 0.5657 to achieve £30,000 per QALY gained.