ARTICLE IN PRESS. Available online at

Transcription

1 Available online at European Journal of Pain xxx (2008) xxx xxx Quality of life, resource consumption and costs of spinal cord stimulation versus conventional medical management in neuropathic pain patients with failed back surgery syndrome (PROCESS trial) Andrea Manca a, *, Krishna Kumar b, Rod S. Taylor c, Line Jacques d, Sam Eldabe e, Mario Meglio f, Joan Molet g, Simon Thomson h, Jim O Callaghan i, Elon Eisenberg j, Germain Milbouw k, Eric Buchser l, Gianpaolo Fortini m, Jonathan Richardson n, Rebecca J. Taylor o, Ron Goeree p, Mark J. Sculpher a a Centre for Health Economics, Alcuin A Block, University of York, York YO10 5DD, United Kingdom b Department of Neurosurgery, Regina General Hospital, Regina, Canada c Peninsula Medical School, Peninsula Technology Assessment Group, Universities of Exeter and Plymouth, United Kingdom d Department of Neurosurgery, Montreal Neurological Institute and Hospital, Montreal, Canada e Department of Pain and Anaesthesia, James Cook University Hospital, Middlesbrough, United Kingdom f Department of Functional Neurosurgery, Gemelli Catholic University Hospital, Rome, Italy g Department of Neurosurgery, Santa Creu i Sant Pau Hospital, Barcelona, Spain h Pain Clinic, Basildon and Thurrock University Hospitals, Basildon, United Kingdom i Pain Clinic, Axxon Pain Medicine, Brisbane, Australia j Pain Relief Unit, Rambam Medical Center, Haifa, Israel k Department of Neurosurgery, Namur Regional Hospital, Namur, Belgium l Pain Management Services and Neuromodulation Centre, EHC-Morges Hospital, Morges, Switzerland m Pain Clinic, Varese Regional Hospital and Macchi Foundation, Varese, Italy n Department of Pain and Anesthesia, Bradford Hospitals, Bradford, United Kingdom o Health Economics Facility, Health Service Management Centre, University of Birmingham, United Kingdom p Program for Assessment of Technology in Health, St. Joseph s Hospital Department of Clinical Epidemiology and Biostatistics, McMaster University, Hamilton, ON, Canada Received 22 August 2007; received in revised form 28 January 2008; accepted 28 January 2008 Abstract Background: Chronic back and leg pain conditions result in patients loss of function, reduced quality of life and increased costs to the society. Aims: To assess health-related quality of life (HRQoL) and cost implications of spinal cord stimulation plus non-surgical conventional medical management (SCS group) versus non-surgical conventional medical management alone (CMM group) in the management of neuropathic pain in patients with failed back surgery syndrome. Methods: A total of 100 patients were randomised to either the SCS or CMM group. Healthcare resource consumption data relating to screening, the use of the implantable generator in SCS patients, hospital stay, and drug and non-drug pain-related treatment were collected prospectively. Resource consumption was costed using UK and Canadian national figures. HRQoL was assessed using the EuroQol-5D (EQ-5D) questionnaire. Costs and outcomes were assessed for each patient over their first 6- months of the trial. * Corresponding author. Tel.: +44 (0) ; fax: +44 (0) address: [email protected] (A. Manca) /$34.00 Ó 2008 European Federation of Chapters of the International Association for the Study of Pain. Published by Elsevier Ltd. All rights reserved. doi: /j.ejpain

2 2 A. Manca et al. / European Journal of Pain xxx (2008) xxx xxx Results: The 6-month mean total healthcare cost in the SCS group (CAN$19,486; 12,653) was significantly higher than in the CMM group (CAN$3994; 2594), with a mean adjusted difference of CAN$15, 395 ( 9997) (p < 0.001). However, the gain in HRQoL with SCS over the same period of time was markedly greater in the SCS group, with a mean EQ-5D score difference of 0.25 [p < 0.001] and 0.21 [p < 0.001], respectively at 3- and 6-months after adjusting for baseline variables. Conclusions: The addition of SCS to CMM in patients with neuropathic leg and back pain results in higher costs to health systems but also generates important improvements in patients EQ-5D over the same period. Ó 2008 European Federation of Chapters of the International Association for the Study of Pain. Published by Elsevier Ltd. All rights reserved. Keywords: Costs; Quality of life; Neuropathic pain; Spinal cord stimulation; Randomised controlled trial 1. Introduction Neuropathic pain has a reported prevalence in the general and primary care population of 1.5 8% (Bennett, 1998; Hall et al., 2006; Torrance et al., 2006), although its presence is often under-diagnosed and under-treated (Taylor, 2006). The healthcare costs associated with this condition are substantial. A recent United States-based analysis revealed that the healthcare costs of patients with neuropathic pain was three-fold that of age- and sex-matched patients without neuropathic pain (Berger et al., 2004). The most common location of chronic neuropathic pain is the back and legs (Dworkin et al., 2003), and 10 40% of patients who have undergone lumbosacral spine surgery to alleviate neuropathic radicular pain instead experience persistent or recurrent pain (North et al., 1993; Wilkinson, 1991). In carefully chosen patients with this condition, which is often referred to as failed back surgery syndrome (FBSS), spinal cord stimulation (SCS) has been shown to provide effective pain relief (Taylor et al., 2005; Turner et al., 2004). Health policy makers not only expect medical technologies, such as spinal cord stimulation (SCS), to provide added clinical effectiveness, they also expect data on their impact on treatment costs, allowing them to make decisions about value for money. A number of studies have reported the costs of SCS (for a review see Taylor et al., 2004) yet the evidence of cost-effectiveness of SCS for FBSS is currently limited to one model-based analysis (Taylor and Taylor, 2005). This analysis suggests that SCS for the treatment of pain in FBSS patients is cost-effective at two years. However, the authors emphasise the preliminary nature of their conclusions and the need for additional data, in particular the direct assessment of health-related quality of life, healthcare resources as well as the costs of both SCS and conventional medical management (CMM) to be collected in a prospective randomised controlled trial setting. The PROCESS (prospective, randomized, controlled, multicenter study of patients with failed back surgery syndrome) trial has recently demonstrated that the addition of SCS to CMM provides better pain relief and improves the health-related quality of life (HRQoL) and functional capacity of patients with FBSS suffering from chronic neuropathic back and leg pain (Kumar et al., 2007). This paper reports the generic health-related quality of life and costs at 6-months from the PROCESS trial. 2. Methods 2.1. Study design and patients Full details of the PROCESS trial are reported elsewhere (Kumar et al., 2005; Kumar et al., 2007). In summary, 100 patients were recruited in a total of 12 centers in Europe, Canada, Australia, and Israel between April 2003 and June Patients of 18 years or older suffering from predominant neuropathic pain of radicular origin in the legs (radiating in dermatomal segments L4 and/or L5 and/or S1) with or without associated less severe back pain were included. The intensity of pain was at least 50 mm on a visual analogue scale (VAS: 0 equalling no pain, to 100 mm representing the worst possible pain) for at least 6-months after at least one anatomically successful surgery for a herniated disc. All patients had a documented history of nerve injury (i.e., root compression by herniated disc, compatible with the pain complaint). The neuropathic nature of the pain was confirmed according to the routine clinical practice of each investigator and included mapping the pain distribution, examining sensory/motor/reflex changes. All patients randomised to the SCS group underwent a screening trial. Those experiencing at least 80% overlap of their pain with stimulation-induced paresthesia and at least 50% leg pain relief received an implantable neurostimulation system. CMM included oral medication (i.e., opioids, non-steroidal anti-inflammatory drugs [NSAIDs], antidepressants, anticonvulsants/antiepileptics, and other analgesic therapies), nerve blocks, epidural corticosteroids, physical and psychological rehabilitative therapy, and/or chiropractic care. In either group, implantable drug delivery systems and re-operation were not allowed. Patients were assessed prior to randomisation (baseline) and at 3- and 6-months after initiation of treatment. Given that SCS therapy induces paraesthesia, it was not possible to blind patients or investigators during the trial.

3 A. Manca et al. / European Journal of Pain xxx (2008) xxx xxx Measurement of resource consumption Healthcare resource consumption data relating to pre-implant screening, the use of the implantable pulse generator (IPG) in SCS patients, associated hospital stay, and drug and non-drug pain-related treatment between baseline and 6-months after hospital discharge was collected prospectively for each patient using case report forms (CRFs). The data collected included: Pre-implant screening and IPG implant: Hospital staff collected individual-level information on time in theatre, use of hardware (leads, anchors, extensions, IPG), and length of hospital stay (including the type of ward). Use of medications: Details on the daily doses and treatment duration of four main drugs classes (i.e., opioids, NSAIDs, antidepressants, and anticonvulsants including gabapentin and pregabalin) were recorded. Non-drug therapy: Similarly, the utilisation of nondrug therapies (i.e., physical rehabilitation and psychotherapy, acupuncture, nerve blocks, massage, chiropractic, occupational and osteopathic therapies, acupressure, and the use of transcutaneous electrical nerve stimulation [TENS]) was captured. Complications: SCS and non-scs related complications, including those related to CMM, were recorded. In addition, the CRFs were designed to collect SCS-specific healthcare resource utilisation such as additional surgery, initial hospitalisation and readmission required to overcome complications (e.g., SCS-lead replacement, repositioning or biological complications) and non-invasive tests (e.g., X-ray, CT-scan, MRI) Unit costs Healthcare resource consumption was costed at prices using UK and Canadian figures. To estimate the total cost for each element of resource consumption identified in the CRF, country-specific unit costs were obtained from published sources, complemented with data from the literature where needed. Equipment and consumables were costed using manufacturers list prices, whereas drug prices were taken from the drug formulary for the country of interest (BNF, 2006; Ontario Ministry of Health, 2006). Similarly, the cost of in-patient hospital stay was estimated using fully allocated cost figures, while non-drug therapies (such as physical rehabilitation) were costed using a combination of published tariffs and estimates from the literature (Curtis and Netten, 2006; UK Department of Health, 2006). Where necessary, cost figures were uprated for inflation using the national healthcare-specific price index (Curtis and Netten, 2006; Statistics Canada, 2007). Total costs for each of the resource use consumption observed in the study were calculated by multiplying the latter by the relevant national average unit cost. The costing exercise was conducted for Canada and the UK since these two countries accounted for the majority of the patients enrolled in the PROCESS trial. Cost results in this paper are therefore reported in UK Sterling ( ), Canadian Dollars (CAN$) and, to facilitate the interpretation of the results among European readers, in Euros ( ) converting the Canadian figures according to the exchange rate (CAN$/ : 1.54) available at the time of writing. While Euro figures could have been derived from the estimated costs in UK Sterling we decided against this practice, in view of the fact that Canada and the UK have different unit costs for the same item of healthcare resource. For each country these unit costs represent essentially weights that are applied to each item of healthcare resource consumed to express them in monetary terms. The resulting total cost estimate can be seen as a weighted average of the sum of the resource consumption observed in the trial. Because prices (i.e., the weights ) differ between countries, converting total costs in UK Sterling and Canadian Dollars into Euros would lead to different total cost estimates. To avoid confusion it was decided to obtain Euro estimates converting from one currency only. As Canada was larger recruiter than the UK in the PROCESS trial it was considered more appropriate to use the figures in Canadian Dollars as a base for the conversion to Euros Health outcomes In PROCESS, HRQoL was assessed using the Short- Form 36 (SF-36) and the EuroQol-5D (EQ-5D). For the purposes of economic evaluation, a preference-based quality of life that weights patient responses according to values collected from the general population is required. The EQ-5D is one of the most widely used preference-based measures (Kind, 1996). It comprises five questions, each relating to a different dimension: mobility, self-care, ability to undertake usual activity, pain/discomfort and anxiety/depression. Each dimension has three possible levels of severity: no problems, moderate problems, and severe problems. Based on their combined answers to the EQ-5D questionnaire, patients can be classified as being in one of 243 possible health states (not including unconsciousness and death). Each of these health states has an associated utility score (on a 0 [equivalent to death] to 1 [equivalent to good health] scale, where negative values are possible) typically derived from a large sample of the population. The EQ-5D utility score ranges from to 1, and its interval properties make it the most

4 4 A. Manca et al. / European Journal of Pain xxx (2008) xxx xxx commonly-used instrument to measure generic HRQoL within economic assessment of health technologies (Drummond et al., 2005). The utility weights used in the present analysis are those obtained from a large sample of the UK population (Dolan, 1997; Kind et al., 1999). The EQ-5D questionnaire was completed by each patient in the trial at baseline, 3- and 6-months follow up Statistical analysis All primary analyses compared groups by intention to treat (ITT) and, given the time horizon of the analysis, costs and EQ-5D have been left undiscounted (CADTH, 2006; NICE, 2004). Ordinary least squares regression was used to analyse cost and EQ-5D data. Cost data were analysed following the regression-based methodology proposed by Hoch and colleagues (Hoch et al., 2002). Analysis of covariance was used to estimate the differential change in EQ-5D weighted index between the SCS and CMM groups in the trial at two time points: from baseline to 3-months and from baseline to 6-months post SCS implant (Roberts and Torgerson, 1999). The results are presented, firstly reporting the unadjusted analyses for costs and changes in the EQ-5D weighted index, to estimate absolute differences in mean total cost, and in change from baseline score between the two treatment arms. A set of exploratory analyses was also conducted to facilitate the assessment of the contributions of a set of patient baseline variables (i.e., gender, age, number of previous back surgeries, time since last back surgery, unilateral or bilateral leg pain, and in the case of the EQ-5D score its baseline value [Manca et al., 2005] to the study results). The latter set of covariates was defined in the study protocol for the analysis of clinical outcomes, the results of which are reported elsewhere (Kumar et al., 2007). 3. Results Eligible subjects were randomly assigned to CMM alone (n = 48) or CMM with SCS (n = 52). Patients baseline characteristics in the trial were equally balanced between the SCS and CMM groups and descriptive statistics are fully reported in the main clinical report (Kumar et al., 2007). On average, the age at recruitment was 51, with a slightly greater proportion of male patients in the SCS group (58% vs. 44%). Time since last back surgery was on average 4.6 years, with a marginally greater proportion of patients in the SCS group having had more than one previous surgery (54% vs. 46%). The distribution of leg pain (bilateral vs. unilateral) was well balanced. Three patients randomised to CMM requested to be screened for SCS (and thus were implanted with trial leads), but did not receive an IPG during the 6-month period. Following an ITT analysis strategy they were analysed as CMM group patients Resource use Table 1 reports the results in terms of healthcare resources consumption for the treatment strategies compared here. There is clearly a difference in the mix and spread of health resource utilisation, which is due to the different nature of the interventions being compared over the 6-months follow up period. It is therefore unsurprising to observe that SCS required hospitalisation during the screening period, the duration of which was on average 2.5 days. The majority (69%) of patients in the SCS arm were implanted with one lead; the length of time over which the operation room was occupied was on average 105 min. Furthermore, three patients in the CMM arm underwent a lead implant during the first 6-months, but the internalisation (IPG implant) for two of these took place after the 6-months visit and the third patient failed the screening phase. Nine patients randomised to SCS did not achieve the trial screening criteria set in the study. However, five of these patients requested to continue on SCS therapy. Therefore, a total of 48 patients in the SCS arm received an IPG (Synergy, Medtronic, MN), and 67% of these patients required hospitalisation with an average inpatient stay of 2.3 days. Drug treatments were different in SCS and CMM groups. Opioids were used for an average of 11 days less by the SCS group (and in a smaller proportion of patients, i.e., 75% vs. 77%). NSAIDs were used for an average of 38 days less by the SCS patients (68 vs. 106 days). Antidepressants were required for almost two weeks less in the SCS group. Finally, anticonvulsants were used more intensively by the CMM group with 65 prescriptions (i.e., change in drug, or daily dose) as opposed to 37 prescriptions in the SCS group (Table 1). A substantial reduction in healthcare resource use can also be observed when looking at non-drug treatments for pain (e.g., physiotherapy, massage, etc). The proportion of patients requiring contacts with healthcare professionals for treatment related to neuropathic pain was lower in the SCS group. This can be observed particularly in the case of physical therapy (7% SCS vs. 44% CMM), psychotherapy (4% vs. 14%), and massage therapy (0% vs. 14%). This pattern is also accompanied by a reduction in the duration and frequency of contacts with these healthcare professionals. SCS resulted in complications which reduced the net healthcare resource reduction for this group (25% of patients required extra resources to resolve complications such as lead migrations). These required an overnight stay in most cases, the mean duration of which was 2.6 days. In two cases these complications required

5 A. Manca et al. / European Journal of Pain xxx (2008) xxx xxx 5 Table 1 Resource consumption in the PROCESS trial at 6-months a Spinal cord stimulation (SCS) group n =52 Conventional medical management (CMM) group n =48 Intervention costs Screening Hospitalisations for screening Day care, n (%) 4 (7) Required overnight stay, n (%) 49 (92) 3 (6) No hospitalisation, n (%) 1 (1) Duration of in-patient stay, mean (SD) [min max] 2.5 (1) [1 5] 8 (NA b ) [2 19] Leads Required one lead, n (%) 36 (69) 3 (6) Required two leads, n (%) 16 (31) Duration of the intervention (mins), mean (SD) [min max] 105 (33) [35 181] 70 (NA b ) [45 70] IPG implant Hospitalisations Day care, n (%) 13 (27) Required overnight stay, n (%) 32 (67) No hospitalisation, n (%) 3 (6) Duration of in-patient stay, mean (SD) [min max] 2.3 (1.4) [1 6] Leads Required one lead, n (%) 3 (0.7) 1 (2) Required two leads, n (%) 0 0 Duration of the intervention (mins), mean (SD) [min max] 70 (23) [30 151] IPG implanted, n (%) 48 (100) Failed screening Hospitalisations Day care, n (%) 1 (2) Required overnight stay, n (%) 2 (4) No hospitalisation, n (%) 1 (2) Duration of in-patient stay (mins), mean (SD) [min max] 2.5 (NA b ) [2 3] Duration of surgery for explant (mins), mean (SD) [min max] 65 (NA b ) [20 125] Non-invasive investigations CT scan, n (%) MRI, n (%) X-ray, n (%) 1 (2) Myelogram, n (%) Others, n (%) SCS related complication Surgery, n (%) 13 c (25) Duration (mins), mean (SD) [min max] 62 (NA b ) [15 140] Hospitalisation Day care, n (%) 3 (6) Required overnight stay, n (%) 9 (17) No hospitalisation, n (%) Duration of in-patient stay, mean (SD) [min max] 2.6 (NA b ) [1 6] Leads, n (%) Required one lead, n (%) 1 (2) Required two leads, n (%) 1 (2) Non-invasive investigations d CT scan, n (%) MRI, n (%) X-ray, n (%) 5 (10) Myelogram, n (%) Others, n (%) Full blood count 8 (15) Abdominal wound swab 1 (2) Aspiration of seroma 1 (2) Scar tissue biopsy 1 (2) Microbiology 2 (4) (continued on next page)

6 6 A. Manca et al. / European Journal of Pain xxx (2008) xxx xxx Table 1 (continued) Spinal cord stimulation (SCS) group n = 52 Conventional medical management (CMM) group n =48 IPG reprogramming sessions, n (%) 0 14 (27) (33) (29) (11) CMM related/non-scs related costs Drug treatment for pain Oral/transdermal opioids e, n (%) 39 (75) 37 (77) Total number of prescriptions f Mean (SD) [min max] duration (days) per prescription 84 (67) [0 201] 95 (72) [0 193] NSAIDs e, n (%) 33 (63) 30 (62) Total number of prescriptions f Mean (SD) [min max] duration (days) per prescription 68 (66) [1 182] 106 (80) [0 182] Antidepressants e, n (%) 21 (40) 29 (60) Total number of prescriptions Mean (SD) [min max] duration (days) per prescription 97 (74) [3 186] 110 (72) [0 182] Anticonvulsive/antiepileptics e, n (%) 30 (57) 31 (64) Total number of prescriptions f Mean (SD) [min max] duration (days) per prescription 95 (74) [0 212] 89 (77) [1 197] Non-drug treatment for pain Physical rehabilitation e, n (%) 4 (7) 21 (44) Total number of treatment episodes f 6 32 Mean (SD) [min max] duration (days) per treatment episode 49 (53) [7 150] 66 (52) [4 182] Psychotherapy e, n (%) 2 (4) 7 (14) Total number of treatment episodes f 2 12 Mean (SD) [min max] duration (days) per treatment episode 49 (NA b ) [7 92] 68 (55) [1 182] Acupuncture e, n (%) 2 (4) 7 (14) Total number of treatment episodes f 2 8 Mean (SD) [min max] duration (days) per treatment episode 40 (NA b )[5 75] 97 (58) [14 182] Blocks e, n (%) 8 Total number of treatment episodes f Mean (SD) [min max] duration (days) per treatment episode Massage e, n (%) 7 (14) Total number of treatment episodes f 7 Mean (SD) [min max] duration (days) per treatment episode 91 (NA b ) [4 182] Chiropractic session e, n (%) 1 (2) 2 (4) Total number of treatment episodes f 1 2 Mean (SD) [min max] duration (days) per treatment episode 35 (NA b ) 91 (NA b ) [1 182] Acupressure e, n (%) 3 (5) Total number of treatment episodes f 1 Mean (SD) [min max] duration (days) per treatment episode (NA b ) TENS sessions e, n (%) Total number of treatment episodes f Mean (SD) [min max] duration (days) per treatment episode Osteopathic therapy e, n (%) 2 (4) 1 (2) Total number of treatment episodes f 2 1 Mean (SD) [min max] duration (days) per treatment episode 182 (NA b ) 182 (NA b ) Occupational therapy e, n (%) 1 (2) Total number of treatment episodes f 1 Mean (SD) [min max] duration (days) per treatment episode 182 (NA b ) SCS: spinal cord stimulation; CMM: conventional medical management; NSAIDs: non-steroidal anti-inflammatory drugs; MRI: magnetic resonance imaging; CT: computer assisted tomography; IPG: implantable pulse generator. a Unless otherwise stated values are mean (SD) and, where appropriate, [min max]. b Insufficient number of observations to estimate SD. c Number of patients needing one or more surgeries following SCS related complication. d Numbers of patients needing one or more investigation following complication. e This section reports the number (%) of patients requiring a given treatment. f Prescription or treatment episodes represents a change in type (e.g., medication brand or molecule), frequency or dose of treatment.

7 A. Manca et al. / European Journal of Pain xxx (2008) xxx xxx 7 Table 2 Unadjusted cost results at 6-months follow-up between randomised groups Spinal cord stimulation (SCS) group n = 52 Conventional medical management (CMM) group n =48 Between group unadjusted differences Mean (SD) Mean (SD) Difference (95% CI) UK Intervention costs Hospitalisations 1886 (1179) 200 (1390) 1686 ( ) Surgery (excluding devices) 2774 (832) 61 (239) 2713 ( ) Leads 1339 (688) 57 (219) 1282 ( ) Non-invasive investigations 3 (20) 3 ( 8 to3) SCS related complications 576 (1320) 576 ( ) IPG devices 7243 (2641) 7243 ( ) Reprogramming 33 (30) 33 (24 41) CMM related/non-scs related costs Drug pain treatment 1096 (1822) 1417 (1772) 321 ( 1035 to 392) Non-drug pain treatment 134 (389) 1835 (3524) 1701 ( 2715 to 686) Total cost (UK ) 15,081 (4194) 3573 (4082) Canada Intervention costs Hospitalisations 2313 (1505) 244 (1690) 2069 ( ) Surgery (excluding devices) 2344 (703) 51 (202) 2293 ( ) Leads 3456 (1966) 133 (522) 3323 ( ) Non-invasive investigations 1 (10) 1 ( 4 to1) SCS related complications 572 (1329) 572 ( ) IPG devices 9461 (3450) 9461 ( ) Reprogramming 29 (26) 29 (21 36) CMM related/non-scs related costs Drug pain treatment 1139 (2489) 1388 (1974) 249 ( 1137 to 639) Non-drug pain treatment 172 (373) 2177 (3744) 2005 ( 3081 to 927) Total cost (CAN$) 19,486 (5784) 3994 (4526) Total cost ( ) 12,653 (3756) 2594 (2939) SCS, spinal cord stimulation; CMM, conventional medical management; IPG, implantable pulse generator. Results are presented separately for the UK and Canada by attaching the relevant country s unit costs to trial-wide resource consumption. a new lead, while in eight (15%) patients one or two noninvasive investigations such as X-ray or full blood count were needed. Optimisation of programming (SCS therapy) required 1 4 sessions in 32 (62%) patients. Fourteen (27%) patients did not require any re-programming Costs Nine main cost categories under two main headings (i.e., Intervention and CMM related and non-scs related costs ) reflecting the pain treatment algorithm were created for cost comparison purposes (Table 2). The mean per patient cost of the hospitalisation related to leads and IPG implant on an ITT basis was 1,886 [CAN$2313] for SCS patients and 200 [CAN$244] for the CMM group. Given the nature of the interventions being compared, surgery costs (operating theatre time, staff and consumables other than the devices themselves) were higher with SCS group (mean 2,774 [CAN$2344]) than CMM (mean 61 [CAN$51]). The mean cost of the leads amounted to 1339 [CAN$3456] and 57 [CAN$133] in the SCS and CMM groups, respectively. The average per patient costs of the IPG device in the SCS arm was 7243 [CAN$9461]. Note that the costs related to leads and surgery in the CMM group are due to three patients requesting trial screening. The cost of pain medication was 321 [CAN$249] lower in the average SCS patient over the 6-months follow up. Similarly, non-drug therapy costs were lower in the SCS than the CMM arm ( 134 [CAN$172] vs [CAN$2177]). Apart from the cost associated with reprogramming sessions and non-invasive investigations in SCS patients, the only other relevant cost component was associated with the cost of the SCS related complications, mostly related to lead migrations (mean 576 [CAN$572]). At the end of the 6-month follow up period, the total average patient cost was 15,081 (CAN$19,486; 12,653) in the SCS group and 3573 (CAN$3994; 2594) in the CMM group, with a statistically significant (adjusted) differential mean cost (Table 3) of 11,373

8 8 A. Manca et al. / European Journal of Pain xxx (2008) xxx xxx Table 3 Estimation of differential total cost at 6-months follow-up between randomised groups a Dependent variable Analysis for the UK Analysis for Canada Coefficient (95% CI) Coefficient (95% CI) Unadjusted analysis Intervention (incremental cost: SCS CMM) 11,508 ( ,150) 15,492 (13,439 17,545) Constant (mean cost in CMM group) 3573 ( ) 3994 ( ) Analysis adjusted for patients characteristics b Intervention (incremental cost: SCS CMM) 11,373 ( ,234) 15,395 (12,990 17,799) Male 436 ( 1416 to 2289) 912 ( 1581 to 3406) Age at recruitment 81 ( 171 to 7) to 10) Time since last back surgery (years) 112 ( 465 to 242) 72 ( 473 to 328) Bilateral leg pain 1352 ( 843 to 3548) 1006 ( 1627 to 3640) Number of previous back surgeries c Two surgeries 562 ( 2479 to 1355) 416 ( 2899 to 2065) Three surgeries 42 ( 2130 to 2215) 395 ( 2801 to 2010) Four surgeries or more 966 ( 3648 to 1715) 897 ( 4078 to 5871) Constant 3496 ( ) 3589 ( ) SCS, spinal cord stimulation; CMM, conventional medical management; CI, confidence interval. a Values are expressed in national currency (i.e., British pounds and Canadian Dollars). b One surgery is the reference group. c These results allow prediction of the mean cost at 6-months for a hypothetical patient based on a set of baseline characteristics observed in the trial. For instance, 6-months after receiving SCS, the mean cost of an individual with the following characteristics male, aged 40 at recruitment in the trial, with three previous back surgeries, unilateral leg pain, who had the last back surgery 2.5 years before is expected to be, on average, 14,225. This is obtained as follows: 3496 (constant) (coefficient for male dummy for male) + 81 ( ) (coefficient for age departure from the mean age in the trial, i.e., [ ] years) (coefficient for three previous back surgeries dummy for three previous back surgeries) (coefficient for unilateral leg pain dummy for unilateral) (coefficient for time since last back surgery number of years since last back surgery) +11,373 1 (coefficient for treatment group dummy for treatment group). [95%, CI: ,234] (CAN$15,395 [95% CI: 12,990 17,799]; 9997 [95% CI: ,557]; p < ). Fig. 1 shows that while SCS requires an upfront cost, it also brings a considerable saving in terms of CMM related and non-scs related events at 6-months follow up. A further analysis to explore the role of patients baseline characteristics on determining total costs (Table 3) suggests that none of the baseline variables investigated had a statistically significant impact on costs Health-related quality of life Patients in the SCS group displayed a similar mean baseline EQ-5D score than their counterparts in the CMM group (0.13 vs. 0.18) as indicated in Table 4. Fig. 2 shows that the mean EQ-5D score improves at 3- and 6-months in both treatment arms, with the CMM group experiencing a non-statistically significant improvement from its mean baseline value (as represented by the overlapping 95% CI bars). In the SCS group, the improvement in the EQ-5D over time is 14,000 12,000 Intervention costs CMM related and non-scs related costs Total cost (Euro) 10,000 8,000 6,000 4,000 2,000 0 SCS Treatment strategy Note: Intervention costs include: hospitalisations, surgery (exclude devices), leads, non-invasive investigations, SCS related complications. IPG devices, and reprogramming sessions. CMM related / non SCS related costs include: drug and non-drug pain treatments CMM Fig. 1. Composition of total costs (figures reported in Euros).

9 A. Manca et al. / European Journal of Pain xxx (2008) xxx xxx 9 Table 4 Unadjusted comparison of EQ-5D weighted index score at baseline, 3- and 6-months follow-up between randomised groups EQ-5D weighted index score a Spinal cord stimulation (SCS) group n = 52 Conventional medical management (CMM) group n =48 Between group unadjusted difference Between group adjusted difference b Mean (SD) Mean (SD) Mean (95% CI) Mean (95% CI) Baseline 0.13 (0.30) 0.18 (0.31) 3-Months 0.49 (0.31) 0.22 (0.31) 0.27 ( ) 0.27 ( ) 6-Months 0.47 (0.32) 0.25 (0.30) 0.22 ( ) 0.23 ( ) EQ-5D, EuroQol-5D; SCS, spinal cord stimulation; CMM, conventional medical management; CI, confidence interval. a Higher score indicates better quality of life. b Adjusted for baseline imbalance in EQ-5D weighted index score. SCS group CMM group.6.6 EQ 5D score.4.2 EQ 5D score Time period (months) Time period (months) Fig. 2. Mean (95% CI) EQ-5D weighted index score at baseline and follow up by treatment group. greater compared to CMM, with statistically significant mean improvements of 0.27 (95% CI: ; p < 0.001) and 0.23 (95% CI: ; p < 0.001) respectively at 3- and 6-months after adjusting for differences in the baseline EQ-5D scores. Adjusted for patients baseline characteristics (Table 5), the difference in change from baseline score at 3- and 6-months is consistently in favour of the SCS group (mean difference at 3-months: 0.25 [95% CI: ; p < 0.001]; mean difference at 6-months: 0.21 [95% CI: ; p < 0.001]). As expected, the most significant predictor of the value of the EQ-5D at follow up is its baseline value (Manca et al., 2005; Morton and Torgerson, 2003; Roberts and Torgerson, 1999; Vickers and Altman, 2001). The number of years from last back surgery was statistically significant (mean coefficient: 0.02 [95% CI: ; p = 0.009]) in predicting the difference in the change from baseline between SCS and CMM at 6-months but not in predicting the change at 3-months. The latter result was investigated further. It was found that the statistical significance of the result at 6-months was due to the skewed distribution of the variable time from last back surgery (mean: 2.6; standard deviation: 2.3; min max = ). When looking at the effect of SCS among those individuals whose time since last back surgery was one year or less (results not reported), it was found that in the latter sub-group SCS was more beneficial compared to those whose last back surgery occurred earlier. This result was consistent with the findings of the exploratory analysis conducted on the primary clinical outcome (i.e., leg pain relief) in the same group of patients. 4. Discussion While initially expensive due to the upfront implant costs, SCS proffers improvements in generic HRQoL as measured by the EQ-5D. Of the total mean additional cost of SCS, 15% is offset in 6-months time by reducing the use of drugs for pain relief and other non-drug pain treatment reduction. Adjusted for baseline covariates,

10 10 A. Manca et al. / European Journal of Pain xxx (2008) xxx xxx Table 5 Estimation of differential change from baseline in EQ-5D at 3- and 6-month follow-up between randomised groups Dependent variable Analysis at 3-months Analysis at 6-months Coefficient (95% CI) Coefficient (95% CI) Unadjusted analysis Intervention (differential change from baseline: SCS CMM) 0.27 ( ) 0.22 ( ) Constant (mean change from baseline in CMM group) 0.21 ( ) 0.25 ( ) Analysis adjusted for baseline EQ-5D Intervention (differential change from baseline: SCS CMM) 0.27 ( ) 0.23 ( ) EQ-5D at baseline 0.36 ( ) 0.48 ( ) Constant 0.16 ( ) 0.16 ( ) Analysis adjusted for patients characteristics b Intervention (differential change from baseline: SCS CMM) 0.25 ( ) 0.21 ( ) EQ-5D at baseline 0.38 ( ) 0.47 ( ) Male 0.04 ( 0.08 to 0.17) 0.05 ( 0.08 to 0.17) Age at recruitment ( to 0.005) ( to 0.005) Time since last back surgery (years) 0.03 ( 0.03 to 0.02) 0.02 ( ) Bilateral leg pain ( 0.15 to 0.13) 0.08 ( 0.22 to 0.06) Number of previous back surgeries a Two surgeries ( 0.15 to 0.15) 0.05 ( 0.19 to 0.08) Three surgeries 0.14 ( 0.06 to 0.35) 0.02 ( 0.15 to 0.21) Four surgeries 0.20 ( 0.42 to 0.02) 0.17 ( 0.40 to 0.06) Constant 0.20 ( 0.17 to 0.58) 0.25 ( 0.14 to 0.65) EQ-5D, EuroQol-5D; SCS, spinal cord stimulation; CMM, conventional medical management; CI, confidence interval. a One surgery is the reference group. b These results allow prediction of the mean EQ-5D score, at either 3- or 6-months, for a hypothetical patient based on a set of baseline characteristics observed in the trial. For instance, 6-months after receiving SCS, the mean EQ5D of an individual with the following characteristics male, aged 40 at recruitment in the trial, with three previous back surgeries, unilateral leg pain, who had the last back surgery 2.5 years before, and with an EQ-5D at baseline equal to the observed mean in the trial is expected to be, on average, equal to This is obtained as follows: 0.25 (constant) (coefficient for male dummy for male) ( ) (coefficient for age departure from the mean age in the trial, i.e., ( ) years) (coefficient for three previous back surgeries dummy for three previous back surgeries) (coefficient for unilateral leg pain dummy for unilateral) (coefficient for time since last back surgery number of years since last back surgery) (coefficient for EQ-5D at baseline overall mean EQ-5D at baseline in the trial) (coefficient for treatment group dummy for treatment group). SCS has an additional mean cost of 11,373 [CAN$15,395; 9997] per patient. This additional cost relates to the SCS implant procedure including hospitalisations, surgery, leads, IPG and the management of complications. The additional cost of SCS over 6-months needs to be compared with the gains in HRQoL seen from the intervention over the same period. At 6-months follow up, patients receiving SCS in the PROCESS trial experienced an adjusted mean gain of 0.23 on the 0 (equivalent to death) to 1 (equivalent to good health) utility scale, compared to those receiving CMM alone. To put this into context, a recent survey of 36,678 US adults using the EQ-5D instrument estimated the HRQoL decrement associated with a number of illnesses whilst controlling for age, co-morbidities and socio-demographic variables (Sullivan and Ghushchyan, 2006). It estimated mean decrements of 0.02 for asthma, 0.04 for angina pectoris, 0.05 for stroke, and 0.06 for depression. The mean baseline EQ-5D of the patients in the PROCESS trial was on average 0.15 between the two arms, which is considerably less than the baseline EQ-5D score of patients hospitalised after ischemic stroke (Calvert et al., 2005), which has been reported to be 0.31 on the EQ-5D scale. The gain in quality of life of 0.23 on the EQ-5D scale experienced by SCS patients in this study with respect to their baseline value is therefore considerable. Moreover, the changes in EQ-5D are consistent with improvements in other patient related outcomes seen in the PROCESS trial. As reported elsewhere (Kumar et al., 2007), patients in the SCS group experienced significant improvements in pain relief, and improved generic HRQoL as measured by the SF-36 scale as well as functional capacity (as measured by the Oswestry Disability Scale version 2). In contrast, patients in the CMM group have little or no pain relief or other outcome benefits. Of course, without blinding we cannot rule out a placebo effect, but blinding is a difficult issue with SCS as the therapy produces paraesthesia, which must be elicited in the area of the pain if SCS is to be efficacious. In addition, the implantation procedure might in itself produce a placebo effect, but sham operations are ethically difficult to justify (Van Zundert, 2007). The PROCESS trial randomised patients from eight countries. The cost estimates presented here are based on resource consumption data collected on all 100 randomised patients, but are valued using unit costs from

11 A. Manca et al. / European Journal of Pain xxx (2008) xxx xxx 11 the UK and Canada, the two largest recruiting countries in the study. This trial is the first to confirm (Fig. 1) indications of early analyses that the addition of SCS to CMM brings an important cost offset in terms of other treatments. It has been argued that in the medium to long term SCS has the potential to be cost-saving compared to CMM (Kumar et al., 2002; Taylor and Taylor, 2005; Taylor et al., 2004). On the basis of an approximate exchange rate of CAN$1 = 0.43 used in our analysis, there is variation in unit costs between the countries most notably in the cost of day-case visits to hospital and of some non-drug pain therapies. International differences in unit costs have been noted elsewhere (Schulman et al., 1998). These unit cost differences seem to result in some important differences between the UK and Canadian analyses in the total mean cost of the therapies: for the UK analysis, the mean cost per patient of CMM is 44% that of SCS, whereas this proportion is 26% for Canada. Exploratory analyses on potential country effect (results not reported) indicate that the difference in total costs for both the UK and Canada do not appear to be statistically significantly different from the trial overall mean. The PROCESS trial is the first randomised study in FBSS patients to prospectively compare the addition of SCS to CMM alone, and one of very few in the area of neuropathic pain management to collect data suitable for economic analysis (Taylor et al., 2004). This includes detailed resource consumption information on all key areas of health service activity including hospitalisation, the cost of devices, drugs, and other pain management. This provides a key source of data to assess the costeffectiveness of SCS in real life practice. A limitation of PROCESS is that randomised evidence is only available over a period of 6-months follow-up: for ethical reasons and in order to recruit clinicians and patients into the study, it was necessary to permit crossover from CMM to SCS beyond 6-months. However, our analyses do show that compared to baseline, quality of life improvements in SCS patients compared to baseline are sustained in the long-term in the subgroup of patients who continue to use their stimulator in the longer-term (Kumar et al., submitted for publication). The limited follow-up period is one reason why the economic data collected in PROCESS alone are not sufficient to assess whether the additional costs of SCS are justified by the additional benefit patients experience (i.e., the cost-effectiveness of SCS). A full cost-effectiveness analysis would need to consider how costs and HRQoL differences would develop beyond 6-months taking into consideration, for example, the maintenance of SCS hardware, cost of other therapies, and long-term effectiveness of SCS. In addition, such an analysis would need to include any other relevant evidence, particularly trial evidence on the effectiveness of SCS, CMM and other forms of management (Ades et al., 2006; Rivero- Arias et al., 2005). In conclusion, at 6-months observation and compared to CMM alone, SCS increases HRQoL in patients with chronic back and leg pain with a neuropathic component after one or multiple surgeries by 0.21 on the EQ-5D scale at additional mean healthcare cost of 11,373 (CAN$15,395; 9997) per patient. Acknowledgement and declaration of interest This study was funded by Medtronic Inc. The authors had full, non-restricted access to the data. The funding source had no role in the analysis or interpretation of the data. The conclusions of the paper are entirely those of authors. References Ades AE, Sculpher MJ, Sutton A, Abrams KR, Cooper NJ, Welton N, et al. Bayesian methods for evidence synthesis in cost-effectiveness analysis. Pharmacoeconomics 2006;24(1):1 19. Bennett GJ. Neuropathic pain: new insights, new interventions. Hosp Pract (Minneap) 1998;33:107 10, p , Berger A, Dukes EM, Oster G. Clinical characteristics and economics costs of patients with painful neuropathic disorders. J Pain 2004;5: BNF British National Formulary (BNF). British Medical Association and the Royal Pharmaceutical Society of Great Britain, London; CADTH. Guidelines for the economic evaluation of health technologies: Canada. 3rd ed. Ottawa: Canadian Agency for Drugs and Technologies in Health; Calvert MJ, Freemantle N, Cleland JGF. The impact of chronic heart failure on health-related quality of life data acquired in the baseline phase of the CARE-HF study. Eur J Heart Failure 2005;7: Curtis L, Netten A. Unit costs of health and social care, PSSRU. Canterbury: University of Kent; Dolan P. Modeling valuations for EuroQol health states. Med Care 1997;35: Drummond MF, Sculpher MJ, Torrance G, O Brien BJ, Stoddart G. Methods for the economic evaluation of health care programmes. 3rd ed. Oxford: Oxford University Press; Dworkin RH, Backonja N, Rowbotham MC, et al. Advances in neuropathic pain: diagnosis, mechanism and treatment recommendations. Arch Neurol 2003;60: Hall GC, Carroll D, Parry D, McQuay HJ. Epidemiology and treatment of neuropathic pain: the UK primary care perspective. Pain 2006;122: Hoch JS, Briggs AH, Willan A. Something old, something new, something borrowed, something BLUE: a framework for the marriage of health econometrics and cost-effectiveness analysis. Health Econ 2002;11: Kind P. The EuroQoL instrument: an index of health-related quality of life. In: Spilker B, editor. Quality of Life and Pharmacoeconomics in Clinical Trials. 2nd ed. Philadelphia: Lippincott-Raven; p Kumar K, Malik IS, Demeria D. Treatment of chronic pain with spinal cord stimulation versus alternative therapies: cost-effectiveness analysis. Neurosurgery 2002;51(1):

12 12 A. Manca et al. / European Journal of Pain xxx (2008) xxx xxx Kumar K, North RB, Taylor RS, et al. Spinal cord stimulation versus conventional medical management: a prospective, randomized, controlled, multicenter study of patients with failed back surgery syndrome (PROCESS study). Neuromodulation 2005;8: Kumar K, Taylor RS, Jacques L, Eldabe S, Meglio M, Molet J, et al. The effects of spinal cord stimulation in neuropathic pain are sustained: a 24-month follow-up of the PROCESS trial. Pain, submitted for publication. Kumar K, Taylor RS, Jacques L, Eldabe S, Meglio M, Molet J, et al. Spinal cord stimulation versus conventional medical management for neuropathic pain: a multicenter randomized controlled trial in patients with failed back surgery syndrome. Pain 2007;132: Manca A, Hawkins N, Sculpher MJ. Estimating mean QALYs in trialbased cost-effectiveness analysis: the importance of controlling for baseline utility. Health Econ 2005;14(5): Morton V, Torgerson DJ. Effect of regression to the mean on decision making in health care. Brit Med J 2003;326: NICE. Guide to the Methods of Technology Appraisal. London: National Institute for Clinical Excellence; April North RB, Kidd DH, Campbell JN, Long DM. Dorsal root ganglionectomy for failed back surgery syndrome: a five-year follow-up study. J Neurosurg 1993;39: Ontario Case Costing Initiative (OCCI). Ontario Case Costing Initiative (OCCI). Ottawa: Ministry of Health; Rivero-Arias O, Campbell H, Gray A, Fairbank J, Frost H, Wilson- MacDonald J. Surgical stabilisation of the spine compared with a programme of intensive rehabilitation for the management of patients with chronic low back pain: cost utility analysis based on a randomised controlled trial. BMJ 2005;330(7502):1239. Roberts C, Torgerson DJ. Baseline imbalance in randomised controlled trials. Brit Med J 1999;319:185. Schulman K, Burke J, Drummond MF, Davies LM, Carlsson P, Gruger J, et al. Resource costing for multinational neurologic clinical trials: methods and results. Health Econ 1998;7: Statistics Canada. Consumer Price Index, health and personal care, by province. Statistics Canada; Sullivan PW, Ghushchyan V. Preference-based EQ-5D index scores for chronic conditions in the United States. Med Decis Mak 2006;26: Taylor RJ, Taylor RS. Spinal cord stimulation for failed back surgery syndrome: a decision analytic model and cost effectiveness analysis. Int J Technol Assess Health Care 2005;21: Taylor RS. Epidemiology of refractory neuropathic pain. Pain Pract 2006;6:22 6. Taylor RS, Taylor RJ, Van Buyten JP, Buchser E, North R, Bayliss S. The cost effectiveness of spinal cord stimulation in the treatment of pain: a systematic review of the literature. J Pain Symp Manage 2004;27: Taylor RS, Van Buyten JP, Buscher E. Spinal cord stimulation for chronic back and leg pain and failed back surgery syndrome: a systematic review and analysis of prognostic factors. Spine 2005;30: Torrance N, Smith BH, Bennett MJ, Lee AJ. The epidemiology of chronic pain of predominantly neuropathic origin. Results from a general population survey. Eur J Pain 2006;7: Turner JA, Loeser JD, Deyo RA, Sanders SB. Spinal cord stimulation for patients with failed back surgery syndrome or complex regional pain syndrome: a systematic review of effectiveness and complications. Pain 2004;108: UK Department of Health. Reference cost 2004/2005, London: Department of Health; Van Zundert J. Clinical research in interventional pain management techniques: the clinician s point of view. Pain Pract 2007;7: Vickers AJ, Altman DG. Analysing controlled trials with baseline and follow up measurements. Brit Med J 2001;323: Wilkinson HA. The failed back syndrome: etiology and therapy. 2nd ed. Philadelphia: Harper&Row; 1991.