Cost-utility of pregabalin as add-on to usual care versus usual care alone in the management of peripheral neuropathic pain in Belgium

Abstract

Background and objectivess:

The cost effectiveness of pregabalin as an add-on to the standard treatment of Belgian patients with post-herpetic neuralgia (PHN) had been demonstrated in a previously published Markov model. The purpose of this study was to update that model with more recent cost data and clinical evidence, and reevaluate the cost effectiveness from the payer’s perspective of add-on pregabalin in a wider set of NeP conditions.

Methods:

The model, featuring 4-week cycles and a 1-year time horizon, consisted in four possible health states: mild, moderate or severe pain and withdrawn from therapy. Three versions of the model were developed, using transition probabilities derived from pain scores reported in three placebo-controlled studies. The two treatment arms were ‘usual care’ or ‘usual care + pregabalin’. Resource use and utility data were obtained from a chart review and unit costs from recent published data. The final outcome of the model was the incremental cost per quality-adjusted life-year (QALY) gained when adding pregabalin to standard care.

Results:

Based on 1000 simulations, two versions of the model showed that pregabalin was dominant respectively in 94.8% and 67.2% of the simulations, while the incremental cost per QALY was below €32,000/QALY in respectively 99.1% and 94.6% of the simulations. The third version did not show cost effectiveness, despite an incremental cost of only €300 after 1 year. However, in the corresponding study, patients seemed less responsive to GABA analogs, since 55% of them had failed to respond to gabapentin before study inclusion.

Limitations:

The studies upon which the model is based have a short follow-up time as compared to the model horizon. The endpoints of two studies were only provided at the aggregated level and do not necessarily reflect the real practice.

Conclusion:

Based on this analysis, it can be concluded that from a Belgium payer perspective pregabalin offers a slight increase in quality of life in the studied populations as compared to standard care. Pregabalin is cost effective in the majority of cases except in one published clinical study, despite a low incremental cost per year (€300).

Keywords::

• Cost effectiveness
• Markov model
• Neuropathy
• Pregabalin

Introduction

Neuropathic pain (NeP) may result from a wide range of clinical symptoms. Common NeP mechanisms include painful diabetic peripheral neuropathy (PDPN), post-herpetic neuralgia (PHN), trigeminal neuralgia (TN), complex regional pain syndrome (CRPS) and multiple sclerosis1. It is generally chronic and disabling2. The prevalence of neuropathic pain is estimated at 7–8% of the population3,4.

The current treatment of peripheral neuropathic pain (PNeP) includes a wide range of drug and non-drug therapies. Guidelines from the European Federation of Neurological Societies (EFNS) considered that tricyclic antidepressants (TCAs), pregabalin, gabapentin, tramadol and opioids, duloxetine, venlafaxine, topical lidocaine and capsaicin patches (in restricted conditions) with level A evidence. The efficacy of combination like TCA-gabapentin and gabapentin-opioids was also acknowledged5. This variability in approach was confirmed in a cross-sectional assessment of the management of patients with neuropathic pain in Belgium6.

Pregabalin is currently approved for use in neuropathic pain in Belgium, including peripheral NeP conditions, such as the pain experienced by diabetic painful diabetic peripheral neuropathy (PDPN), post-herpetic neuralgia (PHN), phantom limb pain, post-traumatic neuralgia, cancer related NeP and post surgical NeP and central neuropathic pain, such as the pain experienced by patients with spinal cord injury7.

Annemans et al.8 developed a cost-utility model comparing pregabalin as an add-on to ‘usual care’ versus ‘usual care’ alone in patients with PHN pain, where ‘usual care’ was defined as a mix of drug therapies excluding anti-epileptic drugs (AEDs). The structure of the model was based on a previous model published in 20029. The health outcomes of the 2008 model were expressed in quality-adjusted life years (QALYs). Clinical data had been obtained from a randomized trial reported by Van Seventer et al.10 while cost and utility data had been obtained from an expert panel6.

Since the publication by Van Seventer et al. and the model by Annemans, several other studies assessing the clinical efficacy of pregabalin as an add-on to usual treatment in different types of neuropathic pain were published11–14. The purpose of this study was to adapt the model developed by Annemans to allow the evaluation of the cost effectiveness of add-on pregabalin versus usual care in the light of clinical evidence provided by the more recent studies in different neuropathic pain conditions (not only PHN).

Methods

Model structure

The model was an adaptation of a Markov model previously published by Annemans8. The Markov model applied cycles of 4 weeks and the time horizon was 1 year (i.e., 13 cycles) (Figure 1). At the beginning of each cycle, patients could be in four possible health states: severe pain, moderate pain, mild pain, and withdrawn from therapy. There was no mortality specified in this model, since deceased patients were accounted for as ‘withdrawn from therapy’. The state ‘withdrawn from therapy’ was an absorbing state, meaning that patients could not leave this state. The definition of the other health states was based on the pain scores reported by the patient using an 11-points scoring instruments (from 0 to 10, with the highest score corresponding to maximum pain and 0 to no pain). Based upon a previous study a pain score below 4 was considered as ‘no-to-mild pain’, between 4 and 7 as moderate pain and from 7 on as severe pain15. Given that a minimum pain score of 4 at baseline was an inclusion criterion in all the clinical studies10–14, all the patients in the cohort started the simulation in the ‘moderate pain’ or ‘severe pain’ health states. At the end of each cycle, patients could move to a different health state according to the evolution of their pain score. If the patient did not achieve a level of mild pain after 12 weeks (= treatment failure), he/she was assumed to be non-responsive and to move to second-line therapy. The transition probabilities were calculated using the evolution of the patients in the different studies10–14.

Figure 1. Structure of the Markov model.

The purpose of the model was to define the value for money of adding pregabalin to ‘usual care’ (being defined as a mix of drug therapies). Therefore, the cohort was simulated following two treatment arms: either treated with ‘usual care only or pregabalin as add-on to ‘usual care’. The final outcome of the model was the incremental cost per quality-adjusted life-year (QALY) gained when adding pregabalin to standard care.

Sources for clinical data

The 2008 model developed by Annemans had used efficacy data from a study that included patients with PHN Van Seventer et al.10. In that trial, double-blind study medication as pregabalin capsules, or matching placebo was administered orally (BID) for 12 weeks following a 1-week baseline phase. Patients were randomized to one of four treatment groups: placebo, 150, 300, or 300/600 mg/day pregabalin. Patients with a creatinine clearance rate (CLcr >30 and ≤60 mL/min (‘low’ CLcr) randomized to the 300/600 mg/day group received 300 mg/day; whereas, patients with a CLcr >60 mL/min (‘normal’ CLcr) randomized to the 300/600 mg/day group received 600 mg/day pregabalin. The concurrent use of different types of analgesics (narcotics, tricyclic antidepressants, SSRIs, non-opioid analgesics and NSAIDs) was allowed. Other anticonvulsants were not allowed. That study was considered as the reference study and used as first source of clinical efficacy data for the model.

Four recent clinical studies were identified as additional possible sources. Despite some differences in design and in populations, all the studies had in common that they evaluated pregabalin as an add-on to the standard treatment of neuropathic pain. However, one study (Xochilcal-Morales et al.)11 was excluded as it was not comparative. A second study (Gilron et al.)12 was rejected due to a different design: it consisted of a 4-week single-blind treatment phase, where pregabalin was administered to the whole sample (n = 256), followed by a 5-week double-blind phase where 77 of the remaining 157 patients were switched to placebo, while 80 remained on pregabalin treatment. The primary endpoint was the average pain score at the end of a 5-week double-blind period, so the study did not provide any baseline scores and the final pain scores were not comparable with the other study, since all patients benefited from the effect of pregabalin during the single-blind period.

The two remaining studies (Moon et al.13; Van Seventer et al.14) had a design comparable to the Van Seventer study, thus clinical data from these trials could be used in the model. The study by Moon et al.13 followed a population of 241 Korean patients with a diagnosis of PDPN, PHN or post-traumatic neuralgia (PTN). This was a phase III, flexible-dose, 10-week, randomized, double-blind, placebo-controlled study with four phases: a 1-week screening period, a 4-week titration phase, a 4-week dose maintenance phase and a 1-week taper phase. Patients were randomized to one of five treatment groups: placebo, 150, 300, 450 or 600 mg/day pregabalin, with an average dose of 480 mg. A total 55% of the recruited patients had tried and failed to respond to gabapentin before study. Of the 240 patients who received study medication (either pregabalin or placebo, both with allowed co-medication), 200 completed the double-blind treatment phase. Even though the primary end-point was the difference in the week 8 least squares mean pain score between pregabalin and placebo, calculated using the average of the last seven available pain scores, clinical data comparable to those inputted in the first model (i.e., absolute mean pain score at baseline, week 4 and week 8 in both the placebo group and the ‘all dosages’ pregabalin group) could be obtained from the clinical data report (data on file). As an alternate scenario, the model was also run using the primary endpoint, since the study was powered for this.

In the flexible-dose study published by Van Seventer et al.14, 254 patients with PTN pain were included. In all, 127 patients received pregabalin, and 127 patients received placebo. The trial consisted in a 4-week titration period followed by 4-week dose maintenance period. Doses of pregabalin used were 150 mg (30% at week 4 or the end of the titration period, 300 mg (46% at week 4) and 600 mg (24% at week 4), at a variable regimen. Medication was given in a BID schedule. The primary efficacy variable was the mean pain score in the intent-to-treat population at week 8, based on the last seven recorded daily pain scores, both in the placebo group and in the aggregated pregabalin group. The clinical data needed for the cost-effectiveness model, i.e. the average pain scores in both treatment arms at week 4 and week 8 among the patients still under treatment at that time, were obtained from the clinical study report (data on file). The characteristics of the populations from the three studies used in the model are shown in Table 1.

Table 1. Baseline characteristics from the three clinical trials retained as clinical source for the cost-effectiveness model.

Population size	Van Seventer et al.10		Moon et al.13		Van Seventer et al.14
	Placebo	Pregabalin	Placebo	Pregabalin	Placebo	Pregabalin
	n = 93	n = 368	n = 78	n = 162	n = 127	n = 127
% male (n male)	43.0 (40)	45.7 (168)	41.0 (32)	48.8 (79)	59.1 (75)	39.4 (50)
Age (years)	70.9	70.7	61.3	59.7	51	52
Weight (kg)	73.03	72.96	62.0	63.6	81	78
Duration of NeP (months)	43.4 (SD 44.8)	40.7 (SD 45.3)	–	–	–	–
Baseline pain score (numeric scale)	6.85 (SD 1.49)	6.67 (SD 1.48)	6.31 (SD 1.45)	6.28 (SD 1.51)	6.27 (SD 1.65)	5.96 (SD 1.58)

†Mean (SD).

Table 2 shows the average pain scores collected from the three clinical trials and used to generate sets of transition probabilities.

Table 2. Clinical efficacy variables from the three trials.

	Van Seventer et al.10		Moon et al.13		Moon et al.13/ primary endpoint†		Van Seventer et al.14
	n*	Mean	n	Mean	n	Mean	n	Mean
Baseline
Placebo	93	6.85 (SD 1.53)	78	6.31 (SD 1.45)	78	6.31 (SD 1.45)	126	6.27 (SD 1.65)
PGB 150 mg	87	6.44 (SD 1.58)	–	–	–	–	–	–
PGB 300 mg	98	6.72 (SD 0.38)	–	–	–	–	–	–
PGB 300/600 mg	90	6.65 (SD 1.42)	–	–	–	–	–	–
PGB all doses	–	–	162	6.28 (SD 1.51)	162	6.28 (SD 1.51)	126	5.96 (SD 1.58)
Week 4
Placebo	83	6.15 (SD 1.81)	68	5.23 (SD 2.01)	–	–	108	5.46 (SD 1.97)
PGB 150 mg	83	5.35 (SD 1.90)	–	–	–	–	–	–
PGB 300 mg	87	4.91 (SD 1.85)	–	–	–	–	–	–
PGB 300/600 mg	75	4.51 (SD 1.81)	–	–	–	–	–	–
PGB all doses	–	–	151	4.81 (SD 1.86)	–	–	108	4.76 (SD 2.07)
Week 8
Placebo	65	5.84 (SD 1.68)	62	4.74 (SD 2.11)	77	5.17 (SD 2.18)	75	5.61 (SD 2.48)
PGB 150 mg	66	4.91 (SD 1.69)	–	–	–	–	–	–
PGB 300 mg	68	4.81 (SD 1.72)	–	–	–	–	–	–
PGB 300/600 mg	64	4.17 (SD 1.75)	–	–	–	–	–	–
PGB all doses	–	–	138	4.41 (SD 2.02)	161	4.61 (SD 2.12)	–	4.44 (SD 2.03)
Week 12
Placebo	60	6.01 (SD 1.61)	–	–	–	–	–	–
PGB 150 mg	61	4.89 (SD 1.70)	–	–	–	–	–	–
PGB 300 mg	62	4.94 (SD 1.64)	–	–	–	–	–	–
PGB 300/600 mg	61	4.08 (SD 1.70)	–	–	–	–	–	–

*n = population still under treatment.

†Average of the last 7 recorded pain scores (at any time-point); this includes patients discontinuing before week 8.

Calculation of transition probabilities from the clinical efficacy variables

Starting from the absolute pain scores in the pregabalin and in the comparator arm at various time-points (every 4 weeks from baseline on, in order to match the cycles used in the model), a Monte Carlo16 simulation was run in order to calculate the proportion of patients being in each possible health state at each 4-weekly time-point (starting from baseline on) and for every dosage for which pain scores were available from the clinical studies. The simulation was based on the assumption that pain scores at each time-point fitted to a normal distribution, with the mean and the standard-deviation respectively equal to the average pain score and the standard deviation observed in the study sample at that time-point. An additional constraint was that no pain score could be lower than 4 at baseline (eligibility criterion in all the clinical studies). Finally, drop-off rates, i.e. the proportions of patients withdrawing from the study at each cycle, were also taken from the clinical studies and applied homogenously across health states.

Transition probabilities were derived from the proportions obtained from the Monte Carlo simulations assuming that patients would only transit between adjacent health states (i.e., no transition from mild to severe or from severe to mild). Transition probabilities between weeks 8 and 12 were assumed to be identical to the transitions between weeks 4 and 8 in the Moon et al.13 and Van Seventer et al.14 versions of the model (respectively, Model 2 and Model 3). In all three models, it was assumed that patients still in the moderate or severe health states after week 12 would be switched to a subsequent treatment line. A maximum effect should have been reached by then, unless the patient was non-responsive. Patients in the mild health state would remain on therapy, but the drop-off rates observed between week 8 and week 12 were applied until the end of the simulation horizon. Similarly, patients in the placebo arm pursue their ‘usual care’ but drop-off rates were also applied throughout the model. Table 3 shows the transition probabilities applied in the three versions of the model.

Table 3. Transition probabilities in the model taking into account drop-offs.

	Moderate-to-mild	Severe-to-moderate	Mild-to-moderate	Moderate-to-severe	Drop-off
Model 1 (Van Seventer et al.,10)
Weeks 0–4
Placebo	0.1892	0.2546	0.0000	0.0000	0.1075
PGB 150 mg	0.3666	0.4448	0.0000	0.0000	0.0460
PGB 300 mg	0.4774	0.6142	0.0000	0.0000	0.1122
PGB 600 mg	0.5528	0.6696	0.0000	0.0000	0.1667
PGB all doses	0.4670	0.5763	0.0000	0.0000	0.1091
Weeks 4–8
Placebo	0.0307	0.1812	0.0000	0.0000	0.2169
PGB 150 mg	0.0772	0.3801	0.0000	0.0000	0.2048
PGB 300 mg	0.0098	0.1977	0.0000	0.0000	0.2184
PGB 600 mg	0.1136	0.3665	0.0000	0.0000	0.1467
PGB all doses	0.0644	0.2516	0.0000	0.0000	0.1918
Weeks 8–12 and subsequent
Placebo	0.0000	0.0000	0.2333	0.0346	0.0769
PGB 150 mg	0.0085	0.0046	0.0000	0.0000	0.0758
PGB 300 mg	0.0000	0.0000	0.1112	0.0071	0.0882
PGB 600 mg	0.0412	0.1285	0.0000	0.0000	0.0469
PGB all doses	0.0161	0.0692	0.0382	0.0024	0.0707
Model 2 (Moon et al.,13)
Weeks 0–4
Placebo	0.3393	0.3213	0.0000	0.0000	0.1282
PGB all doses	0.4548	0.5724	0.0000	0.0000	0.0679
Weeks 4–8 and subsequent
Placebo	0.1702	0.2177	0.0000	0.0000	0.0882
PGB all doses	0.1474	0.1550	0.0000	0.0000	0.0861
Model 2 (Moon et al.,13); based on study primary endpoint
Weeks 0–4*
Placebo	0.2206	0.1852	0.0000	0.0000	0.0064
PGB all doses	0.2903	0.2932	0.0000	0.0000	0.0031
Weeks 4–8 and subsequent
Placebo	0.2527	0.2276	0.0000	0.0000	0.0064
PGB all doses	0.3412	0.4153	0.0000	0.0000	0.0031
Model 3 (Van Seventer et al.,14)
Weeks 0–4
Placebo	0.3105	0.3034	0.0000	0.0000	0.1429
PGB all doses	0.4245	0.4134	0.0000	0.0000	0.1429
Weeks 4–8 and subsequent
Placebo	0.0422	0.0000	0.0000	0.0907	0.3056
PGB all doses	0.0803	0.2106	0.0000	0.0000	0.2778

*Pain score missing at week 4. Probabilities were calculated assuming a linear progression between baseline and week 8.

Cost input

The resource use corresponding to the mild, moderate and severe health states were obtained from a chart review conducted in a sample of 88 Belgian patients with peripheral neuropathic pain recruited during a standard follow-up visit, with a neurologist, anesthesiologist or physiotherapy specialist6. In order to serve as a possible benchmark against which a treatment with pregabalin could be compared, patients in that chart review were not allowed to be treated with other antiepileptics. Originally, 97 patients with peripheral NeP were included, of whom nine were taking other antiepileptics and were excluded from analysis. The study period was limited to 1 month as from study inclusion, during which participating patients had been reporting resource consumption on a daily basis in a diary. The monthly costs corresponding to the different health states in the cost-effectiveness model were obtained by multiplying the resources from the chart review with 2011 unit costs as available from public sources (Table 4).

Table 4. Cost per pain level in the observational one month study (unit costs 2011).

Pain intensity			Mean cost (€)
1 (Mild)	Mean		261.49 (SE 30.30)
	95% confidence interval for mean	Lower bound	€202.10
		Upper bound	€320.89
2 (Moderate)	Mean		606.35 (SE 136.12)
	95% confidence interval for mean	Lower bound	€339.59
		Upper bound	€873.15
3 (Severe)	Mean		714.49 (SE 195.70)
	95% confidence interval for mean	Lower bound	€330.91
		Upper bound	€1098.07

Patients withdrawing from treatment or switching to second-line therapy were assigned the cost corresponding to the severe health state.

The acquisition cost of pregabalin depends on the dose. Table 5 shows the acquisition costs of the four different packages that are currently available on the Belgian market, from both the public and the INAMI perspective. Two packages are available for the 150-mg and the 300-mg doses (per 56 units and per 200 units); according to figures from the Belgian Pharmaco-Epidemiology Institute (IPhEB; data March 2011), for both dosages, the market shares are 55% and 45%, respectively for the 56 unit and 200-unit packages. These proportions were taken into account when calculating the acquisition cost of pregabalin in the model (Table 6). Since pregabalin is administered on a bi-daily regimen and caps are not divisible, patients receiving 150 mg are actually administered 75 mg BID. Similarly, patients receiving 300 mg are administered 150 mg BID and patients receiving 600 mg are administered 300 mg BID. When calculating the drug cost in Model 1, the fact that 28.9% of the patients in the 600 mg group actually use 300 mg due to their low CLcr level (30–60 mL/min) was taken into account8.

Table 5. Cost of available packages of pregabalin (€).

Package	Public price (€)	INAMI General (€)
56 × 75 mg	49.89	38.89
56 × 150 mg	84.28	73.28
200 × 150 mg	241.92	228.42
56 × 300 mg	124.77	113.77
200 × 300 mg	357.60	344.10

Table 6. Monthly and daily cost of pregabalin at various dosages.

Dosage	Public price per month (€)	Public price per day (€)	INAMI price per month(€)	INAMI price per day (€)
75 mg	24.95	0.89	19.45	0.69
150 mg	49.89	1.78	38.89	1.39
225 mg	63.36	2.26	55.51	1.98
300 mg	76.84	2.74	69.08	2.47
450 mg	91.98	3.28	84.23	3.01
600 mg	113.68	4.06	105.93	3.78

In Model 1, the cost applied to the ‘all doses’ mix was calculated based on real-life proportions per dosage as provided by the Belgian Pharmaco-Epidemiology Institute, i.e. 36.5% of patients on 150 mg, 44.5% of patients on 300 mg and 19% of patients on 600 mg (May 2011). For Models 2 and 3, since the clinical outcomes were only reported at the aggregated level, the cost applied to the ‘all doses’ mix was calculated using the proportions of patients on each dosage at baseline in the corresponding study. In Moon et al.2, the distribution after the titration phase was: 6.2% of patients on 150 mg, 21.9% on 300 mg, 16.4% on 450 mg, and 55.5% on 600 mg. In Van Seventer et al.14, 30.2% of the patients were on 150 mg, 46.0% were on 300 mg and 23.8% were on 600 mg.

Utility input

The utility values associated to the different health states in the model (Table 7) were drawn from the same source as the health states maintenance costs5. In the Belgian cross-sectional study described above, patients were also asked to fill in a SF-36 questionnaire. Items from the questionnaires were converted into utilities using the SF-6D algorithm17, a generic six-dimensional classification system for health states. Utilities are valued on a scale between 0 and 1, where 0 is the value for death and 1 the value for perfect health. The utility values decrease with an increasing pain severity, demonstrating the major impact of the pain on the NeP patients’ daily life.

Table 7. Utility values in the cross-sectional assessment of current therapy, per level of pain severity.

Pain intensity			Statistic
Mild	Mean		0.573 (SE 0.025)
	95% confidence interval for mean	Lower bound	0.520
		Upper bound	0.625
Moderate	Mean		0.566 (SE 0.014)
	95% confidence interval for mean	Lower bound	0.538
		Upper bound	0.594
Severe	Mean		0.481 (SE 0.025)
	95% confidence interval for mean	Lower bound	0.429
		Upper bound	0.534

Model endpoint

By applying the transition probabilities, the costs per health state and the utility per health state as described above, the three versions of the model calculated the cumulative costs and utility over the full-year period (13 cycles) in both treatment arms and the incremental cost-effectiveness ratio (ICER, in €/QALY gained). All three models provided results for the aggregated pregabalin arm (all doses together); additionally, Model 1 calculated results per dose.

Sensitivity analysis

The key variables in the analysis were the cost and the utility per health state. It was observed that the average monthly maintenance cost increased with pain intensity, the more severe health states being clearly associated with higher maintenance costs.

As guidelines recently recommend18, a probabilistic sensitivity analysis (PSA) was conducted in order to assess the robustness of the above reported results. In the PSA the cost data per pain level were assumed to follow a triangular distribution and the utility data per pain level were considered as beta-distributed. Clinical efficacy variables were not included in the sensitivity analysis since they were already obtained through a probabilistic method (Monte Carlo simulations).

Results

Results provided by the three versions of the model are shown in Table 8. Both Model 1 and 3 concluded that pregabalin was dominant (i.e. both cost savings and health benefit were obtained), confirming the conclusions from previous cost-effectiveness models (Annemans et al., 2008), and so did the version of Model 2 based on primary endpoints.

Table 8. Cost-effectiveness results.

	Total cost (€)	Incremental cost (€)	Utility	Incremental utility	ICER
Model 1
Usual care	8634		0.512
PGB 150 mg	8085	−549	0.526	0.014	Dominant
PGB 300 mg	8483	−151	0.520	0.008	Dominant
PGB 600 mg	7763	−870	0.532	0.020	Dominant
PGB all doses	8123	−511	0.526	0.013	Dominant
Model 2
Usual care	7216		0.528
PGB all doses	7501	285	0.532	0.004	65,958
Model 2 (alternate; with primary endpoints)
Usual care	5541		0.557
PGB all doses	5521	−20	0.564	0.006	Dominant
Model 3
Usual care	8738		0.498
PGB all doses	8647	−92	0.505	0.007	Dominant

Figure 2 shows the results of the probabilistic sensitivity analysis applied respectively to Model 1, Model 2 and Model 3, considering all doses of pregabalin together (pregabalin mix). In each graph, the 1000 points represent the results obtained with 1000 simulations. Points situated above the horizontal axis correspond to cases were pregabalin was shown to be more costly and points on the right side of the vertical axis correspond to cases were pregabalin provided health gains. Points in the right under quadrant correspond to cases were pregabalin was dominant (i.e., both cost-saving and gaining QALYs).

Figures 2–5 Distribution of the incremental cost effectiveness of add-on pregabalin(all doses) vs. usual care in the four versions of the model.

Model 1 concluded that pregabalin was cost effective in 99.1% of the simulations (using a reference threshold of 32,000 €/QALY), cost-saving in 95.5% of the simulations and dominant in 94.8% of the simulations. Model 2 reached cost effectiveness in 11.7% and 85.7% of the simulations respectively for the base case and the alternate version. The relatively weak result in the base case (as compared to the two other models) was due to the fact that, in the Moon study, the decrease in pain score reported at week 8 in the placebo arm was quite similar to the pregabalin arm, resulting in a very low incremental utility for the pregabalin arm. The average incremental cost of 1 year of treatment with add-on pregabalin, calculated on 1000 simulations, was below €300; however, since the average incremental utility was as low as 0.004, the ICER reached high values. Model 3 concluded that pregabalin was cost effective in 94.6% and dominant in 67.2% of the simulations.

Discussion

This paper presents the methodology and the results of the adaptation of a cost-utility model of pregabalin as an add-on option to the standard treatment of neuropathic pain. The comparator was the current usual care, including a wide range of drugs. The model horizon was 1 year and the results were presented from the Belgian healthcare payer’s perspective.

The model structure was taken from a previous cost-effectiveness model published by Annemans et al.8. The clinical efficacy data inputted in the model were successively obtained from three recent clinical trials where pregabalin was compared with current usual care. Three versions of the model were adapted, corresponding to the three different sets of clinical data available. The three studies covered a wide scope of indications commonly defined as NeP: PHN, PDPN and PTN (including postsurgical neuralgia). The cost data were obtained from a sample of 88 Belgian PNeP patients who were followed during 1 month and were not treated with AEDs at the time of the observation, in order to match the characteristics of the population observed in the clinical trial described by Van Seventer et al.10

There is currently no official guideline in Belgium regarding thresholds for cost effectiveness but some informal thresholds are often referred to: a threshold of €20,000 is commonly referred to in the Netherlands, while NICE usually considers £20,000–30,000 (≈€24,000–36,000) a ‘range of acceptable cost effectiveness’19. In this study, we used a cost-effectiveness threshold of 32,000 €/QALY, which corresponds to the average GDP per capita in Belgium (as published by the Bureau Federal du Plan).

Two of the three studies showed that at any of the time-points, the percentage of patients in the mild and moderate health states was higher in the pregabalin arm than in the comparator arm. In the one study where the pregabalin cohort was further divided per dose (Van Seventer et al.10), the same observation was made for each dose. Therefore, more patients in the pregabalin arm ended up in health states associated to lower costs and higher utilities. As a consequence, the two models based on clinical data from these two trials (Model 1 and Model 3) concluded that pregabalin was at least cost effective, if not dominant, in patients with PHN, PDPN and PTN pain.

The conclusions drawn from Model 2 were different, despite the fact that the indications were matching those observed in the other studies. This was due to the fact that, in the corresponding clinical trial (Moon et al.13), the clinical benefit of pregabalin was less than what could be observed in other studies; the efficacy observed in the placebo arm was quite similar to the effect observed in the pregabalin arm, thus resulting in a small incremental utility associated to patients in the pregabalin arm. Even though the additional cost in the pregabalin arm after 1 year was less than €300, the extremely low incremental utility resulted in an ICER that was higher than the reference threshold of 32,000 €/QALY. Moreover, several specificities of that clinical trial have to be highlighted. First of all, 55% of the patients had failed to respond to gabapentin before study inclusion; that could indicate that at least a certain proportion of the population might have been less responsive to GABA analogs (the common group for gabapentin and pregabalin) though patients not responding to gabapentin can benefit from pregabalin20. Secondly, co-medications were reported in 92.3% of the placebo patients, versus 79.6% of the pregabalin patients. Finally, the results may have been impacted by cultural factors conditioning attitude towards the reporting of pain scores; the study was conducted in an Asian population (South Korea) and patients in Asia tend to report less serious pain-related health problems than in other regions21.

More generally, comparisons between the three models should be made with caution. The three models were based upon data from three clinical trials in different types of peripheral NeP: PDPN (Moon et al.13), PHN (Van Seventer et al.10 and Moon et al.13) and PTN (Van Seventer et al.14). Whereas PDPN and PHN are associated with injury to the neural tissue, reflecting abnormal functioning of the pain system and considered as ‘pure’ NeP, PTN might have – besides the NeP component – a nociceptive pain component resulting from noxious stimuli and inflammation in otherwise intact tissue22.

Except for the different types of peripheral NeP included and the prohibited use of AEDs in one clinical trial (Van Seventer et al.10), the inclusion and exclusion criteria were similar for the three studies: patients needed to be 18 years of age, suffering from pain for at least 3 months and having a mean daily pain score of 4 or more; patients with abnormal blood cell counts, patients with unstable hepatic, respiratory, hematologic, cardiovascular and psychologic conditions (such as major depression), patients with drug and alcohol abuse were excluded from the studies.

A first limitation to this study lies in the short follow-up time, the longest follow-up available in the three studies being 12 weeks. Since the horizon of the model was a full year, an extrapolation had to be made to obtain transition probabilities from the fourth month onwards. In that extrapolation, patients who were still in the ‘severe’ health state after 3 months were switched to second line while the same transition probabilities as in the third cycle were applied until the end of the simulation for the other health states. A 15-month open-label study of pregabalin in neuropathic pain, with a recurrent ‘drug holiday period’ after every 3 months, demonstrated that the improvement in pain after the first 3-month treatment period was consistent throughout the five 3-month pregabalin treatment periods, demonstrating the sustained effect of pregabalin on pain relief beyond the 3-month time period20.

Another limitation results from the overlaps observed in the confidence intervals associated to the maintenance costs (between the moderate and the severe health states) and to the average utilities (between the mild and the moderate health states). However, there is a significant gain in utility from moving from severe to moderate and a significant reduction of cost when moving from moderate to mild. The combination of these two factors is enough to show the benefit of pregabalin.

The response to treatment varies a lot according to the dose. Therefore, the final ICER is expected to be sensitive to the proportions of patients in each dose group. A second limitation to this study comes from the fact that only one trial allowed an efficacy assessment per dose. In the version of the model based on that trial, the efficacy in the aggregated pregabalin arm (pooled doses) was calculated using proportions based on real-life figures provided by Pfizer (March 2011) REF. In the two other trials, endpoints were only provided at the aggregated level, i.e. pooled doses. In the corresponding models, proportions were taken directly from the studies and did not necessarily reflect the real practice in the Belgian settings. Nevertheless, it is impossible to estimate the result of a re-calculation of the ICER based on real-life proportions. Several other cost-effectiveness analyses of pregabalin versus gabapentin in NeP have been performed. A Canadian cost-effectiveness analysis in DPN and PHN concluded that a treatment with pregabalin was dominant versus a treatment with gabapentin23. A Spanish cost-effectiveness analysis in DPN and PHN demonstrated that pregabalin was cost effective versus gabapentin (€20,535/QALY)24, an Italian cost-effectiveness analysis resulted in an ICER of €468/QALY25. In the HTA analysis performed by NICE, PHN pregabalin 150–600 mg/day was more cost effective than gabapentin 3600 and 1800 mg/day for PHN and pregabalin 300 and 600 mg/day was more cost effective than gabapentin 3600 mg/day in DPN26.

Based on the analysis described here, add-on pregabalin offers a slight, but significant increase in quality of life in the studied populations as compared to standard care from a Belgium healthcare/payer perspective. Pregabalin can be considered cost effective in the majority of cases except in one published clinical study, despite a low incremental cost per year (€300).

Transparency

Declaration of funding

This study was sponsored by Pfizer. Pfizer Ltd provided funding to IMS HEOR Belgium to perform this study and to prepare the manuscript.

Declaration of financial/other relationship

Pierre Chevalier and Marc Lamotte were employees of IMS HEOR Belgium who were paid consultants to Pfizer Ltd in connection with the development of this manuscript. H.v.C and R.E. were employees of Pfizer Belgium at the time of the study. L.A. is an employee of Ghent and Brussels University who received financial support from Pfizer Ltd in connection with the development of this manuscript and for his contribution to this work. JME peer reviewers on this manuscript have no relevant financial relationships to disclose.