NCBI Bookshelf. A service of the National Library of Medicine, National Institutes of Health.

Centre for Clinical Practice at NICE (UK). Neuropathic Pain: The Pharmacological Management of Neuropathic Pain in Adults in Non-Specialist Settings. London: National Institute for Health and Clinical Excellence (UK); 2010 Mar. (NICE Clinical Guidelines, No. 96.)

  • This publication is provided for historical reference only and the information may be out of date.

This publication is provided for historical reference only and the information may be out of date.

Cover of Neuropathic Pain

Neuropathic Pain: The Pharmacological Management of Neuropathic Pain in Adults in Non-Specialist Settings.

Show details

2How this guideline was developed

‘Neuropathic pain: the pharmacological management of neuropathic pain in adults in non-specialist settings’ (NICE clinical guideline 96) is a NICE short clinical guideline. For a full explanation of how this type of guideline is developed, see ‘The guidelines manual’ (2009) at www.nice.org.uk/GuidelinesManual

2.1. Introduction

2.1.1. Pharmacological treatments, key outcomes and analysis

Based on the guideline scope, neuropathic pain is treated as a ‘blanket condition’ in this guideline regardless of its aetiologies, unless there is valid and robust clinical and health economics evidence that shows the clinical efficacy and cost effectiveness of a particular treatment for a specific neuropathic pain condition.

It was agreed during the scoping workshop and consultation on the scope, and by the Guideline Development Group (GDG), to consider 34 different pharmacological treatments for neuropathic pain within the four main drug classes (antidepressants, anti-epileptics, opioid analgesics and topical treatments). These are listed in table 2. The different neuropathic pain conditions that were included in the searches are listed in table 3. Systematic literature searches were carried out to identify randomised placebo-controlled trials on these 34 different pharmacological treatments for neuropathic pain, as well as any head-to-head comparative trials and combination therapy trials.

Table 2. Pharmacological treatments considered for the clinical guideline on neuropathic pain.

Table 2

Pharmacological treatments considered for the clinical guideline on neuropathic pain.

Table 3. Neuropathic pain conditions (search terms) included in the searches.

Table 3

Neuropathic pain conditions (search terms) included in the searches.

A total of 23,207 studies were retrieved by the systematic searches (antidepressants = 2781, anti-epileptics = 4757, opioid analgesics = 9612, topical capsaicin and topical lidocaine = 6057). From the 23,207 studies, 90 randomised placebo-controlled trials, 10 head-to-head comparative trials and four combination therapy trials were included, based on the inclusion and exclusion criteria suggested by the GDG through two short questionnaires13. The searches did not identify any placebo-controlled studies that met the inclusion and exclusion criteria for 15 of the pharmacological treatments (table 4). The 104 included studies are summarised in table 5.

Table 4. Pharmacological treatments for which no studies met the inclusion and exclusion criteria.

Table 4

Pharmacological treatments for which no studies met the inclusion and exclusion criteria.

Table 5. Summary of included randomised placebo-controlled trials on antidepressants, anti-epileptics, opioid analgesics and topical treatments, and head-to-head comparative and combination therapy trials, for the treatment of neuropathic pain.

Table 5

Summary of included randomised placebo-controlled trials on antidepressants, anti-epileptics, opioid analgesics and topical treatments, and head-to-head comparative and combination therapy trials, for the treatment of neuropathic pain.

Analysis and synthesis

The primary outcomes for meta-analysis, based on the Initiative on Methods, Measurement, and Pain Assessment in Clinical Trials (IMMPACT) recommendations (Dworkin et al. 2005; Dworkin et al. 2008), were: at least 30% pain reduction; at least 50% pain reduction; patient-reported global improvement; and adverse effects. Specific adverse effects for each drug class were selected by the GDG (see appendix 9.3A), based on the expert knowledge and experience of GDG members (including that of patient and carer members). A fixed-effects model meta-analysis by subclass of the pharmacological treatment (for example, antidepressants: TCAs, SSRIs, SNRIs) or by individual drug of the pharmacological treatment (for example, anti-epileptics: pregabalin, gabapentin, oxcarbazepine, lamotrigine, carbamazepine, phenytoin, sodium valproate, topiramate) was carried out on the primary outcomes. Where there was significant heterogeneity, a random-effects model was adopted for the meta-analysis (for further information on methodology, see the review protocol in appendix 9.2). All results from the meta-analyses (relative risk or risk ratio [RR], absolute risk reduction [ARR], absolute risk increase [ARI], number-needed-to-treat to benefit [NNTB] and number-needed-to-treat to harm [NNTH]) are presented in GRADE profiles (for GRADE methodology, see appendix 9.9) and subsequent evidence statements. No studies were excluded on the basis of outcomes.

For the completeness of the evidence base, included studies that did not report the primary outcomes recommended by the IMMPACT recommendations (at least 30% pain reduction; at least 50% pain reduction; patient-reported global improvement; adverse effects) (Dworkin et al. 2005; Dworkin et al. 2008) were summarised in evidence tables (see appendix 9.9). Pain outcomes (other than the primary outcomes) reported in these studies are presented in GRADE profiles and evidence statements as ‘other reported pain outcomes’. The ‘other reported pain outcomes’ included mean pain relief score, mean pain intensity score, mean change in pain relief score from baseline, mean change in pain intensity score from baseline and mean change in daily pain score. Only evidence on the primary outcomes recommended by the IMMPACT recommendations (at least 30% pain reduction; at least 50% pain reduction; patient-reported global improvement; adverse effects) was used to generate recommendations. However, where evidence on the primary outcomes for particular pharmacological treatments was scarce or limited, evidence from ‘other reported pain outcomes’ was used to assist and generate discussion among the GDG to reach consensus, but not as the sole basis for making recommendations. For included studies that did not report either primary outcomes or ‘other reported pain outcomes’, study characteristics were summarised in the evidence tables for information (see the evidence tables in appendix 9.9 for full information on each included study).

2.1.2. Health economics

No health economic modelling was undertaken for this guideline because there was a relevant health technology assessment (HTA) monograph in development to which the GDG had been given access (Fox-Rushby JA, Griffith GL, Ross JR et al. [2010] The clinical and cost-effectiveness of different treatment pathways for neuropathic pain [NP]. NIHR Health Technology Assessment [HTA] programme, ref. 05/30/03. In press. Project abstract available from www.hta.ac.uk/1527). The GDG reviewed, appraised and summarised the HTA report, and the results of the economic analyses from the HTA report informed this guideline as appropriate.

The HTA report focused on two neuropathic pain populations: people with post-herpetic neuralgia (PHN) and people with painful diabetic neuropathy (PDN). A systematic review of the economic evidence was also performed as part of the evidence review for this guideline. A systematic search found a total of 2273 papers. Full details on the search strategy can be found in appendix 9.7.

For the purposes of this guideline, the GDG decided at the outset that neuropathic pain would be treated as a ‘blanket condition’ where possible or necessary. However, it was clear that the treatment of various subpopulations would differ considerably and that it would not be possible to extrapolate from one subgroup to all people with neuropathic pain. In addition, the GDG decided that the HTA report included thorough data on the cost effectiveness of treatment pathways (sequences) for the subpopulations with PHN and PDN. On this basis, the economic evidence review for this guideline excluded papers on people with PHN or PDN.

2.1.3. Summaries of included studies

2.2. Evidence statements

2.2.1. Antidepressants

(see table 6)

Table 6. Characteristics of included studies: antidepressants (placebo-controlled trials).

Table 6

Characteristics of included studies: antidepressants (placebo-controlled trials).

Primary outcomes

TCAs (as monotherapy – placebo-controlled trials)

For evidence relating to the following evidence statements, see table 11 (GRADE profiles).

For these evidence statements, the TCAs referred to are amitriptyline, nortriptyline, desipramine and imipramine.

Outcomes on pain
  • Patients receiving TCAs were significantly more likely to report at least 30% pain reduction and global improvement compared with patients receiving placebo (moderate-quality evidence).
Adverse effects
  • Patients receiving TCAs were significantly more likely to withdraw from treatment because of adverse effects compared with patients receiving placebo (low-quality evidence).
  • Patients receiving TCAs were significantly more likely to report dry mouth (moderate-quality evidence) and sedation (low-quality evidence) compared with patients receiving placebo.
  • For incidences of blurred vision, dizziness, vomiting and gastrointestinal disturbances, there were no significant differences between patients receiving TCAs and patients receiving placebo (low-quality evidence).
  • Patients receiving TCAs were significantly more likely to report any adverse effects (non-specified) compared with patients receiving placebo (high-quality evidence).
Lofepramine, trimipramine, dothiepin and doxepin (as monotherapy – placebo-controlled trials)
  • No studies on lofepramine, trimipramine, dosulepin (dothiepin) or doxepin met the inclusion and exclusion criteria. Therefore there was no appropriate evidence that lofepramine, trimipramine, dosulepin (dothiepin) or doxepin is clinically effective in treating neuropathic pain.
SSRIs (as monotherapy – placebo-controlled trials)
  • No studies on SSRIs met the inclusion and exclusion criteria. Therefore there was no appropriate evidence that any SSRI is clinically effective in treating neuropathic pain.
SNRIs (as monotherapy – placebo-controlled trials)

For evidence relating to the following evidence statements, see table 12 (GRADE profiles).

For these evidence statements, the SNRIs referred to are duloxetine and venlafaxine.

Outcomes on pain
  • Patients receiving SNRIs were significantly more likely to report at least 30% pain reduction (duloxetine) and at least 50% pain reduction (duloxetine and venlafaxine) (moderate-to-high-quality evidence).
  • The number of patients reporting global improvement was not significantly different between patients receiving venlafaxine and patients receiving placebo (moderate-quality evidence).
Adverse effects
  • Patients receiving SNRIs were significantly more likely to withdraw from treatment because of adverse effects compared with patients receiving placebo (moderate-quality evidence).
  • For incidences of dry mouth and gastrointestinal disturbances, there were no significant differences between patients receiving SNRIs and patients receiving placebo (low-quality evidence).
  • For incidences of blurred vision and vomiting, there were no significant differences between patients receiving SNRIs and patients receiving placebo (very-low-quality evidence).
  • For the incidence of any adverse effects (non-specified), there was no significant difference between patients receiving SNRIs and patients receiving placebo (very-low-quality evidence).

Other reported pain outcomes

For evidence relating to the following evidence statements, see table 13 (GRADE profiles).

For mean pain intensity scores:

  • There was conflicting low-quality evidence on the efficacy of amitriptyline in reducing pain intensity scores.
  • There was no significant difference in pain intensity scores between patients receiving venlafaxine and patients receiving placebo (low-quality evidence).

For mean pain relief scores:

  • There was no significant difference in pain relief scores between patients receiving amitriptyline and patients receiving placebo (low-quality evidence).

2.2.2. Anti-epileptics

(see table 7)

Table 7. Characteristics of included studies: anti-epileptics (placebo-controlled trials).

Table 7

Characteristics of included studies: anti-epileptics (placebo-controlled trials).

Primary outcomes

Gabapentin (as monotherapy – placebo-controlled trials)

For evidence relating to the following evidence statements, see table 14 (GRADE profiles).

Outcomes on pain
  • Patients receiving gabapentin were significantly more likely to report at least 50% pain reduction and global improvement compared with patients receiving placebo (moderate-to-high-quality evidence).
  • The number of patients reporting at least 30% pain reduction was not significantly different between patients receiving gabapentin and patients receiving placebo (moderate-quality evidence).
Adverse effects
  • Patients receiving gabapentin were significantly more likely to withdraw from treatment because of adverse effects compared with patients receiving placebo (moderate-quality evidence).
  • Patients receiving gabapentin were significantly more likely to report dizziness, somnolence (moderate-quality evidence) and fatigue (low-quality evidence) compared with patients receiving placebo.
  • For incidences of sedation and gait disturbances, there were no significant differences between patients receiving gabapentin and patients receiving placebo (very-low-quality evidence).
  • Patients receiving gabapentin were significantly more likely to report any adverse effects (non-specified) compared with patients receiving placebo (high-quality evidence).
Pregabalin (as monotherapy – placebo-controlled trials)

For evidence relating to the following evidence statements, see table 15 (GRADE profiles).

Outcomes on pain
  • Patients receiving pregabalin were significantly more likely to report at least 30% pain reduction, at least 50% pain reduction and global improvement compared with patients receiving placebo (high-quality evidence).
Adverse effects
  • Patients receiving pregabalin were significantly more likely to withdraw from treatment because of adverse effects compared with patients receiving placebo (high-quality evidence).
  • Patients receiving pregabalin were significantly more likely to report dizziness, somnolence (high-quality evidence), weight gain and gait disturbances (low-quality evidence) compared with patients receiving placebo.
  • For the incidence of fatigue, there was no significant difference between patients receiving pregabalin and patients receiving placebo (very-low-quality evidence).
  • Patients receiving pregabalin were significantly more likely to report any adverse effects (non-specified) compared with patients receiving placebo (moderate-quality evidence).
Lamotrigine (as monotherapy – placebo-controlled trials)

For evidence relating to the following evidence statements, see table 16 (GRADE profiles).

Outcomes on pain
  • The numbers of patients reporting at least 30% pain reduction and at least 50% pain reduction were not significantly different between patients receiving lamotrigine and patients receiving placebo (moderate-quality evidence).
  • Patients receiving lamotrigine were significantly more likely to report global improvement compared with patients receiving placebo (moderate-quality evidence).
Adverse effects
  • Patients receiving lamotrigine were significantly more likely to withdraw from treatment because of adverse effects compared with patients receiving placebo (moderate-quality evidence).
  • For incidences of dizziness, fatigue (low-quality evidence) and sedation (very-low-quality evidence), there were no significant differences between patients receiving lamotrigine and patients receiving placebo.
  • For the incidence of any adverse effects (non-specified), there was no significant difference between patients receiving lamotrigine and patients receiving placebo (high-quality evidence).
Oxcarbazepine (as monotherapy – placebo-controlled trials)

For evidence relating to the following evidence statements, see table 17 (GRADE profiles).

Outcomes on pain
  • Patients receiving oxcarbazepine were significantly more likely to report at least 30% pain reduction and at least 50% pain reduction compared with patients receiving placebo (moderate-quality evidence).
  • The number of patients reporting global improvement was not significantly different between patients receiving oxcarbazepine and patients receiving placebo (moderate-quality evidence).
Adverse effects
  • Patients receiving oxcarbazepine were significantly more likely to withdraw from treatment because of adverse effects compared with patients receiving placebo (moderate-quality evidence).
  • Patients receiving oxcarbazepine were significantly more likely to report dizziness and somnolence compared with patients receiving placebo (low-quality evidence).
  • For the incidence of fatigue, there was no significant difference between patients receiving oxcarbazepine and patients receiving placebo (low-quality evidence).
Topiramate (as monotherapy – placebo-controlled trials)

For evidence relating to the following evidence statements, see table 18 (GRADE profiles).

Outcomes on pain
  • Patients receiving topiramate were significantly more likely to report at least 30% pain reduction, at least 50% pain reduction and global improvement compared with patients receiving placebo (moderate-quality evidence).
Adverse effects
  • Patients receiving topiramate were significantly more likely to withdraw from treatment because of adverse effects compared with patients receiving placebo (high-quality evidence).
  • Patients receiving topiramate were significantly more likely to report somnolence, fatigue (moderate-quality evidence) and sedation (very-low-quality evidence) compared with patients receiving placebo.
  • For the incidence of dizziness, there was no significant difference between patients receiving topiramate and patients receiving placebo (very-low-quality evidence).
Carbamazepine (as monotherapy – placebo-controlled trials)

For evidence relating to the following evidence statements, see table 19 (GRADE profiles).

Outcomes on pain
  • The number of patients reporting global improvement was not significantly different between patients receiving carbamazepine and patients receiving placebo (moderate-quality evidence).
Adverse effects
  • Patients receiving carbamazepine were significantly more likely to report any adverse effects (non-specified) compared with patients receiving placebo (very-low-quality evidence).
Sodium valproate (as monotherapy – placebo-controlled trials)

For evidence relating to the following evidence statements, see table 20 (GRADE profiles).

Outcomes on pain
  • No study on sodium valproate that reported the primary outcomes on pain met the inclusion and exclusion criteria.
Adverse effects
  • For the incidence of withdrawal from treatment because of adverse effects, there was no significant difference between patients receiving sodium valproate and patients receiving placebo (low-quality evidence).
  • For the incidence of any adverse effects (non-specified), there was no significant difference between patients receiving sodium valproate and patients receiving placebo (high-quality evidence).
Phenytoin
  • No study on phenytoin met the inclusion and exclusion criteria. Therefore there was no appropriate evidence that phenytoin is clinically effective in treating neuropathic pain.

Other reported pain outcomes

For evidence relating to the following evidence statements, see table 21 (GRADE profiles).

For sodium valproate:

  • There was conflicting low-quality evidence on the efficacy of sodium valproate in relation to pain intensity scores and pain relief scores.

For pregabalin:

  • Pain intensity scores for patients receiving pregabalin were significantly lower than those for patients receiving placebo (low-quality evidence).

For gabapentin:

  • The mean change in pain intensity score from baseline was significantly greater for patients receiving gabapentin than for patients receiving placebo (low-quality evidence).

For oxcarbazepine:

  • There was no significant difference in pain relief scores between patients receiving oxcarbazepine and patients receiving placebo (low-quality evidence).

2.2.3. Opioids

(see table 8)

Table 8. Characteristics of included studies: opioid analgesics (placebo-controlled trials).

Table 8

Characteristics of included studies: opioid analgesics (placebo-controlled trials).

Primary outcomes

Morphine (as monotherapy – placebo-controlled trials)

For evidence relating to the following evidence statements, see table 22 (GRADE profiles).

Outcomes on pain
  • Patients receiving morphine were significantly more likely to report at least 30% pain reduction and at least 50% pain reduction compared with patients receiving placebo (moderate-quality evidence).
  • The number of patients reporting global improvement was not significantly different between patients receiving morphine and patients receiving placebo (moderate-quality evidence).
Adverse effects
  • Patients receiving morphine were significantly more likely to withdraw from treatment because of adverse effects compared with patients receiving placebo (very-low-quality evidence).
  • Patients receiving morphine were significantly more likely to report constipation and somnolence/drowsiness compared with patients receiving placebo (low-quality evidence).
  • For incidences of nausea and dizziness, there were no significant differences between patients receiving morphine and patients receiving placebo (low-quality evidence).
Tramadol (as monotherapy – placebo-controlled trials)

For evidence relating to the following evidence statements, see table 23 (GRADE profiles).

Outcomes on pain
  • Patients receiving tramadol were significantly more likely to report at least 50% pain reduction compared with patients receiving placebo (moderate-quality evidence).
Adverse effects
  • Patients receiving tramadol were significantly more likely to withdraw from treatment because of adverse effects compared with patients receiving placebo (low-quality evidence).
  • Patients receiving tramadol were significantly more likely to report constipation, nausea and dizziness compared with patients receiving placebo (low-quality evidence).
  • For incidences of somnolence/drowsiness and vomiting, there were no significant differences between patients receiving tramadol and patients receiving placebo (very-low-quality evidence).
Oxycodone (as monotherapy – placebo-controlled trials)

For evidence relating to the following evidence statements, see table 24 (GRADE profiles).

Outcomes on pain
  • No studies on oxycodone that reported the primary outcomes on pain met the inclusion and exclusion criteria.
Adverse effects
  • For the incidence of withdrawal from treatment because of adverse effects, there was no significant difference between patients receiving oxycodone and patients receiving placebo (low-quality evidence).
  • Patients receiving oxycodone were significantly more likely to report somnolence/drowsiness, nausea, dizziness and vomiting compared with patients receiving placebo (very-low-quality evidence).
Co-codamol, co-dydramol, dihydrocodeine, buprenorphine, fentanyl and codeine phosphate (as monotherapy – placebo-controlled trials)
  • No studies on co-codamol, co-dydramol, dihydrocodeine, buprenorphine, fentanyl or codeine phosphate met the inclusion and exclusion criteria. Therefore there was no appropriate evidence that co-codamol, co-dydramol, dihydrocodeine, buprenorphine, fentanyl or codeine phosphate is clinically effective in treating neuropathic pain.

Other reported pain outcomes

For evidence relating to the following evidence statements, see table 25 (GRADE profiles).

For tramadol:

  • Pain intensity scores and pain relief scores for patients receiving tramadol were significantly lower than those for patients receiving placebo (low-quality evidence).

For oxycodone:

  • The mean change in pain intensity score from baseline for patients receiving oxycodone was significantly greater than that for patients receiving placebo (low-quality evidence).

2.2.4. Topical treatments

(see table 9)

Table 9. Characteristics of included studies: topical capsaicin and topical lidocaine (placebo-controlled trials).

Table 9

Characteristics of included studies: topical capsaicin and topical lidocaine (placebo-controlled trials).

Primary outcomes

Topical capsaicin (as monotherapy – placebo-controlled trials)

For evidence relating to the following evidence statements, see table 26 (GRADE profiles).

Outcomes on pain
  • The numbers of patients reporting at least 40% pain reduction, at least 50% pain reduction and global improvement were not significantly different between patients receiving topical capsaicin and patients receiving placebo (moderate-quality evidence).
Adverse effects
  • Patients receiving topical capsaicin were significantly more likely to withdraw from treatment because of adverse effects compared with patients receiving placebo (low-quality evidence).
  • Patients receiving topical capsaicin were significantly more likely to report a burning sensation compared with patients receiving placebo (high-quality evidence).
  • For the incidence of skin irritation, there was no significant difference between patients receiving topical capsaicin and patients receiving placebo (very-low-quality evidence).
Topical lidocaine (as monotherapy – placebo-controlled trials)

For evidence relating to the following evidence statements, see table 27 (GRADE profiles).

Outcomes on pain
  • No studies on topical lidocaine were identified that reported the primary outcomes on pain.
Adverse effects
  • For the incidence of withdrawal from treatment because of adverse effects, there was no significant difference between patients receiving topical lidocaine and patients receiving placebo (low-quality evidence).
  • For incidences of rash and skin irritation, there were no significant differences between patients receiving topical lidocaine and patients receiving placebo (very-low-quality evidence).

Other reported pain outcomes

Topical capsaicin (as monotherapy – placebo-controlled trials)

For evidence relating to the following evidence statements, see table 28 (GRADE profiles).

  • There was conflicting low-quality evidence on the efficacy of topical capsaicin in reducing mean pain intensity scores.
  • The mean change in pain intensity score from baseline was significantly greater for patients receiving topical capsaicin than for patients receiving placebo (low-quality evidence).
Topical lidocaine (as monotherapy – placebo-controlled trials)

For evidence relating to the following evidence statements, see table 29 (GRADE profiles).

  • There were no significant differences in pain intensity scores and pain relief scores between patients receiving topical lidocaine and patients receiving placebo (low-quality evidence).
  • There was conflicting low-quality evidence on the efficacy of topical lidocaine in reducing pain intensity scores from baseline.
  • The mean change in pain relief score from baseline was significantly greater for patients receiving topical lidocaine than for patients receiving placebo (low-quality evidence).

2.2.5. Comparative trials and combination therapy

(see table 10)

Table 10. Characteristics of included studies: comparative trials and combination therapy (randomised controlled trials).

Table 10

Characteristics of included studies: comparative trials and combination therapy (randomised controlled trials).

Cross-class comparative trials

Amitriptyline (TCA) compared with gabapentin (anti-epileptic)
Primary outcomes

For evidence relating to the following evidence statements, see table 30 (GRADE profiles).

Outcomes on pain
  • Patients receiving amitriptyline were significantly more likely to report at least 30% pain reduction compared with patients receiving gabapentin (moderate-quality evidence).
  • The number of patients reporting global improvement was not significantly different between patients receiving amitriptyline and patients receiving gabapentin (moderate-quality evidence).
Adverse effects
  • For the incidence of withdrawal from treatment because of adverse effects, there was no significant difference between patients receiving amitriptyline and patients receiving gabapentin (low-quality evidence).
  • For incidences of dry mouth, dizziness, blurred vision, sedation, fatigue and weight gain, there were no significant differences between patients receiving amitriptyline and patients receiving gabapentin (very-low-quality evidence).
  • For the incidence of any adverse effects (non-specified), there was no significant difference between patients receiving amitriptyline and patients receiving gabapentin (very-low-quality evidence).
Other reported pain outcomes

For evidence relating to the following evidence statements, see table 31 (GRADE profiles).

  • The mean change in pain relief score from baseline was significantly greater for patients receiving gabapentin than for patients receiving amitriptyline (very-low-quality evidence).
Nortriptyline (TCA) compared with gabapentin (anti-epileptic)
Primary outcomes

For evidence relating to the following evidence statements, see table 32 (GRADE profiles).

Outcomes on pain
  • The number of patients reporting at least 50% pain reduction was not significantly different between patients receiving nortriptyline and patients receiving gabapentin (moderate-quality evidence).
Adverse effects
  • For incidences of somnolence, dry mouth and fatigue, there were no significant differences between patients receiving nortriptyline and patients receiving gabapentin (very-low-quality evidence).
Amitriptyline (TCA) compared with carbamazepine (anti-epileptic)
Primary outcomes

For evidence relating to the following evidence statements, see table 33 (GRADE profiles).

Outcomes on pain
  • The number of patients reporting global improvement was not significantly different between patients receiving amitriptyline and patients receiving carbamazepine (moderate-quality evidence).
Adverse effects
  • For the incidence of any adverse effects (non-specified), there was no significant difference between patients receiving amitriptyline and patients receiving carbamazepine (very-low-quality evidence).
Pregabalin (anti-epileptic) compared with oxycodone (opioid analgesic)
Primary outcomes

For evidence relating to the following evidence statements, see table 34 (GRADE profiles)

Outcomes on pain
  • No studies comparing pregabalin with oxycodone that reported the primary outcomes on pain met the inclusion and exclusion criteria.
Adverse effects
  • For the incidence of withdrawal from treatment because of adverse effects, there was no significant difference between patients receiving pregabalin and patients receiving oxycodone (very-low-quality evidence).
Pregabalin (anti-epileptic) compared with topical lidocaine
Primary outcomes

For evidence relating to the following evidence statements, see table 35 (GRADE profiles).

Outcomes on pain
  • The numbers of patients reporting at least 30% pain reduction, at least 50% pain reduction and global improvement were not significantly different between patients receiving pregabalin and patients receiving topical lidocaine (very-low-quality evidence).
Adverse effects
  • Patients receiving pregabalin were significantly more likely to withdraw from treatment because of adverse effects compared with patients receiving topical lidocaine (very-low-quality evidence).
  • Patients receiving pregabalin were significantly more likely to report any adverse effects (non-specified) compared with patients receiving topical lidocaine (very-low-quality evidence).
Amitriptyline (TCA) compared with topical capsaicin
Primary outcomes

For evidence relating to the following evidence statements, see table 36 (GRADE profiles).

Outcomes on pain
  • No studies comparing amitriptyline with topical capsaicin that reported the primary outcomes on pain met the inclusion and exclusion criteria.
Adverse effects
  • Patients receiving amitriptyline were significantly more likely to report sedation compared with patients receiving topical capsaicin (very-low-quality evidence).
  • Patients receiving topical capsaicin were significantly more likely to report a burning sensation compared with patients receiving amitriptyline (very-low-quality evidence).
Other reported pain outcomes

For evidence relating to the following evidence statements, see table 37 (GRADE profiles).

  • There were no significant differences in pain relief scores or the mean change in pain intensity score from baseline between patients receiving amitriptyline and patients receiving topical capsaicin (low-quality evidence).

Within-class comparative trials

Imipramine (TCA) compared with venlafaxine (SNRI)
Primary outcomes

For evidence relating to the following evidence statements, see table 38 (GRADE profiles).

Outcomes on pain
  • The number of patients reporting global improvement was not significantly different between patients receiving imipramine and patients receiving venlafaxine (moderate-quality evidence).
Adverse effects
  • For incidences of dizziness, dry mouth, blurred vision and any adverse effects (non-specified), there were no significant differences between patients receiving imipramine and patients receiving venlafaxine (very-low-quality evidence).
Amitriptyline (TCA) compared with nortriptyline (TCA)
Primary outcomes

For evidence relating to the following evidence statements, see table 39 (GRADE profiles).

Outcomes on pain
  • No studies comparing amitriptyline with nortriptyline that reported the primary outcomes on pain met the inclusion and exclusion criteria.
Adverse effects
  • For incidences of dry mouth, dizziness, drowsiness and any adverse effects (non-specified), there were no significant differences between patients receiving amitriptyline and patients receiving nortriptyline (very-low-quality evidence).

Combination therapy

Pregabalin plus oxycodone (combination) compared with pregabalin alone (anti-epileptics)
Primary outcomes

For evidence relating to the following evidence statements, see table 40 (GRADE profiles).

Outcomes on pain
  • No studies comparing pregabalin plus oxycodone with pregabalin alone that reported the primary outcomes on pain met the inclusion and exclusion criteria.
Adverse effects
  • For the incidence of withdrawal from treatment because of adverse effects, there was no significant difference between patients receiving pregabalin plus oxycodone and patients receiving pregabalin alone (very-low-quality evidence).
Gabapentin plus oxycodone (combination) compared with gabapentin alone (anti-epileptics)
Primary outcomes

For evidence relating to the following evidence statements, see table 41 (GRADE profiles).

Outcomes on pain
  • No studies comparing gabapentin plus oxycodone with gabapentin alone that reported the primary outcomes on pain met the inclusion and exclusion criteria.
Adverse effects
  • Patients receiving gabapentin plus oxycodone were significantly more likely to withdraw from treatment because of adverse effects compared with patients receiving gabapentin alone (very-low-quality evidence).
  • Patients receiving gabapentin plus oxycodone were significantly more likely to report constipation, nausea, fatigue, dizziness, somnolence and any adverse effects (non-specified) compared with patients receiving gabapentin alone (very-low-quality evidence).
  • For the incidence of vomiting, there was no significant difference between patients receiving gabapentin plus oxycodone and patients receiving gabapentin alone (very-low-quality evidence).
Other reported pain outcomes

For evidence relating to the following evidence statements, see table 42 (GRADE profiles).

  • The mean change in pain relief score from baseline was significantly greater for patients receiving gabapentin plus oxycodone than for patients receiving gabapentin alone (low-quality evidence).
Pregabalin plus oxycodone (combination) compared with oxycodone alone (opioid analgesic)
Primary outcomes

For evidence relating to the following evidence statements, see table 43 (GRADE profiles).

Outcomes on pain
  • No studies comparing pregabalin plus oxycodone with oxycodone alone that reported the primary outcomes on pain met the inclusion and exclusion criteria.
Adverse effects
  • For the incidence of withdrawal from treatment because of adverse effects, there was no significant difference between patients receiving pregabalin plus oxycodone and patients receiving oxycodone alone (very-low-quality evidence).
Gabapentin plus nortriptyline (combination) compared with gabapentin alone and nortriptyline alone
Other reported pain outcomes

For evidence relating to the following evidence statements, see table 44 (GRADE profiles).

  • The mean change in daily pain score was significantly greater for patients receiving gabapentin plus nortriptyline than for patients receiving gabapentin alone (low-quality evidence).
  • The mean change in daily pain score was significantly greater for patients receiving gabapentin plus nortriptyline than for patients receiving nortriptyline alone (low-quality evidence).

2.2.6. Health economics evidence statements

For patients with painful diabetic neuropathy:

  • One high-quality study provided evidence that duloxetine, especially in dosages of up to 60 mg per day, is the most cost-effective treatment.
  • One high-quality study concluded that amitriptyline is less cost effective than duloxetine, but its cost effectiveness is similar to that of pregabalin at a willingness to pay (WTP) threshold of between £20,000 and £30,000 per quality-adjusted life year (QALY) gained.
  • One high-quality study concluded that pregabalin is less cost effective than duloxetine, but its cost effectiveness is similar to that of amitriptyline at a WTP threshold of between £20,000 and £30,000 per QALY gained.

For patients with painful diabetic neuropathy or post-herpetic neuralgia:

  • There is evidence from one high-quality HTA report that pregabalin is more cost effective than gabapentin.

See section 2.4 for a review of the health economics evidence.

2.3. Clinical evidence reviews

2.3.1. Antidepressants as monotherapy for neuropathic pain

Fifteen antidepressants (nine TCAs, four SSRIs and two SNRIs) were included in this review (see table 2) and a total of 2781 studies were retrieved by the systematic searches. From the 2781 studies, 23 randomised placebo-controlled trials on antidepressants were included, based on the inclusion and exclusion criteria suggested by the GDG through two short questionnaires14. No placebo-controlled studies on lofepramine, trimipramine, dosulepin (dothiepin), doxepin or SSRIs (citalopram, fluoxetine, paroxetine and sertraline) that were identified met the inclusion and exclusion criteria. The characteristics of the 23 included studies are summarised in table 6 (for detailed full evidence tables, see appendix 9.9).

2.3.2. Anti-epileptics as monotherapy for neuropathic pain

Eight anti-epileptics were included in this review (see table 2) and a total of 4757 studies were retrieved by the systematic searches. A total of 46 randomised placebo-controlled trials on anti-epileptics were included from the retrieved 4757 studies, based on the inclusion and exclusion criteria15. None of the placebo-controlled studies identified on phenytoin met the inclusion and exclusion criteria. The characteristics of the 46 included studies are summarised in table 7 (for detailed full evidence tables, see appendix 9.9). Meta-analysis was carried out for individual anti-epileptics (gabapentin, pregabalin, lamotrigine, oxcarbazepine, topiramate, carbamazepine and sodium valproate) for primary outcomes and specific adverse effects. See section 2.1.1 for details of the analysis and synthesis of outcomes.

2.3.3. Opioid analgesics as monotherapy for neuropathic pain

Nine opioid analgesics were included in this review (see table 2). A total of 9612 studies were retrieved by the systematic searches, and eight randomised placebo-controlled trials were included based on the inclusion and exclusion criteria16. None of the placebo-controlled studies identified on co-codamol, co-dydramol, dihydrocodeine, buprenorphine, fentanyl or codeine phosphate met the inclusion and exclusion criteria. The eight studies that were included were on morphine, tramadol and oxycodone for adult patients with neuropathic pain. The characteristics of the eight included studies on opioid analgesics are summarised in table 8 (for detailed full evidence tables, see appendix 9.9). Meta-analysis was carried out for individual opioid analgesics (morphine, tramadol and oxycodone) for primary outcomes and adverse effects if there were sufficient data. See section 2.1.1 for details of the analysis and synthesis of outcomes.

2.3.4. Topical capsaicin and topical lidocaine as monotherapy for neuropathic pain

Two topical treatments, capsaicin and lidocaine, were included in this review (see table 2). A total of 6057 studies were retrieved by the systematic searches and 14 randomised placebo-controlled trials were included based on the inclusion and exclusion criteria17 (nine studies for topical capsaicin and five studies for topical lidocaine). The characteristics of the 14 included studies are summarised in table 9 (for detailed full evidence tables, see appendix 9.9). Meta-analysis was carried out for both topical treatments for primary outcomes and adverse effects if sufficient data were available. See section 2.1.1 for details of the analysis and synthesis of outcomes.

2.3.5. Comparative trials on pharmacological treatments and combination therapy for neuropathic pain

Any head-to-head comparative trials and combination therapy trials that included the 34 pharmacological treatments were selected in this review (see table 2). Within the 23,207 studies that were retrieved by the systematic searches, 13 randomised trials were included based on the inclusion and exclusion criteria18 (head-to-head comparative = 10 studies, combination therapy = 3 studies. The characteristics of the 13 included studies are summarised in table 10 (for detailed full evidence tables, see appendix 9.9). Meta-analysis was carried out for different comparisons or combinations for primary outcomes and specific adverse effects if sufficient data were available. See section 2.1.1 for details of the analysis and synthesis of outcomes.

Cross-class comparative trials: antidepressants vs anti-epileptics

Primary outcomes for amitriptyline vs gabapentin as monotherapy

Table 30. GRADE profiles

Other reported pain outcomes for amitriptyline vs gabapentin as monotherapy

Table 31. GRADE profiles

Cross-class comparative trials: anti-epileptics vs opioid analgesics

Primary outcomes

The only study identified that compared pregabalin with oxycodone did not report the primary outcomes of pain.

Table 34. GRADE profiles – pregabalin vs oxycodone as monotherapy

Cross-class comparative trials: anti-epileptics vs topical treatments

Cross-class comparative trials: antidepressants vs topical treatments

Primary outcomes

The only study identified that compared amitriptyline with topical capsaicin did not report the primary outcomes of pain.

Table 36. GRADE profiles – amitriptyline vs topical capsaicin as monotherapy

Within-class comparative trials: antidepressants (TCAs) vs antidepressants (SNRIs)

Within-class comparative trials: antidepressants (TCAs) vs antidepressants (TCAs)

Primary outcomes

The only study identified that compared amitriptyline with nortriptyline did not report the primary outcomes of pain.

Table 39. GRADE profiles – amitriptyline vs nortriptyline as monotherapy

Combination therapy: anti-epileptics + opioid analgesics vs anti-epileptics alone

Primary outcomes

The only combination study identified that compared pregabalin plus oxycodone with pregabalin alone did not report the primary outcomes of pain.

Table 40. GRADE profiles – pregabalin + oxycodone vs pregabalin alone

The only combination study identified that compared gabapentin plus oxycodone with gabapentin alone did not report the primary outcomes of pain.

Table 41. GRADE profiles – gabapentin + oxycodone vs gabapentin alone

Combination therapy: anti-epileptics + opioid analgesics vs opioid analgesics alone

Primary outcomes

The only combination study identified that compared pregabalin plus oxycodone with oxycodone alone did not report the primary outcomes of pain.

Table 43. GRADE profiles – pregabalin + oxycodone vs oxycodone alone

Combination therapy: anti-epileptics + antidepressants vs anti-epileptics alone and antidepressants alone

Primary outcomes

The only combination study identified that compared gabapentin plus nortriptyline with gabapentin alone and with nortriptyline alone did not report the primary outcomes of pain or adverse effects.

2.4. Health economics evidence review

A systematic review of economic evidence on the pharmacological management of neuropathic pain found a total of 2273 papers. Full details of the search strategy are given in appendix 9.7.

In addition, the GDG had access to a relevant health technology assessment (HTA) report that had not been published during guideline development. This HTA report (Fox-Rushby JA, Griffith GL, Ross JR et al. [2010] The clinical and cost-effectiveness of different treatment pathways for neuropathic pain [NP]. NIHR Health Technology Assessment [HTA] programme, ref. 05/30/03. In press. Project abstract available from www.hta.ac.uk/1527) reviewed the clinical and cost effectiveness of different treatment pathways for neuropathic pain. The initial review included all subpopulations for the various conditions associated with neuropathic pain. However, because of the availability of evidence, the HTA report focused on two distinct neuropathic pain populations: people with painful diabetic neuropathy (PDN) and people with post-herpetic neuralgia (PHN).

In our own review of the economic literature, 479 economic studies were found on antidepressants, of which 39 were relevant based on title and abstract scanning. For anti-epileptic (anticonvulsant) drugs, 482 papers were retrieved, of which 40 were shortlisted. The search for opioids yielded 1125 hits, and a total of 140 papers were shortlisted for the review. Finally, 187 articles on topical treatments were found, of which 27 were shortlisted.

Of the 246 papers shortlisted, only 15 were ordered in full text because it was clear that the remaining papers either were studies of the wrong population or were not economic analyses. Papers on people with PHN or PDN (which made up the bulk of the 246 retrieved papers) were excluded because these populations were covered by the HTA report. Of the 15 papers ordered in full text, no study could be included. Reasons for exclusion included: study design (not an economic study); wrong patient population (no neuropathic pain; immediate post-surgery pain); wrong clinical indication (general anaesthetics); wrong route of administration (injection, infusion); and a follow-up period of less than 1 week. Appendix 10.4 lists the excluded studies and reasons for exclusion, in accordance with the economic profiles as set out in ‘The guidelines manual’ (NICE 2009).

For the purposes of this guideline, the GDG decided at the outset that neuropathic pain would be treated as a ‘blanket condition’ where possible or necessary. However, it was clear that the treatment of various subpopulations would differ considerably and that it would not be possible to extrapolate from one subgroup to all people with neuropathic pain.

No health economic modelling was undertaken for this guideline, because the GDG decided that the HTA report that was in development contained thorough data on the cost effectiveness of treatment pathways (sequences) for two common neuropathic pain conditions. The GDG reviewed, appraised and summarised the HTA report, and the results of the economic analyses from the HTA report informed this guideline as appropriate.

2.4.1. HTA report: methods

In order to present the best available evidence on the cost effectiveness of alternative pharmacological treatment pathways for people with PHN and people with PDN, the HTA report reviewed the effectiveness evidence systematically for each subpopulation. Further searches for data on resource use, drug costs and utilities associated with health states were conducted. This information was synthesised in a meta-analysis as appropriate and entered into cost-effectiveness decision models of different treatment pathways.

Markov models were developed for the evaluation of the cost effectiveness of pharmacological treatments for both the PHN and PDN populations. For each population, four separate analyses were conducted. The efficacy review fed data into an indirect comparison of all drugs for which useable data were available. The indirect analysis produced a drug hierarchy in terms of net benefit. Two additional analyses were conducted which will not be fully reported here since they did not inform the GDG decisions: a sequential analysis based on clinical convention to titrate individual tolerated drugs upwards before switching to another drug, and a Bayesian value of information (VOI) analysis that estimated the value of future research, given the underlying uncertainty.

HTA post-herpetic neuralgia (PHN) model

The model had a 10-year time horizon, with 6-week cycles in order to represent the average expected interval between clinical consultations and to capture adverse events and relapses. A cohort of patients aged 70 years was modelled. The model included eight health states: pain relief and no adverse events; pain relief and minor adverse events; no pain relief and no adverse events; no pain relief and minor adverse events; severe adverse events leading to withdrawal from treatment; spontaneous subsidence of pain; drug terminated; and death.

For the model, effectiveness in terms of pain reduction was defined as binary, with a cut-off of at least 50% pain reduction. The outcome (at least 50% pain reduction) was pooled with moderate and greater improvement outcomes on global improvement scales. Pooled estimates of pain reduction from the meta-analysis were transformed to reflect 6-week cycles and applied probabilistically by assigning a distribution to the drug and placebo. From these, estimates could be sampled and relative risk (RR) calculated. The same method was used to obtain estimates of RR of minor and major adverse events.

Data on spontaneous subsidence of pain were obtained from a separate, specific search that identified nine papers, four of which were included. Information on health state utilities was searched for in the literature. There was a complete lack of adverse-event utility data in the published literature. A Google Scholar search found data on utility estimates where dizziness and drowsiness were experienced by patients receiving TCAs. Compliance was assumed to be 100% at base case, but this was lowered to 50% in sensitivity analysis to test uncertainty.

Cost data were relevant for a UK scenario and the model adopted an NHS perspective, accounting for health outcomes in terms of quality-adjusted life years (QALYs). Discounting was in line with NICE's reference case (for details, see chapter 7 of ‘The guidelines manual’ (NICE 2009).

As no published resource use data were available, a survey of healthcare professionals was undertaken. For the PHN model, three pathways were described: GP-led, consultant-led (by an anaesthetist/pain specialist, neurologist or ophthalmologist), and jointly led care by a GP and a consultant. Results were then incorporated into the model and a separate sensitivity analysis was undertaken to test the associated uncertainty. Unit costs were taken from established sources (Personal Social Services Research Unit [PSSRU], ‘British National Formulary’ [BNF] and NHS drug tariff). Following changes in pricing during stakeholder consultation on the guideline in November 2009, the unit costs of all drugs and adverse events were updated to be correct as of December 2009. The analyses were rerun using these up-to-date unit costs and reported in January 2010 (see the draft updated analysis report for details of unit costs – available from www.hta.ac.uk/1527). The dosing regimen of pregabalin was also adjusted in the rerun analyses, to reflect the GDG's recommendations (see recommendation 1.1.10).

The indirect comparison involved probabilistic modelling of pregabalin (150, 300 and 600 mg/day), gabapentin (1800, 2400 and 3600 mg/day), oxycodone 60 mg/day, lidocaine 5% patch, epidural methylprednisolone 60 mg and lidocaine intrathecal 90 mg.

Important assumptions for the PHN model

These assumptions are adapted from those in the HTA report.

  • The cycle length was 6 weeks, within which period a clinical change (pain relief or adverse event) would be expected in practice.
  • Pain relief was assumed to relate to a reduction in the symptoms, and not the duration, of pain.
  • Beneficial effects and adverse events were assumed to start from the second cycle, or after 6 weeks.
  • Patients who did not achieve pain relief within the period of time for which the trial data were available were assumed to not respond to the drug. These patients were prescribed a new drug in the sequential analysis.
  • Patients who experienced severe adverse events leading to withdrawal had the drug terminated immediately. Adverse events were treated if necessary.
  • Effectiveness period of trial: pain relief and adverse event data from trials with durations of less than 6 weeks were not extrapolated beyond the trial duration (that is, it was assumed that there was no more pain relief or adverse events than were found during the trial).
  • Patients who experienced pain relief were assumed to remain on the drug and to continue to get pain relief until spontaneous subsidence of pain or death.
  • Patients who experienced pain relief and minor adverse events were assumed to have been titrated to the minimum dose that gives pain relief. They would continue to experience the adverse events or require drugs to alleviate them until spontaneous subsidence of pain or death.
  • Medication prescribed for minor adverse events was assumed to make the adverse events tolerable.
  • Less than 50% pain reduction was considered insufficient pain relief that did not result in a change in health state utility or QALYs.
  • Failure to respond to one drug was assumed not to affect the likelihood of responding to another.
  • The trials of clinical effectiveness identified from the systematic review did not distinguish between patients who did and did not obtain pain relief when reporting minor adverse events. It was assumed that all patients randomised to the treatment arm had an equal probability of adverse events regardless of whether or not they obtained pain relief.
  • Because PHN is not associated with increased mortality, all-cause mortality for the general public was applied in the model.
  • In the base case, adherence to drug dose and frequency was assumed to be 100%, which reflects the trial conditions under which the clinical effectiveness data were collected.

HTA painful diabetic neuropathy (PDN) model

The model was described as simulating a cohort of 2000 people aged 55 years over a lifetime horizon, with a maximum of 360 cycles each of 6 weeks' duration. Similar to the PHN model, this model was based on two pain states (at least 50% pain reduction or no pain reduction), following the conventional, dichotomous representation of the natural history of pain relief in the literature. Because spontaneous pain resolution was deemed unlikely, this health state was omitted from the PDN model.

As in the PHN model, a 6-week cycle was selected to represent the average expected interval between clinical consultations and over which the symptoms would change. This cycle length was also described as being suitable to represent increased mortality as a result of myocardial infarction (MI), which might be a relevant clinical endpoint for people receiving certain doses of drugs such as amitriptyline. Model assumptions differed slightly from those of the PHN model.

For the model, effectiveness in terms of pain relief was defined and implemented in the model using the same methods as in the PHN model. However, in the absence of PDN-specific mortality data, the PDN model used age-adjusted all-cause mortality data for people with diabetes. Spontaneous subsidence of pain was deemed not to be applicable to this model. Utilities for health states and adverse events were derived using the same methods as outlined for the PHN model. Compliance was assumed to be 100% at base case, but was lowered to 50% in sensitivity analysis to test uncertainty.

Because of a complete lack of published data, resource use was estimated via a survey to elicit expert opinion. For PDN, five main care pathways were described: GP-led care; pain-specialist-led care; diabetologist-led care; jointly led care by a GP and a pain specialist; and jointly led care by a GP and a diabetologist. At base case, resource use was consistent with patients being under the care of a diabetologist. The model adopted an NHS perspective, accounting for health outcomes in terms of QALYs. Unit costs were taken from established sources (PSSRU, BNF, NHS drug tariff). Following changes in pricing during stakeholder consultation on the guideline in November 2009, the unit costs of all drugs and adverse events were updated to be correct as of December 2009. The analyses were rerun using these up-to-date unit costs and reported in January 2010 (see the draft updated analysis report for details of unit costs – available from www.hta.ac.uk/1527). The dosing regimen of pregabalin was also adjusted in the rerun analyses, to reflect the GDG's recommendations (see recommendation 1.1.10). Discounting of costs and outcomes was in line with NICE methods (NICE 2009). Results were incorporated into the model and a separate sensitivity analysis was undertaken to test the associated uncertainty.

For the indirect modelling, some drugs could only be modelled deterministically – namely lamotrigine (400 mg/day) and nortriptyline plus fluphenazine (60 + 3 mg/day). Only the drugs that could be modelled probabilistically were included in the sequential analysis. These were pregabalin (150, 300 and 600 mg/day), gabapentin (900 and 3600 mg/day), oxcarbazepine (600, 1200 and 1800 mg/day), zonisamide (600 mg/day), topiramate (400 mg/day), amitriptyline (75 mg/day), duloxetine (20, 60 and 120 mg/day) and venlafaxine (75 and 225 mg/day).

Important assumptions for the PDN model

Assumptions for the PDN model were the same as for PHN model, except for the following:

  • Patients who experienced pain relief were assumed to remain on the drug and to continue to get pain relief for the remainder of their lifetime.
  • Patients who experienced pain relief and minor adverse events were assumed to have been titrated to the minimum dose that gives pain relief. They would continue to experience the adverse events or require drugs to alleviate them for their lifetime.
  • All-cause mortality for people with type 2 diabetes was applied in the model.

2.4.2. HTA report: results

Modelling indirect comparisons: base-case results

The indirect analysis presented results in terms of decreasing mean net benefit associated with each drug at a WTP threshold of £30,000 per QALY gained.

PHN model

For the PHN model, the indirect analysis found that pregabalin 150 mg/day is the most cost-effective treatment, as it provides the highest mean net benefit. If this treatment does not provide sufficient pain relief, the next most cost-effective option is pregabalin 300 mg/day, followed by pregabalin 600 mg/day, gabapentin 3600 mg/day, gabapentin 1800 mg/day, gabapentin 2400 mg/day, oxycodone 60 mg/day, lidocaine intrathecal 90 mg, lidocaine 5% patch and epidural methylprednisolone 60 mg.

At base case for the PHN population, pregabalin 150 mg/day had the highest probability of being most cost effective of 59.5%, followed by pregabalin 300 mg/day (30.2%) and pregabalin 600 mg/day (10.3%). All other modelled drugs, including gabapentin 1800, 2400 and 3600 mg/day and oxycodone, had zero probability of being the most cost-effective treatment option.

PDN model

For the PDN model, the hierarchy of cost effectiveness from most to least cost effective in terms of mean net benefit was duloxetine 60 mg/day, duloxetine 20 mg/day, amitriptyline 75 mg/day, duloxetine 120 mg/day, pregabalin 600 mg/day, oxcarbazepine 1200 mg/day, pregabalin 300 mg/day, oxcarbazepine 600 mg/day, gabapentin 3600 mg/day, oxcarbazepine 1800 mg/day, pregabalin 150 mg/day, topiramate 400 mg/day, venlafaxine 225 mg/day, venlafaxine 75 mg/day, gabapentin 900 mg/day and zonisamide 600 mg/day. When results for all doses of each drug were added together, oxcarbazepine appears more cost effective than pregabalin at a WTP threshold of £30,000 per QALY gained. However, this difference in net benefit was very small, although it increased slightly at the lower WTP threshold of £20,000 per QALY. The difference between pregabalin and the next best drug, gabapentin, was much greater, and so were the decrements in net benefit moving down the hierarchy to the least cost-effective drug, zonisamide.

At base case for the PDN population, duloxetine 60 mg/day had the highest probability of being most cost effective of 34.5%, followed by duloxetine 20 mg/day (33.2%), amitriptyline 75 mg/day (21.1%), pregabalin 300 mg/day (3.41%) and duloxetine 120 mg/day (3.37%).The remaining 4.42% was distributed among a further five treatment options, with seven options having a zero probability of being most cost effective.

Modelling indirect comparisons: sensitivity analysis and uncertainty

PHN model

For the PHN model, the sensitivity analysis revealed little uncertainty at the WTP threshold of £30,000 per QALY gained, with various dosages of pregabalin (see the ‘base-case results’ section above) having a combined probability of 100% of being most cost effective. This result did not change at the lower WTP threshold of £20,000 per QALY. This corresponds to pregabalin being expected to provide the highest net benefit compared with the other drugs included in the model.

PDN model

Probabilistic sensitivity analysis for the PDN model showed that, at the WTP threshold of £30,000 per QALY gained, duloxetine at various dosages (see the ‘base-case results’ section above for probabilities for individual dosages) had a 70.8% probability of being most cost effective, followed by amitriptyline at 21.3%, pregabalin at 4.4% and oxcarbazepine at 3.6%, with the remaining drugs having a negligible (less than 2%) probability of being most cost effective.

At the WTP threshold of £20,000 per QALY, the order of drugs in terms of their probability of being the most cost effective changed to: duloxetine 20 mg/day (37.8%), duloxetine 60 mg/day (28.9%; resulting in a combined probability for duloxetine of 66.7%) and amitriptyline 75 mg/day (29.8%). Oxcarbazepine in all three modelled doses had a low probability of being most cost effective (2.7%), whereas pregabalin was very unlikely to be most cost effective (0.3% probability). All of the remaining drugs, including gabapentin, had a zero probability of being most cost effective.

The cost-effectiveness acceptability curve (CEAC) for the PDN model shows that at a WTP threshold of between £20,000 and £30,000 per additional QALY, two drug treatments are highly likely to be cost effective compared with the remaining comparators evaluated: duloxetine 20 mg/day and duloxetine 60 mg/day. Duloxetine 20 mg/day was most likely to be most cost effective at WTP thresholds between £14,000 and approximately £28,000 per additional QALY. Above a threshold of £28,000 per additional QALY, duloxetine 60 mg/day became most likely to be most cost effective. Below a threshold of £14,000 per additional QALY, amitriptyline had the highest probability of being most cost effective.

In order to recommend the most cost-effective drug, it is necessary to check the consistency of a drug being most cost effective and providing the highest expected net benefit. For the PDN model, the data show that this relationship is consistent at both WTP thresholds. Duloxetine, particularly at the lower doses of 20 and 60 mg/day, provides the highest net benefit and has the highest probability of being most cost effective (with the same dosages again coming first and second in rank order). Gabapentin has a zero probability of being most cost effective and provides a lower net benefit than similar drugs in its class, including pregabalin and oxcarbazepine. If duloxetine is not an option, amitriptyline provides the second highest net benefit, followed by pregabalin and oxcarbazepine. In the clinical context of this guideline, pregabalin and amitriptyline both seem to be viable options after duloxetine. (See section 2.5 for a discussion of the clinical interpretation.)

For the drugs modelled deterministically for the PDN model, one-way and multi-way sensitivity analyses did not change the hierarchy of cost effectiveness, with nortriptyline plus fluphenazine being consistently more cost effective than lamotrigine. Nortriptyline was combined with fluphenazine in the trial to mask differences from placebo and strengthen the blinding, and hence was a single active treatment.

2.4.3. Discussion

This discussion will contrast the approaches used for the HTA report and the current clinical guideline and discuss their potential impact on interpretation and generalisability for this guideline. Then the remaining limitations of the model will be discussed.

An evidence statement summarising the (draft) findings from the HTA report is given in section 2.2.6.

Differences between the HTA report and the current guideline

It is recognised that the methodology adopted for the HTA report, in relation to both the efficacy review and the health economic evaluation, was of high quality. Therefore the information provided below does not aim to appraise the validity of the HTA report, but to assess the generalisability of the HTA report in relation to the current guideline.

Efficacy review for HTA modelling – comments on generalisability

The current guideline addresses neuropathic pain as a blanket condition, whereas the HTA report reviewed the evidence on only two conditions, namely PHN and PDN. The health economic evidence base is better for these two subpopulations than for other subgroups. Conducting de novo economic modelling in the time frame of the current guideline would not have produced a different result from that reached by the HTA, as we would have had to base our models on the same evidence base. Other subpopulations would have been difficult to model because of lack of data availability, as shown by our effectiveness and economic reviews of the literature. As the information is presented in the HTA review, the GDG was able to appraise and discuss its generalisability. The GDG agreed that the results of the cost-effectiveness analysis for individual drugs may inform the recommended sequence for neuropathic pain as a blanket condition, and that specific recommendations may be possible for subgroups.

The HTA report had no restrictions on which drugs to include in the reviews. On the other hand, the scope of the current guideline listed specific drugs to be covered, and a number of the drugs in the HTA report were not covered by the guideline scope. However, only three of these drugs were modelled, and none of them had a notable chance of being most cost effective. Therefore this is unlikely to have adversely affected the interpretation of the decision modelling results.

The exclusion criteria for the current guideline, which were agreed by the GDG members based on their expertise and experience, differed from those of the HTA report. Exclusion criteria that were used for the current guideline but not for the HTA report are listed in table 45.

Table 45. Exclusion criteria used for the current guideline but not the HTA report.

Table 45

Exclusion criteria used for the current guideline but not the HTA report.

Although the minimum trial duration (criteria 8 and 11 in table 45) was not specified in the inclusion criteria of the HTA report, the efficacy data used in the modelling were taken from RCTs with trial durations of at least 4 weeks. Limiting the route of drug administration (criteria 4 and 5) would not seem to make a lot of difference, as only lidocaine was modelled. Washout periods (criterion 10) were not taken into consideration explicitly in the HTA report; however, a summary value for rate ratios was taken from the meta-analysis, which gives some confidence in the magnitude of effect used. Applying the guideline's exclusion criterion (criterion 9) on study size (that is, excluding studies with very few participants) in the meta-analysis and probabilistic modelling would have no notable impact on the overall findings, as no studies in the probabilistic modelling in the HTA report had a sample size of below 10.

In the HTA review, the primary pain outcomes for meta-analysis were 50% response to pain (or 50% improvement in pain) and 30% response to pain (or 30% improvement in pain). The HTA review dichotomised ‘global improvement’ measures to construct 50% pain improvement and 30% pain improvement, and then pooled them with the 50% and 30% pain reduction in meta-analysis to give categories of ‘50% response to pain’ and ‘30% response to pain’. The GDG agreed that pain reduction and global improvement are two distinct outcomes that measure different aspects in pain research, a notion supported by IMMPACT. Therefore results for pain reduction and global improvement are pooled and presented separately in the current guideline.

For the health economic modelling in the HTA report, pain relief was used to define the health states, to which a global valuation of quality of life was assigned – that is, a utility estimate. Pain and other outcome data are used widely to feed into utility estimates, and pain is a dimension on the EQ-5D tool that is frequently used to measure quality of life for economic evaluations. Similar approaches to choosing and defining pain outcomes are taken in order to be as inclusive as possible and to avoid discarding data unnecessarily. From a purely conceptual viewpoint, more levels of pain states (such as 30% pain reduction) could have been modelled, but it is unlikely that this would have altered the results of the analysis, especially for those drugs most likely to be cost effective. However, it should be noted that most studies presented 50% pain reduction as the cut off, few studies used both 30% and 50% pain reduction, and fewer still provided data on 30–49% and 50% or more pain reduction.

A drug that fails to provide less than 50% pain reduction does not incur any health benefits in the model. However, introducing a lower cut-off point may result in some benefit, albeit smaller than that obtained with a drug that reduces pain by at least 50%. Thus the differences between the more effective and less effective drugs may become smaller with this approach, but the rank order in the indirect analysis would not change. In terms of the probability of a treatment being the most cost effective, those treatments that currently have a zero probability of being cost effective may achieve a small probability, but this would not alter the interpretation of the findings.

Health economic evaluation in the HTA report – comments on generalisability

A number of drugs were included in the probabilistic modelling (or deterministic modelling) and sequential analysis in the HTA report that were not covered by the scope of the current guideline, including epidural methylprednisolone and intrathecal lidocaine in the PHN model, and venlafaxine and zonisamide in the PDN model. None of these drugs was among those most likely to be cost effective. Taking these drugs out of the modelling and the modelled sequence would not change the rank order of the remaining drugs.

The HTA report decision analysis for PDN modelled amitriptyline at a dose of 75 mg/day. This is relatively high; in practice a patient may start at a lower dose followed by dose titration up to an effective dose that may still be lower than 75 mg/day. This is a limitation of the modelling, and the GDG carefully considered this when making its recommendations.

As discussed above, both PHN and PDN models were based on two pain states, which were ‘at least 50% pain reduction’ and ‘no pain reduction’. The assumption was that less than 50% pain reduction is considered insufficient and does not result in a change in health state utility or QALYs. In addition to the assumptions and implications for the guideline discussed above, basing the modelling on this meta-analysed outcome resulted in numerous drugs not being evaluated in the modelling (especially TCAs for PHN) because of a lack of data. This is a serious limitation to the completeness and applicability of the analysis, and the GDG carefully considered complementing the recommended drugs with others for the treatment of all subgroups of people with neuropathic pain.

For both the PHN and PDN models, expert opinion supplemented the data where insufficient data were available. Six experts in PHN and four experts in PDN completed a questionnaire, and the answers obtained informed the costing, as well as providing information on adverse events. The model was not sensitive to changing the three care settings for PHN or the five settings for PDN, and the rank order of the most cost-effective drugs remained unchanged.

For both the PHN and PDN models, information on the resource use of different care pathways was collected from experts (through questionnaires; see above). The care pathways used in the deterministic and probabilistic modelling do not appear to match the definition of ‘non-specialist settings’ used in the current guideline. This has two possible implications: first, cost estimates may not reflect those relevant for the current guideline; secondly, the drugs may not be suitable to be prescribed in a non-specialist setting. For example, healthcare professionals who are not pain specialists may have different levels of experience and confidence in prescribing and managing the long-term use of opioids. Moreover, some drugs that need specific monitoring, such as venlafaxine and epidural methylprednisolone, are not appropriate for use in non-specialist settings, especially in general practice. Again, the GDG discussed the results of the modelling and made recommendations based on both the presented evidence and its own judgement. In terms of the model results, an example is that venlafaxine was modelled for PDN, and was ranked in eighth place in the sequence. The decision that venlafaxine is not an alternative for this particular guideline will not affect the results, since disregarding one option from the indirect findings will not alter the ranking of treatments.

The methods used in the HTA report are of high quality, although data synthesis techniques such as network meta-analysis might have enabled the analysts to evaluate a wider network of evidence. This may have resulted in the inclusion of more drugs in the models.

HTA model limitations resulting from the reliance on RCT data

The reliance of the HTA model on data from clinical trials means that it is susceptible to the weaknesses associated with trials, such as failing to reflect real clinical practice.

The different drug doses used in the models were based on the efficacy trials. However, drug doses in trials do not necessarily reflect the doses prescribed in practice, and may be substantially higher. This is an important issue and affects evidence of both clinical and cost effectiveness. Making recommendations in an evidence-based way requires careful consideration of valid inferences. Deviations from the evidence are possible only where transparent reasoning allows this.

In addition, it is possible that the data on minor adverse events are unrepresentative. In a drug trial a patient experiencing minor adverse events may be asked to continue to take the drug for the short duration of the trial. In contrast, a member of the public under the care of their GP and/or a specialist may agree to try an alternative drug in the hope of obtaining pain relief without unpleasant adverse events.

Neither the PHN model nor the PDN model in the HTA report included combination therapy, which was a key question to be addressed by the current guideline. The limitations of the clinical evidence that informed the modelling did not allow combination therapies to be modelled. There may have been some crossover effect, as some trials allowed patients to take co-analgesics or did not report on this matter. It was not possible to estimate the implications for pain relief or adverse-event data recorded in the trials. In the absence of reliable evidence, any recommendations relating to treatment combinations should be made with caution. The deliberations and decision-making of the GDG have been recorded and are presented transparently.

The clinical trials did not report outcomes at titration stages and thus it was not possible to model movement between pain states and brief adverse events experienced during titration. Also, the clinical characteristics of pain may change over time, and patients may try a drug that has been unsuccessful at relieving their pain in the past. The paucity of data on this topic prevented this practice from being modelled, and the GDG took this into consideration when making its recommendations.

Comorbidities associated with PDN and diabetes, such as cardiovascular disease and peripheral vascular disease, were not accounted for in the model, because the systematic review excluded efficacy trials that included patients with comorbidities.

Because the RCTs for patients with PHN or PDN evaluated chronic neuropathic pain of moderate to severe intensity, the findings from such studies cannot be generalised to patients with mild PHN or PDN pain. This was taken into account by the GDG when making its recommendations.

Mortality rate imputation has been based on the best available evidence. However, there are a number of issues, including the similarity or otherwise of the mortality rates of patients with type 2 diabetes and patients with PDN. As a result, the modelling may underestimate or overestimate the survival QALYs associated with the prescription of analgesic drugs to PDN patients. This was considered by the GDG when making its recommendations.

In conclusion, all of the items listed above were discussed by the GDG, and consistency between the effectiveness review and the indirect cost-effectiveness evidence was checked. The debate and reasoning behind recommendations is recorded in section 2.5.

2.5. Evidence to recommendations

2.5.1. Antidepressants

The GDG agreed that there is good evidence (of high to moderate quality) on the efficacy of antidepressants, namely TCAs and SNRIs, for the primary outcomes on pain.

TCAs

Amitriptyline: first-line or second-line treatment for neuropathic pain

The GDG acknowledged that the majority of the evidence for TCAs is from studies on amitriptyline, and that the evidence covers various study populations with different neuropathic pain conditions. Since amitriptyline is widely used for treating neuropathic pain in current practice, the GDG agreed that amitriptyline should be recommended as either first-line or second-line treatment, depending on the person's condition, other lifestyle factors and current medication usage. Amitriptyline is not licensed for neuropathic pain, but the evidence base for treatment efficacy was deemed sufficient to make this positive recommendation.

Because amitriptyline is not licensed for neuropathic pain, the GDG came to the consensus that its initial dosage and titration should be lower than is recommended in the ‘British National Formulary’ (BNF). The GDG agreed that clear statements on drug dosage and titration in the recommendations are crucial for non-specialist settings, to emphasise the importance of titration to achieve maximum benefit. The GDG also agreed that the adverse effects of amitriptyline, as well as the special warnings and precautions for its use as specified in the SPC (based on advice from the Medicines and Healthcare Products Regulatory Agency [MHRA]), should be discussed with the person and weighed against the benefit provided.

Amitriptyline for painful diabetic neuropathy

Based on the evidence of clinical and cost effectiveness, duloxetine was recommended as first-line treatment for people with painful diabetic neuropathy (see below). However, the GDG came to consensus that if duloxetine is contraindicated, amitriptyline should be offered as an alternative first-line treatment for people with this condition.

Nortriptyline and imipramine: alternatives to amitriptyline

The GDG was concerned that many people who achieve satisfactory pain reduction with amitriptyline as first-line or second-line treatment would not be able to tolerate its adverse effects. The GDG reached a consensus that in these cases other TCAs, namely nortriptyline and imipramine, should be recommended as alternatives to amitriptyline, because there is evidence on efficacy in relation to global improvement for these drugs. Both are relatively low-cost drugs, and for this patient population they are potentially good value for money, provided that they do not cause other adverse effects that would reduce the potential gain in quality of life obtained by switching from amitriptyline.

Desipramine

Although there was some evidence for the efficacy of desipramine, it is no longer in the BNF, and so should not be used in clinical practice.

SNRIs

Duloxetine: first-line treatment for painful diabetic neuropathy

The GDG agreed that there is high-to-moderate-quality evidence for the efficacy of duloxetine in treating neuropathic pain. However, all three included studies on duloxetine were in patients with painful diabetic neuropathy, and evidence of cost effectiveness is specifically for the treatment of this condition (see section 2.4). Cost-effectiveness evidence demonstrated that duloxetine was the most cost-effective treatment for painful diabetic neuropathy. Therefore the GDG decided that duloxetine should be recommended as first-line treatment specifically for people with painful diabetic neuropathy. The GDG also agreed that the adverse effects of duloxetine, as well as the special warnings and precautions for its use as specified in the SPC (based on MHRA advice), should be discussed with the person and weighed against the benefit provided.

Venlafaxine

There is high-to-moderate-quality evidence for the efficacy of venlafaxine in treating neuropathic pain. However, based on information from the MHRA, the GDG agreed that the use of venlafaxine for the treatment of neuropathic pain would need specialist care and regular monitoring, and so it should not be initiated in non-specialist settings.

Second-line treatment after first-line treatment with an antidepressant

The GDG agreed that if satisfactory pain reduction was not achieved with amitriptyline (or nortriptyline and imipramine as alternatives) (that is, an antidepressant) as first-line treatment, a drug from another therapeutic class (namely an anti-epileptic – see section 2.5.2) should be recommended as second-line treatment, either as monotherapy or as combination therapy with first-line treatment, instead of trying another antidepressant. As described in section 2.5.2, the recommended anti-epileptic drug is pregabalin.

For people with painful diabetic neuropathy, the GDG concluded that if satisfactory pain reduction was not achieved with duloxetine as first-line treatment, it is possible that amitriptyline could be effective for a small subpopulation of people with this condition, even though both of these drugs are antidepressants. The economic model for the painful diabetic neuropathy population indicated that amitriptyline was more cost effective than duloxetine 120 mg/day, suggesting that amitriptyline is potentially cost effective for second-line treatment. The GDG was also concerned that a person may have a rapid escalation of treatment options in non-specialist settings if amitriptyline, which may be another effective option, was not offered. Therefore the GDG came to the consensus that if duloxetine is not effective as first-line treatment, people with painful diabetic neuropathy should be given the opportunity to switch to amitriptyline or pregabalin, or to combine duloxetine with pregabalin, as second-line treatment.

2.5.2. Anti-epileptics

The GDG agreed that there was insufficient evidence on the efficacy of lamotrigine, sodium valproate or phenytoin for the treatment of neuropathic pain.

Pregabalin as first-line and/or second-line treatment for neuropathic pain, in comparison with gabapentin

The GDG agreed that there is evidence (of high to moderate quality) for the efficacy of pregabalin and gabapentin for the treatment of neuropathic pain. The evidence covered various study populations with different neuropathic pain conditions. The GDG discussed the evidence on these two drugs and agreed that pregabalin is a better treatment than gabapentin for neuropathic pain for the following reasons:

  • Evidence from indirect comparisons of meta-analyses of the two treatments showed that pregabalin has lower NNT values for at least 30% pain reduction and at least 50% pain reduction compared with gabapentin, with a similar adverse-effect profile.
  • Pregabalin has simple dosing and titration compared with gabapentin.
  • Cost-effectiveness modelling showed that pregabalin is more cost effective than gabapentin (see section 2.4.2).

Following changes in pricing during stakeholder consultation on the guideline in November 2009, the unit costs of all drugs and adverse events were updated and the cost-effectiveness analyses were rerun (see section 2.4.1). The results showed that pregabalin was still more cost effective than gabapentin, although the difference was slightly reduced. Overall, pregabalin provided more mean net benefit than gabapentin, and had a higher probability of being more cost effective. With the updated prices and an amended dosing regimen for pregabalin (to reflect the GDG's recommendations), gabapentin 3600 mg/day (at 9th place in hierarchy) was the only instance where gabapentin provided a marginally higher mean net benefit than pregabalin (at 150 mg/day) (11th place). This result was considered by the GDG in a clinical context. Pregabalin 150 mg/day is a comparatively low dose and most people would be titrated up to a higher dose. In contrast, gabapentin 3600 mg/day is a very high dose, which can be difficult to manage and titrate. The GDG concluded that these considerations, together with the clinical evidence, justified the decision to recommend pregabalin rather than gabapentin, despite the minor change in the modelling results. The evidence presented, especially the health economic evaluation, has shown the certainty of the treatment sequence, and therefore supports the recommendation.

Because pregabalin and gabapentin have similar pharmacological profiles (that is, both have high affinity for the alpha-2-delta subunit of the voltage-dependent calcium channel in the central nervous system – therefore if a person had unsatisfactory pain reduction with one drug, it is highly unlikely that they will achieve pain reduction with the other), and the evidence showed that pregabalin is better than gabapentin, the GDG agreed that pregabalin should be recommended as either first-line or second-line treatment, depending on the person's condition, other lifestyle factors and current medication usage. The GDG also agreed that clear statements on drug dosage and titration in the recommendations are crucial for non-specialist settings, to emphasise the importance of titration to achieve maximum benefit. The GDG further agreed that the adverse effects of pregabalin should be discussed with the person and weighed against the benefit provided.

Second-line treatment after first-line treatment with pregabalin

The GDG agreed that if satisfactory pain reduction was not achieved with pregabalin (an anti-epileptic) as first-line treatment, a drug from another therapeutic class (namely an antidepressant – see section 2.5.1) should be recommended as second-line treatment, either as monotherapy or as combination therapy with first-line treatment, instead of trying another anti-epileptic. As described above, the recommended antidepressant is amitriptyline, with imipramine or nortriptyline as an alternative if amitriptyline is effective but the person cannot tolerate the adverse effects.

Topiramate and oxcarbazepine

There was only limited evidence (mainly from studies on patients with painful diabetic neuropathy) for the efficacy of topiramate and oxcarbazepine, and this evidence showed that patients on either of these drugs were more likely to withdraw because of adverse effects than patients on gabapentin or pregabalin (that is, the NNTH values were lower for topiramate and oxcarbazepine).

The economic model for the painful diabetic neuropathy population included three doses of oxcarbazepine. Using prices that were correct at December 2009, the model showed only very small differences in net benefit between the various doses of oxcarbazepine and pregabalin at a WTP threshold of £30,000 per QALY gained. Despite the fact that the pragmatic sequence in the HTA report changed following the rerun of the cost-effectiveness analyses so that oxcarbazepine preceded pregabalin, it was felt that this evidence was not certain enough to balance the concern posed by the clinical evidence. Given the data on adverse effects and withdrawal, the GDG did not think that oxcarbazepine should be offered and managed routinely in a non-specialist setting. In addition, topiramate was unlikely to be cost effective when other drugs, such as pregabalin, are available. Therefore it was agreed that topiramate and oxcarbazepine should not be recommended for the treatment of neuropathic pain in non-specialist settings.

Carbamazepine for the treatment of trigeminal neuralgia

The GDG recognised that the evidence on carbamazepine for the treatment of neuropathic pain overall is very limited and dated. Therefore the GDG agreed that carbamazepine should not be recommended for use across all neuropathic pain conditions. However, although only one study on carbamazepine for treating trigeminal neuralgia met the inclusion and exclusion criteria of this guideline, the GDG acknowledged that carbamazepine (within its licensed indication) has been the routine treatment for trigeminal neuralgia in clinical practice since the 1960s, and anecdotal evidence from clinical experience showed that carbamazepine may be effective for treating this condition. Because trigeminal neuralgia is an extremely painful condition, and currently there is no good-quality evidence on which to base specific recommendations for treating it, the GDG agreed that carbamazepine may have a specific role in treating trigeminal neuralgia, and expected that current routine practice will continue. Consequently, the GDG came to the consensus that a research recommendation should be made in order to further investigate the efficacy of carbamazepine for treating trigeminal neuralgia (see section 3.1).

2.5.3. Opioids

The GDG discussed the evidence on the efficacy of opioid analgesics, namely tramadol, morphine and oxycodone, for treating neuropathic pain and agreed that it was of moderate to low quality and lacked reliability. Hence the GDG recognised that the evidence does not fully reflect current clinical practice.

Tramadol: third-line treatment for neuropathic pain

There was moderate-quality evidence on primary pain outcomes for both tramadol and morphine. However, the number of patients withdrawing from studies because of adverse effects and the incidence of constipation were both lower for patients on tramadol compared with those on morphine. The GDG also acknowledged that although tramadol may lead to dependence in some people, this drug is commonly used in non-specialist settings compared with other opioid analgesics. Hence the GDG felt that it is valid and appropriate to recommend tramadol as third-line treatment for neuropathic pain in non-specialist settings, either as monotherapy or as combination therapy with second-line treatment (as rescue analgesics19). This will ensure that treatment can be continued while a person is waiting for referral to a specialist pain service and/or a condition-specific service. The GDG also agreed that clear statements on drug dosage and titration in the recommendations are crucial for non-specialist settings, to emphasise the importance of titration to achieve maximum benefit.

The acquisition costs of tramadol are relatively low for 50 mg preparations (between approximately £30 and £40 per year), and higher for higher-dose, modified-release preparations (12-hour release preparations: 100 mg, £70 to £100 per year; 150 mg, £110 to £170 per year; 200 mg, £150 to £220 per year; 24-hour release preparations: 300 mg, £270 to £300 per year; 400 mg, £370 to £400 per year). If tramadol provides pain relief for people for whom first-line and second-line treatments were ineffective and who are experiencing intolerable pain, this acquisition cost is likely to represent value for money

The adverse effects of tramadol should be discussed with the person and weighed against the benefit provided. The GDG stressed that if tramadol is used as combination therapy, more conservative dosage and titration may be required.

Morphine and oxycodone

The GDG agreed that the evidence on the efficacy of morphine and oxycodone for treating neuropathic pain was limited and of only low or moderate quality. In addition, as described above, the evidence showed that patients treated with morphine were more likely to withdraw because of adverse effects (that is, lower NNTH values) than patients treated with tramadol. There was insufficient evidence on the primary pain outcomes for oxycodone. Moreover, the GDG was concerned about the risk of long-term dependence, the severe adverse effects and the potential fatality of overdose with morphine and oxycodone. The GDG was also concerned that clinicians in non-specialist settings have very different levels of experience in prescribing and managing the long-term use of morphine and oxycodone. Therefore the GDG came to the consensus that morphine and oxycodone should not be started in non-specialist settings without an assessment by a specialist pain service or a condition-specific service. The GDG also concluded that if an assessment is carried out and treatment with morphine or oxycodone is started by a specialist pain service or condition-specific service, this treatment may be continued in a non-specialist setting provided that there is a multidisciplinary care plan, local shared care agreements and careful management of adverse effects.

2.5.4. Topical treatments

Topical capsaicin

The GDG agreed that there is limited, moderate-quality evidence indicating that topical capsaicin has no efficacy for pain reduction or global improvement for neuropathic pain overall. Based on the clinical experience of members, the GDG did acknowledge that a subgroup of people with ‘localised neuropathic pain’ may benefit from topical capsaicin. However, in view of the limited evidence available, the GDG felt that it could not recommend the use of topical capsaicin across all neuropathic pain conditions in non-specialist settings.

Topical lidocaine

Because none of the included studies on topical lidocaine reported the primary outcomes of pain, the GDG referred to the evidence statements for ‘other reported pain outcomes’ to generate discussion. The GDG agreed that there is a lack of evidence (especially placebo-controlled trials) for the efficacy of topical lidocaine for treating neuropathic pain in non-specialist settings. Moreover, in health-economic modelling, lidocaine was modelled for the patient population with painful diabetic neuropathy and provided the lowest mean net benefit at WTP thresholds between £20,000 and £30,000 per QALY gained, and had a zero probability of being the most cost-effective treatment when pregabalin is an option. However, based on the clinical experience of members, the GDG acknowledged that a subgroup of people with ‘localised neuropathic pain’ who are unable to take oral medication because of medical conditions and/or disability may benefit from topical lidocaine. In view of the limited evidence available, the GDG felt that it could not recommend the use of topical lidocaine as first-line or second-line treatment across all neuropathic pain conditions in non-specialist settings. However, topical lidocaine may play a role as a rescue analgesic (while waiting for a referral to a specialist pain service) in a very small subgroup of people with localised pain who are unable to take oral medication because of medical conditions and/or disability.

2.5.5. Comparative and combination trials

The GDG acknowledged that there were few studies involving comparative trials and combination therapy trials, and that most of the resulting evidence was of low or very low quality.

Amitriptyline or nortriptyline vs gabapentin

The GDG agreed that there was inconsistent, moderate-quality evidence on the efficacy of amitriptyline or nortriptyline compared with gabapentin. Moreover, as there is uncertainty in terms of the low-quality comparative evidence on adverse effects, and in the evidence from the cost-effectiveness analysis (see section 2.4.2), the GDG agreed that amitriptyline (with nortriptyline as an alternative) should be recommended as an option for first-line and second-line treatment.

Pregabalin vs oxycodone and pregabalin vs topical lidocaine

The evidence from both comparisons was of very low quality. In the economic modelling, topical lidocaine has virtually no probability of being a cost-effective treatment option when pregabalin is available. Therefore the GDG agreed that no recommendations should be made based on the comparative evidence.

Amitriptyline vs topical capsaicin

The comparative evidence was of low or very low quality. Therefore the GDG agreed that no recommendations should be made based on this evidence.

Imipramine vs venlafaxine

Although the GDG agreed that there was moderate-quality evidence suggesting that there is no difference between the efficacy of imipramine and venlafaxine, it concluded that safety information from the MHRA meant that venlafaxine should not be recommended for use in non-specialist settings.

Amitriptyline vs nortriptyline

The comparative evidence was of low or very low quality, and so the GDG agreed that no recommendations should be made based on this evidence. However, based on the limited evidence for the efficacy of nortriptyline (one placebo-controlled trial) and the clinical experience of GDG members, the GDG agreed that the recommendation that nortriptyline can be offered as an alternative to amitriptyline if a person achieves satisfactory pain reduction with amitriptyline but cannot tolerate its adverse effects should remain.

Combination therapy

Included studies on combination therapy compared pregabalin + oxycodone with pregabalin alone, gabapentin + oxycodone with gabapentin alone, pregabalin + oxycodone with oxycodone alone, and gabapentin + nortriptyline with gabapentin alone and nortriptyline alone. The evidence from these studies was of low or very low quality. Therefore the GDG agreed that no recommendations should be made based on these limited studies on specific combinations.

However, based on current clinical practice and the experiences of patients and carers, the GDG came to the consensus that amitriptyline (or nortriptyline or imipramine as alternatives) in combination with pregabalin, and duloxetine in combination with pregabalin, should be options for second-line treatment as combination therapy (see recommendation 1.1.13) where unsatisfactory pain reduction is achieved with a single drug, or where switching or stopping drugs is inappropriate for a particular person. A similar consensus was reached for tramadol in combination with second-line treatment (see recommendation 1.1.14).

2.5.6. Key principles of care

The GDG acknowledged that the low-quality evidence on adverse effects for both antidepressants and anti-epileptics was restricted by which, and how, data on particular adverse effects were collected in the trials. Based on the knowledge and experience of GDG members in clinical practice, the evidence did not fully reflect a complete picture of the adverse effects that people would experience in real life. Issues such as the person's vulnerability to specific adverse effects because of comorbidities, contraindications and safety considerations, current medication usage, mental health, lifestyle factors, daily activities and participation20, patient preference and patients' information needs should all be taken into consideration when selecting pharmacological treatments. The GDG further discussed that extra caution is needed when switching or combining drugs.

The GDG stressed that both early and regular clinical reviews are important in order to assess the effectiveness of the treatment and to monitor drug titration, tolerability, adverse effects and the need to continue treatment (including the possibility of gradually reducing the dose if sustained improvement is observed). The GDG acknowledged that patient diaries may be a useful tool for recording progress and informing the clinical reviews. The principle of carrying out regular clinical reviews should apply to all treatments throughout the care pathway to ensure that people receive appropriate care.

As referral to specialist pain services is not an exit from non-specialist care, but the start of a collaborative, ongoing approach to management, the GDG felt that the gateway for referrals to specialist pain services, as well as other condition-specific services, should not be at the end of the care pathway. Clinicians or healthcare professionals in non-specialist settings should consider making referrals at any stage of the care pathway, including at initial presentation and at the regular clinical reviews, if the person has severe pain or there are changes in, or deterioration of, the person's pain, health condition, and/or daily activities and participation.

To ensure continuity of care, the GDG also came to a consensus that existing treatments should be continued for people whose neuropathic pain was already effectively managed before the publication of this guideline.

2.6. Recommendations

Key principles of care

1.1.1.

Consider referring the person to a specialist pain service and/or a condition-specific service21 at any stage, including at initial presentation and at the regular clinical reviews (see recommendation 1.1.9), if:

  • they have severe pain or
  • their pain significantly limits their daily activities and participation22 or
  • their underlying health condition has deteriorated.
1.1.2.

Continue existing treatments for people whose neuropathic pain is already effectively managed23.

1.1.3.

Address the person's concerns and expectations when agreeing which treatments to use by discussing:

  • the benefits and possible adverse effects of each pharmacological treatment
  • why a particular pharmacological treatment is being offered
  • coping strategies for pain and for possible adverse effects of treatment
  • that non-pharmacological treatments are also available in non-specialist settings and/or through referral to specialist services (for example, surgical treatments and psychological therapies).
1.1.4.

When selecting pharmacological treatments, take into account:

  • the person's vulnerability to specific adverse effects because of comorbidities
  • safety considerations and contraindications as detailed in the SPC
  • patient preference
  • lifestyle factors (such as occupation)
  • mental health problems (such as depression and/or anxiety)
  • any other medication the person is taking.
1.1.5.

Explain both the importance of dosage titration and the titration process, providing written information if possible.

1.1.6.

When withdrawing or switching treatment, taper the withdrawal regimen to take account of dosage and any discontinuation symptoms.

1.1.7.

When introducing a new treatment, consider overlap with the old treatments to avoid deterioration in pain control.

1.1.8.

After starting or changing a treatment, perform an early clinical review of dosage titration, tolerability and adverse effects to assess the suitability of the chosen treatment.

1.1.9.

Perform regular clinical reviews to assess and monitor the effectiveness of the chosen treatment. Each review should include assessment of:

  • pain reduction
  • adverse effects
  • daily activities and participation24 (such as ability to work and drive)
  • mood (in particular, whether the person may have depression and/or anxiety25)
  • quality of sleep
  • overall improvement as reported by the person.

First-line treatment

1.1.10.

Offer oral amitriptyline* or pregabalin as first-line treatment (but see recommendation 1.1.11 for people with painful diabetic neuropathy).

  • For amitriptyline*: start at 10 mg per day, with gradual upward titration to an effective dose or the person's maximum tolerated dose of no higher than 75 mg per day (higher doses could be considered in consultation with a specialist pain service).
  • For pregabalin: start at 150 mg per day (divided into two doses; a lower starting dose may be appropriate for some people), with upward titration to an effective dose or the person's maximum tolerated dose of no higher than 600 mg per day (divided into two doses).
1.1.11.

For people with painful diabetic neuropathy, offer oral duloxetine as first-line treatment. If duloxetine is contraindicated, offer oral amitriptyline*.

  • For duloxetine: start at 60 mg per day (a lower starting dose may be appropriate for some people), with upward titration to an effective dose or the person's maximum tolerated dose of no higher than 120 mg per day.
  • For amitriptyline*: see recommendation 1.1.10.
1.1.12.

Based on both the early and regular clinical reviews:

  • if there is satisfactory improvement, continue the treatment; consider gradually reducing the dose over time if improvement is sustained
  • if amitriptyline* as first-line treatment results in satisfactory pain reduction but the person cannot tolerate the adverse effects, consider oral imipramine* or nortriptyline* as an alternative.

Second-line treatment

1.1.13.

If satisfactory pain reduction is not achieved with first-line treatment at the maximum tolerated dose, offer treatment with another drug instead of or in combination with the original drug, after informed discussion with the person.

  • If first-line treatment was with amitriptyline* (or imipramine* or nortriptyline*), switch to or combine with oral pregabalin.
  • If first-line treatment was with pregabalin, switch to or combine with oral amitriptyline* (or imipramine* or nortriptyline* as an alternative if amitriptyline* is effective but the person cannot tolerate the adverse effects; see recommendation 1.1.12).
  • For people with painful diabetic neuropathy:

    if first-line treatment was with duloxetine, switch to amitriptyline* or pregabalin, or combine with pregabalin

    if first-line treatment was with amitriptyline*, switch to or combine with pregabalin.

    Dosage and titration should be the same as in recommendation 1.1.10.

Third-line treatment

1.1.14.

If satisfactory pain reduction is not achieved with second-line treatment

  • refer the person to a specialist pain service and/or a condition-specific service26 and
  • while waiting for referral:

    consider oral tramadol as third-line treatment instead of or in combination27 with the second-line treatment

    consider topical lidocaine28 for treatment of localised pain for people who are unable to take oral medication because of medical conditions and/or disability.

1.1.15.

For tramadol as monotherapy, start at 50 to 100 mg not more often than every 4 hours, with upward titration if required to an effective dose or the person's maximum tolerated dose of no higher than 400 mg per day. If tramadol is used as combination therapy, more conservative titration may be required.

Other treatments

1.1.16.

Do not start treatment with opioids (such as morphine or oxycodone) other than tramadol without an assessment by a specialist pain service or a condition-specific service26.

1.1.17.

Pharmacological treatments other than those recommended in this guideline that are started by a specialist pain service or a condition-specific service26 may continue to be prescribed in non-specialist settings, with a multidisciplinary care plan, local shared care agreements and careful management of adverse effects.

Footnotes

13

For the full search strategies, see appendix 9.7; for the two GDG short questionnaires on inclusion and exclusion criteria, see appendix 9.3A; for the full review protocol, see appendix 9.2.

14

For the full search strategies, see appendix 9.7; for the two GDG short questionnaires on inclusion and exclusion criteria, see appendix 9.3A; for the full review protocol, see appendix 9.2; for study selection flowcharts and list of excluded studies, see appendix 9.4.

15

For the full search strategies, see appendix 9.7; for the two GDG short questionnaires on inclusion and exclusion criteria, see appendix 9.3A; for the full review protocol, see appendix 9.2; for study selection flowcharts and list of excluded studies, see appendix 9.4.

16

For the full search strategies, see appendix 9.7; for the two GDG short questionnaires on inclusion and exclusion criteria, see appendix 9.3A; for the full review protocol, see appendix 9.2; for study selection flowcharts and list of excluded studies, see appendix 9.4.

17

For the full search strategies, see appendix 9.7; for the two GDG short questionnaires on inclusion and exclusion criteria, see appendix 9.3A; for the full review protocol, see appendix 9.2; for study selection flowcharts and list of excluded studies, see appendix 9.4.

18

For the full search strategies, see appendix 9.7; for the two GDG short questionnaires on inclusion and exclusion criteria, see appendix 9.3A; for the full review protocol, see appendix 9.2; for study selection flowcharts and list of excluded studies, see appendix 9.4.

19

Rescue analgesics are analgesics for breakthrough pain, which is pain that comes on suddenly for short periods of time and is not alleviated by the patient's normal pain suppression management.

20

The World Health Organization ICF (International Classification of Functioning, Disability and Health) (2001) defines participation as ‘A person's involvement in a life situation.’ It includes the following domains: learning and applying knowledge, general tasks and demands, mobility, self-care, domestic life, interpersonal interactions and relationships, major life areas, community, and social and civil life.

21

A condition-specific service is a specialist service that provides treatment for the underlying health condition that is causing neuropathic pain. Examples include neurology, diabetology and oncology services.

22

The World Health Organization ICF (International Classification of Functioning, Disability and Health) (2001) defines participation as ‘A person's involvement in a life situation.’ It includes the following domains: learning and applying knowledge, general tasks and demands, mobility, self-care, domestic life, interpersonal interactions and relationships, major life areas, community, and social and civil life.

23

Note that there is currently no good-quality evidence on which to base specific recommendations for treating trigeminal neuralgia. The GDG expected that current routine practice will continue until new evidence is available (see also section 3.1).

24

The World Health Organization ICF (International Classification of Functioning, Disability and Health) (2001) defines participation as ‘A person's involvement in a life situation.’ It includes the following domains: learning and applying knowledge, general tasks and demands, mobility, self-care, domestic life, interpersonal interactions and relationships, major life areas, community, and social and civil life.

25

Refer if necessary to ‘Anxiety’ (NICE clinical guideline 22), ‘Depression’ (NICE clinical guideline 90) and/or ‘Depression in adults with a chronic physical health problem’ (NICE clinical guideline 91) (available at www​.nice.org.uk).

*

In these recommendations, drug names are marked with an asterisk if they do not have UK marketing authorisation for the indication in question at the time of publication (March 2010). Informed consent should be obtained and documented.

26

A condition-specific service is a specialist service that provides treatment for the underlying health condition that is causing neuropathic pain. Examples include neurology, diabetology and oncology services.

27

The combination of tramadol with amitriptyline, nortriptyline, imipramine or duloxetine is associated with only a low risk of serotonin syndrome (the features of which include confusion, delirium, shivering, sweating, changes in blood pressure and myoclonus).

28

Topical lidocaine is licensed for post-herpetic neuralgia, but not for other neuropathic pain conditions.

Copyright © 2010, National Institute for Health and Clinical Excellence.

All rights reserved. This material may be freely reproduced for educational and not-for-profit purposes. No reproduction by or for commercial organisations, or for commercial purposes, is allowed without the express written permission of NICE.

Bookshelf ID: NBK82982

Views

Recent Activity

Your browsing activity is empty.

Activity recording is turned off.

Turn recording back on

See more...