Drugs for Pain Associated With Postherpetic Neuralgia
Drugs for Pain Associated With Postherpetic Neuralgia
The results of the literature search with number of studies reviewed and eligible for inclusion in the study are shown in Figure 1. Of 1898 articles obtained from the search, 262 full-text publications were reviewed for eligibility. One hundred and thirty-one of these were excluded after reviewing the full-text publication based on the study exclusion criteria. Thirty-three of the remaining studies were in PHN, two in MS-related pain, four in CPSP pain, 10 in HIV-related neuropathy and three in posttraumatic pain. Six studies were excluded after data extraction – five PHN studies [one because of non-randomised subpopulation, three because of inability to extract relevant outcomes and one because of mixed pain conditions were reported] and one CPSP study [evaluated prophylactic use rather than treatment].
(Enlarge Image)
Figure 1.
Literature search results.
Study durations ranged from 4 weeks to 1 year for PHN, 3 months for MS-related pain, 4–12 weeks for CPSP, 4–8 weeks for posttraumatic pain and 4–16 weeks for HIV-related neuropathy. Trials were mostly conducted in the US (45%) and Europe (27%). Overall, studies reported low reporting and performance bias ( Appendix ) when assessed using the Cochrane Risk of Bias tool. Selection bias was unclear for several of reviewed studies.
Among NeP conditions of interest, studies examining treatments for PHN provided the most data for inclusion into an MTC ( Table 2 ). As very few studies reported NeP reduction in MS pain, CPSP, HIV-related pain, or posttraumatic pain, efficacy and safety outcomes for these studies are only presented qualitatively.
Data from 28 studies examining 21 interventions (including placebo) for PHN across 4317 patients were included in the quantitative overview (Figure 2). Ten of the 28 studies included in the MTC had sample sizes less than 50 ( Table 2 ). Publication bias was not evident from the graphical plots of 11-point NRS and 100 mm VAS treatment efficacy vs. sample size (Figure 3). Although small studies gave more favourable treatment effects (lower pain reduction), the full range of outcomes were indeed represented by several small studies. Larger studies, which were noticeably fewer in number also reported favourable results, which were concentrated in the middle range relative to the results from smaller studies.
(Enlarge Image)
Figure 2.
Network of head-to-head and placebo-controlled comparisons in PHN. Numbered lines denote more than one study available for the indicated comparison.
(Enlarge Image)
Figure 3.
Funnel plot of 11-point NRS and 100 mm VAS outcomes vs. study sample size for all PHN treatments.
Of the 28 studies included, five were cross-over in study design. All studies reporting imputation for missing data used the method of last observation carried forward. One study reported a head-to-head trial between opioids (morphine or methadone) and TCAs (nortriptyline or desipramine), which were considered to be independent treatment arms and analysed by drug class rather than individual agents.
Overall, no clear evidence of heterogeneity was identified among interventions with three or more comparisons against placebo for any efficacy intervention or outcome, the exception being pregabalin ≥ 300 mg for discontinuations and NGX-4010 for AEs. Visual inspection of these latter data suggested heterogeneity could be because of variation in study durations and mean patient age. However, the small number of studies precluded definitive identification of any study-specific trends.
On the 11-point NRS, all treatments identified were guideline-recommended except memantine. Five of the 10 treatments were studied in at least 50 patients, of which the MTC estimates for the mean reduction in 11-point NRS pain over that of placebo for opioids [morphine or methadone] were most effective (Figure 4a). Memantine (studied in 24 patients) was the only treatment not statistically more effective than placebo in PHN [−0.40, (95% CrI: −2.12, 1.30)]. Treatments with the most number of comparisons (including multiple comparisons within a study) were gabapentin (n = 10 from six studies), NGX-4010 (n = 7 from five studies) and pregabalin ≥ 300 mg (n = 5 from four studies) with net reductions of −0.88 (−1.13, −0.63), −0.41 (−0.65, −0.17) and −1.57 (−1.93, −1.22), respectively.
(Enlarge Image)
Figure 4.
Mixed treatment comparisons estimating the mean reduction in PHN pain over that of placebo on the (A) 11-pt NRS; (B) 100 mm VAS. Mixed treatment comparisons estimating the relative risk of (C) achieving ≥ 30% and ≥ 50% reduction in pain; (D) experiencing an AE; (E) discontinuing relative to placebo. For all panels, numbers in parentheses indicate number of available comparisons for each treatment, italicised treatments displayed in grey indicate those studied in a total of < 50 patients, guideline-recommended treatments are indicated by '**' (Level A) or '*' (Level B)
Half of the treatments (4) with available VAS data were not guideline-recommended (Figure 4b). Divalproex sodium, capsaicin (topical 0.075%) and gabapentin had effects statistically higher than placebo, although only gabapentin (which is guideline-recommended) was studied in ≥ 50 patients. Similar to the NRS outcome, the VAS outcome for gabapentin had the most data available for comparison (n = 6 from three studies).
Model validation analyses showed indirect results of the MTC were consistent with the available head-to-head (direct) comparisons from the clinical trial data. Comparison of nortriptyline vs. gabapentin (95% CrI: −5.10, 5.14), pregabalin ≥ 300 mg vs. lidocaine 5% plaster (95% CrI: −6.07, 6.04), NRS model outcomes and the amitriptyline vs. desipramine (95% CrI: −82.3, 79.3), amitriptyline vs. fluoxetine (95% CrI: −82.3, 79.1), VAS model outcomes with their respective direct trial data, each revealed differences centred around zero, indicating good agreement between the model results and the published data.
All identified PHN treatments with ≥ 30% or ≥ 50% pain responder data were guideline-supported with Level A evidence (Figure 4c), except tramadol. All PHN treatments except lidocaine 5% plaster and nortriptyline were studied in ≥ 50 patients. Pregabalin ≥ 300 mg/day was the most effective PHN treatment studied in ≥ 50 patients in reducing pain by ≥ 30% and ≥ 50% [relative risk (RR) vs. placebo = 2.13 and 2.44, respectively]. NGX-4010 was also studied in ≥ 50 patients and was least effective on both of those scales for PHN. For PHN treatments where both response levels were reported by at least one study, the overall relative probability of achieving ≥ 50% response compared with placebo was found to be higher than the probability of ≥ 30% response.
Among treatments studied for PHN, all had statistically more AEs than placebo except lidocaine 5% plaster [RR (95% CrI): 0.93, (0.53, 1.32)], tramadol [0.95 (0.63, 1.27)], nortriptyline [1.31 (0.37, 1.77)] desipramine [1.50 (0.85, 1.81)] and amitriptyline [1.38 (0.89, 1.70)]. Lorazepam, opioids and maprotiline reported the highest risk of AEs (Figure 4d).
Postherpetic neuralgia discontinuations were similar to placebo for all treatments except TCAs and opioids, which were studied within the same trial. This study reported aggregate results of nortriptyline and desipramine ('TCA') to demonstrate more discontinuations than placebo [4.03 (1.37, 6.38)]. However, estimates of nortriptyline [0.66 (0.06, 2.17)] and desipramine [1.75 (0.05, 5.41)] obtained from other studies did not reveal statistically higher discontinuation probabilities compared with placebo (Figure 4e).
Outcomes for the two MS-related pain studies are listed in Table 3 . MS-related pain can be central pain due paroxysmal symptoms, trigeminal neuralgia, or painful tonic spasms or pain as a sequel to MS symptoms. Additionally, pain could be related to treatment for MS. Both studies identified here examined efficacy for the treatment of central MS-related pain. Patients with chronic pain conditions in addition to MS that could interfere with MS-related pain and patients with trigeminal neuralgia and other painful manifestations (back pain, visceral pain and painful tonic spasms) were excluded from the studies.
No direct or indirect comparisons were performed because of insufficient data. The study with the largest sample size included only 20 patients in the trial. The Breuer 2007 study reported pain reduction on the NRS and the proportion of patients reaching ≥ 30% reduction in pain, neither of which showed significant improvement of lamotrigine over placebo. The Rossi 2009 study reported a very large treatment effect of levetiracetam on the VAS, which the authors report to be a significant reduction over placebo (p < 0.05). Available AE and discontinuation data were not significantly greater than placebo.
Three CPSP studies were identified, with pregabalin being the only treatment examined in ≥ 50 patients (n = 219) ( Table 4 ). No direct or indirect comparisons of treatment efficacy or safety were performed because of insufficient data. Two studies reported modest treatment effects for pregabalin, amitriptyline and carbamazepine. One study provided a measure of uncertainty around these estimates indicating net treatment effect of pregabalin was not greater than placebo for pain NRS [−0.2 (95% CI: −0.7 to 0.4)] or VAS [−1 (95% CI: −7.0 to 5.0)] outcomes. None of the CPSP studies reported binary pain reduction outcomes. AEs were reported to be similar to placebo for lamotrigine [RR 0.94 (0.62–1.45)] and slightly higher than placebo for pregabalin [RR 1.27 (1.03–1.57)]. Two studies reported discontinuation data showing lamotrigine [RR 1.3 (0.41–4.11)] and pregabalin [RR 0.89 (0.49–1.61)] had similar withdrawal rates to placebo.
Results of the three posttraumatic pain studies are listed in Table 5 . Pregabalin was studied in a larger sample of patients (n = 254) and showed significantly better outcomes compared with placebo for mean reduction on the pain NRS [−0.62 (95% CI: −1.09, −0.15)], ≥ 50% responders [1.57 (95% CI: 0.64–3.89)] and AEs [1.47 (95% CI: 1.25–1.73)]. In comparison, a cross-over study examining gabapentin demonstrated beneficial effects with pain reduction using the VAS, but was not significantly better than placebo on the ≥ 30% or ≥ 50% pain responder outcomes. A noticeable difference in the mean reduction on VAS observed during the two cross-over periods was not significant when adjusted for baseline pain intensity (p = 0.20). The third study of mexiletine was published over 20 years ago, reported only the median NRS outcome (−3) in 11 patients.
HIV-associated neuropathy had more available data than CPSP, MS-related pain, or posttraumatic pain ( Table 6 ). Four of the nine studies were conducted in ≥ 50 patients with the remaining in < 50. Consistent with treatment guidelines, capsaicin was the only treatment suggesting a meaningful treatment effect on the 11-point NRS with a −2.5 greater NRS reduction in pain over placebo. However, the study was small (n = 26) and an uncertainty interval around this measure was not reported and could not be calculated. Memantine showed worsening of pain on the 11-point NRS and NGX-4010 was the only treatment to demonstrate significant ≥ 30% or ≥ 50% reduction in pain relative to placebo (overall average of three doses: RR 1.85; 95% CI: 1.34, 2.55).
NGX-4010 and pregabalin both had slightly higher rates of AEs than placebo. Eight studies reported discontinuation rates for seven interventions, allowing for a quantitative indirect comparison of discontinuations relative to placebo (Figure 5). Estimates showed that, except for capsaicin 0.075%, discontinuations with each treatment were generally similar and not significantly different than placebo, as the 95% CrI included 1.
(Enlarge Image)
Figure 5.
Mixed treatment comparisons estimating the relative risk of discontinuing compared with placebo for HIV NeP. Italicised treatments displayed in grey indicate those studied in a total of < 50 patients, guideline-recommended treatments are indicated by '**' (Level A) or '*' (Level B)
The only study (n = 88) to report a change in EQ-5D health state utility (range: −0.11 to 1) from baseline to end-point showed a higher absolute increase in utility for lidocaine 5% plaster (0.12) over pregabalin (0) for the treatment of PHN. Kim et al. reported a zero net increase in utility for pregabalin (relative to placebo) in a CPSP population. In contrast, this study reported a net increase of 2.47 (95% CI: −1.8, 7.9, p = 0.22) in VAS health state utility (range: 0–100) for pregabalin in the same study population. No studies of other NeP conditions reported utility outcomes.
Results
Literature Search
The results of the literature search with number of studies reviewed and eligible for inclusion in the study are shown in Figure 1. Of 1898 articles obtained from the search, 262 full-text publications were reviewed for eligibility. One hundred and thirty-one of these were excluded after reviewing the full-text publication based on the study exclusion criteria. Thirty-three of the remaining studies were in PHN, two in MS-related pain, four in CPSP pain, 10 in HIV-related neuropathy and three in posttraumatic pain. Six studies were excluded after data extraction – five PHN studies [one because of non-randomised subpopulation, three because of inability to extract relevant outcomes and one because of mixed pain conditions were reported] and one CPSP study [evaluated prophylactic use rather than treatment].
(Enlarge Image)
Figure 1.
Literature search results.
Study durations ranged from 4 weeks to 1 year for PHN, 3 months for MS-related pain, 4–12 weeks for CPSP, 4–8 weeks for posttraumatic pain and 4–16 weeks for HIV-related neuropathy. Trials were mostly conducted in the US (45%) and Europe (27%). Overall, studies reported low reporting and performance bias ( Appendix ) when assessed using the Cochrane Risk of Bias tool. Selection bias was unclear for several of reviewed studies.
Among NeP conditions of interest, studies examining treatments for PHN provided the most data for inclusion into an MTC ( Table 2 ). As very few studies reported NeP reduction in MS pain, CPSP, HIV-related pain, or posttraumatic pain, efficacy and safety outcomes for these studies are only presented qualitatively.
Postherpetic Neuralgia
Data from 28 studies examining 21 interventions (including placebo) for PHN across 4317 patients were included in the quantitative overview (Figure 2). Ten of the 28 studies included in the MTC had sample sizes less than 50 ( Table 2 ). Publication bias was not evident from the graphical plots of 11-point NRS and 100 mm VAS treatment efficacy vs. sample size (Figure 3). Although small studies gave more favourable treatment effects (lower pain reduction), the full range of outcomes were indeed represented by several small studies. Larger studies, which were noticeably fewer in number also reported favourable results, which were concentrated in the middle range relative to the results from smaller studies.
(Enlarge Image)
Figure 2.
Network of head-to-head and placebo-controlled comparisons in PHN. Numbered lines denote more than one study available for the indicated comparison.
(Enlarge Image)
Figure 3.
Funnel plot of 11-point NRS and 100 mm VAS outcomes vs. study sample size for all PHN treatments.
Of the 28 studies included, five were cross-over in study design. All studies reporting imputation for missing data used the method of last observation carried forward. One study reported a head-to-head trial between opioids (morphine or methadone) and TCAs (nortriptyline or desipramine), which were considered to be independent treatment arms and analysed by drug class rather than individual agents.
Overall, no clear evidence of heterogeneity was identified among interventions with three or more comparisons against placebo for any efficacy intervention or outcome, the exception being pregabalin ≥ 300 mg for discontinuations and NGX-4010 for AEs. Visual inspection of these latter data suggested heterogeneity could be because of variation in study durations and mean patient age. However, the small number of studies precluded definitive identification of any study-specific trends.
On the 11-point NRS, all treatments identified were guideline-recommended except memantine. Five of the 10 treatments were studied in at least 50 patients, of which the MTC estimates for the mean reduction in 11-point NRS pain over that of placebo for opioids [morphine or methadone] were most effective (Figure 4a). Memantine (studied in 24 patients) was the only treatment not statistically more effective than placebo in PHN [−0.40, (95% CrI: −2.12, 1.30)]. Treatments with the most number of comparisons (including multiple comparisons within a study) were gabapentin (n = 10 from six studies), NGX-4010 (n = 7 from five studies) and pregabalin ≥ 300 mg (n = 5 from four studies) with net reductions of −0.88 (−1.13, −0.63), −0.41 (−0.65, −0.17) and −1.57 (−1.93, −1.22), respectively.
(Enlarge Image)
Figure 4.
Mixed treatment comparisons estimating the mean reduction in PHN pain over that of placebo on the (A) 11-pt NRS; (B) 100 mm VAS. Mixed treatment comparisons estimating the relative risk of (C) achieving ≥ 30% and ≥ 50% reduction in pain; (D) experiencing an AE; (E) discontinuing relative to placebo. For all panels, numbers in parentheses indicate number of available comparisons for each treatment, italicised treatments displayed in grey indicate those studied in a total of < 50 patients, guideline-recommended treatments are indicated by '**' (Level A) or '*' (Level B)
Half of the treatments (4) with available VAS data were not guideline-recommended (Figure 4b). Divalproex sodium, capsaicin (topical 0.075%) and gabapentin had effects statistically higher than placebo, although only gabapentin (which is guideline-recommended) was studied in ≥ 50 patients. Similar to the NRS outcome, the VAS outcome for gabapentin had the most data available for comparison (n = 6 from three studies).
Model validation analyses showed indirect results of the MTC were consistent with the available head-to-head (direct) comparisons from the clinical trial data. Comparison of nortriptyline vs. gabapentin (95% CrI: −5.10, 5.14), pregabalin ≥ 300 mg vs. lidocaine 5% plaster (95% CrI: −6.07, 6.04), NRS model outcomes and the amitriptyline vs. desipramine (95% CrI: −82.3, 79.3), amitriptyline vs. fluoxetine (95% CrI: −82.3, 79.1), VAS model outcomes with their respective direct trial data, each revealed differences centred around zero, indicating good agreement between the model results and the published data.
All identified PHN treatments with ≥ 30% or ≥ 50% pain responder data were guideline-supported with Level A evidence (Figure 4c), except tramadol. All PHN treatments except lidocaine 5% plaster and nortriptyline were studied in ≥ 50 patients. Pregabalin ≥ 300 mg/day was the most effective PHN treatment studied in ≥ 50 patients in reducing pain by ≥ 30% and ≥ 50% [relative risk (RR) vs. placebo = 2.13 and 2.44, respectively]. NGX-4010 was also studied in ≥ 50 patients and was least effective on both of those scales for PHN. For PHN treatments where both response levels were reported by at least one study, the overall relative probability of achieving ≥ 50% response compared with placebo was found to be higher than the probability of ≥ 30% response.
Among treatments studied for PHN, all had statistically more AEs than placebo except lidocaine 5% plaster [RR (95% CrI): 0.93, (0.53, 1.32)], tramadol [0.95 (0.63, 1.27)], nortriptyline [1.31 (0.37, 1.77)] desipramine [1.50 (0.85, 1.81)] and amitriptyline [1.38 (0.89, 1.70)]. Lorazepam, opioids and maprotiline reported the highest risk of AEs (Figure 4d).
Postherpetic neuralgia discontinuations were similar to placebo for all treatments except TCAs and opioids, which were studied within the same trial. This study reported aggregate results of nortriptyline and desipramine ('TCA') to demonstrate more discontinuations than placebo [4.03 (1.37, 6.38)]. However, estimates of nortriptyline [0.66 (0.06, 2.17)] and desipramine [1.75 (0.05, 5.41)] obtained from other studies did not reveal statistically higher discontinuation probabilities compared with placebo (Figure 4e).
MS-related Pain Outcomes
Outcomes for the two MS-related pain studies are listed in Table 3 . MS-related pain can be central pain due paroxysmal symptoms, trigeminal neuralgia, or painful tonic spasms or pain as a sequel to MS symptoms. Additionally, pain could be related to treatment for MS. Both studies identified here examined efficacy for the treatment of central MS-related pain. Patients with chronic pain conditions in addition to MS that could interfere with MS-related pain and patients with trigeminal neuralgia and other painful manifestations (back pain, visceral pain and painful tonic spasms) were excluded from the studies.
No direct or indirect comparisons were performed because of insufficient data. The study with the largest sample size included only 20 patients in the trial. The Breuer 2007 study reported pain reduction on the NRS and the proportion of patients reaching ≥ 30% reduction in pain, neither of which showed significant improvement of lamotrigine over placebo. The Rossi 2009 study reported a very large treatment effect of levetiracetam on the VAS, which the authors report to be a significant reduction over placebo (p < 0.05). Available AE and discontinuation data were not significantly greater than placebo.
Central Poststroke Pain
Three CPSP studies were identified, with pregabalin being the only treatment examined in ≥ 50 patients (n = 219) ( Table 4 ). No direct or indirect comparisons of treatment efficacy or safety were performed because of insufficient data. Two studies reported modest treatment effects for pregabalin, amitriptyline and carbamazepine. One study provided a measure of uncertainty around these estimates indicating net treatment effect of pregabalin was not greater than placebo for pain NRS [−0.2 (95% CI: −0.7 to 0.4)] or VAS [−1 (95% CI: −7.0 to 5.0)] outcomes. None of the CPSP studies reported binary pain reduction outcomes. AEs were reported to be similar to placebo for lamotrigine [RR 0.94 (0.62–1.45)] and slightly higher than placebo for pregabalin [RR 1.27 (1.03–1.57)]. Two studies reported discontinuation data showing lamotrigine [RR 1.3 (0.41–4.11)] and pregabalin [RR 0.89 (0.49–1.61)] had similar withdrawal rates to placebo.
Posttraumatic Pain
Results of the three posttraumatic pain studies are listed in Table 5 . Pregabalin was studied in a larger sample of patients (n = 254) and showed significantly better outcomes compared with placebo for mean reduction on the pain NRS [−0.62 (95% CI: −1.09, −0.15)], ≥ 50% responders [1.57 (95% CI: 0.64–3.89)] and AEs [1.47 (95% CI: 1.25–1.73)]. In comparison, a cross-over study examining gabapentin demonstrated beneficial effects with pain reduction using the VAS, but was not significantly better than placebo on the ≥ 30% or ≥ 50% pain responder outcomes. A noticeable difference in the mean reduction on VAS observed during the two cross-over periods was not significant when adjusted for baseline pain intensity (p = 0.20). The third study of mexiletine was published over 20 years ago, reported only the median NRS outcome (−3) in 11 patients.
HIV-associated Neuropathy
HIV-associated neuropathy had more available data than CPSP, MS-related pain, or posttraumatic pain ( Table 6 ). Four of the nine studies were conducted in ≥ 50 patients with the remaining in < 50. Consistent with treatment guidelines, capsaicin was the only treatment suggesting a meaningful treatment effect on the 11-point NRS with a −2.5 greater NRS reduction in pain over placebo. However, the study was small (n = 26) and an uncertainty interval around this measure was not reported and could not be calculated. Memantine showed worsening of pain on the 11-point NRS and NGX-4010 was the only treatment to demonstrate significant ≥ 30% or ≥ 50% reduction in pain relative to placebo (overall average of three doses: RR 1.85; 95% CI: 1.34, 2.55).
NGX-4010 and pregabalin both had slightly higher rates of AEs than placebo. Eight studies reported discontinuation rates for seven interventions, allowing for a quantitative indirect comparison of discontinuations relative to placebo (Figure 5). Estimates showed that, except for capsaicin 0.075%, discontinuations with each treatment were generally similar and not significantly different than placebo, as the 95% CrI included 1.
(Enlarge Image)
Figure 5.
Mixed treatment comparisons estimating the relative risk of discontinuing compared with placebo for HIV NeP. Italicised treatments displayed in grey indicate those studied in a total of < 50 patients, guideline-recommended treatments are indicated by '**' (Level A) or '*' (Level B)
EQ-5D Outcomes
The only study (n = 88) to report a change in EQ-5D health state utility (range: −0.11 to 1) from baseline to end-point showed a higher absolute increase in utility for lidocaine 5% plaster (0.12) over pregabalin (0) for the treatment of PHN. Kim et al. reported a zero net increase in utility for pregabalin (relative to placebo) in a CPSP population. In contrast, this study reported a net increase of 2.47 (95% CI: −1.8, 7.9, p = 0.22) in VAS health state utility (range: 0–100) for pregabalin in the same study population. No studies of other NeP conditions reported utility outcomes.
