An Evidence-Based Algorithm for the Treatment of Neuropathic Pain

Introduction

Given the significant and growing prevalence of neuropathic pain, estimated to affect up to 3% of the population[1] (a pooled estimate[1-7]) and that approximately 1 in 5 European adults reports chronic pain,[8] the value of proper treatment options derived from the systematic review of randomized, placebo-controlled clinical trials cannot be underestimated. Neuropathic pain is distinct from other (nociceptive) types of commonly reported pain conditions, including headache, back pain, and other types of musculoskeletal pain. Neuropathic pain may occur in diabetic sensorimotor polyneuropathy, the most common type of generalized polyneuropathy,[9] which affects approximately 54% of patients with type 1 diabetes and 45% of patients with type 2 diabetes, and neuropathic pain thus develops in up to 25% of patients with diabetes.[10,11] Approximately 800,000 cases of shingles are reported each year in the United States, with 25% to 50% of those cases developing postherpetic neuralgia (PHN), a neuropathic pain condition resulting from infection by the herpes zoster virus.[3] Neuropathic pain can develop from HIV/AIDS, various toxins (eg, neurotoxins), alcohol abuse, acute trauma (including surgery), chronic trauma (eg, repetitive motion disorders, such as carpal tunnel syndrome, the most common of the mononeuropathies, affecting 2.8% to 4.6% of the adult population[12-14]), central nervous system diseases (such as stroke, multiple sclerosis, and spinal cord injury), or autoimmune conditions (such as celiac disease).[15]

The necessity of such an evidence-based treatment algorithm for use by primary care physicians (PCPs) is further highlighted as patients presenting with pain (complaints of all types of pain) represent the most common reason to seek medical attention,[16,17] and account for approximately 25% to 50% of primary care visits, with 20% of these visits due to persistent chronic pain conditions.[18,19] In the United States, 9 out of 10 adults 18 years and older report experiencing pain at least once a month and 42% reported pain on a daily basis.[16] Compared with patients without chronic pain, a study using data from a Danish multidisciplinary pain center found that those patients with chronic pain are 5 times more likely to use healthcare services.[20] As such, it is even more critical for the PCP to identify those cases of neuropathic pain that can be best treated either by the PCP alone, through referral to a pain specialist who will, with the PCP, manage the patient with the help of 1 or 2 other clinicians in those cases of moderate complexity (eg, persistent, treatment-resistant neuropathic pain, comorbid psychiatric condition, necessity of invasive procedures), or with a multidisciplinary team approach in cases of high complexity.

{mospagebreak}

Pain signals are carried by 2 types of nerve fibers termed Adelta and C-fibers. These fibers become activated in response to intense mechanical, chemical, or thermal stimuli from the periphery and transmit the pain signals into the spinal cord. Two major systems exist: First, fibers from lamina V of the spinal cord projecting to the lateral hypothalamus and somatosensory cortex seem to be responsible for the sensory discriminative elements of the noxious signal; and second, fibers originating in lamina I of the spinal cord and projecting to the hypothalamus, amygdala, and anterior cingulate area mediate the affective component of pain. The signal can be modulated locally, within the spinal cord or from higher supraspinal areas, all of which can either facilitate or inhibit transmission of the signal.[22-29]

Neuropathic pain is initiated or caused by a primary lesion or dysfunction in the nervous system and can be caused by numerous insults, including infection, metabolic disease, and physical trauma. The condition is characterized by ongoing and/or evoked types of pain in an area of sensory dysfunction. Subtle changes in the sensitivity of the pain circuitry, specifically increased sensitivity of the peripheral pain receptors and fibers and increased excitability of nerve cells in the central nervous system, termed peripheral and central sensitization, respectively, are thought to be an underlying mechanism for the allodynia (previously non-noxious stimuli perceived as painful) and hyperalgesia (noxious stimuli producing exaggerated and prolonged pain) observed in some patients with neuropathic pain. (See reviews by Coderre and Katz[30] and Jensen and colleagues.[31])

Effective pharmacotherapies to manage neuropathic pain include tricyclic antidepressants (TCAs); newer antidepressants, such as duloxetine and venlafaxine; the lidocaine patch 5%; anticonvulsants, such as gabapentin and pregabalin; opioids; and tramadol[19,32,33] — all of which are reflective of not only the heterogeneity of the patient population, but also the varying underlying pathophysiologies of neuropathic pain.

Of the pharmacotherapeutic strategies used to treat neuropathic pain, some are based on empirical evidence, whereas some are derived from clinical trials. In general, evidence-based treatment algorithms for neuropathic pain have been based on data generated from small and, in some cases, poorly designed clinical studies or anecdotal evidence.[34-36] Furthermore, there is a scarcity of clinical trials assessing efficacy and side effects in direct comparisons of one (analgesic) drug with another.[37] An alternative is to estimate the relative efficacy and safety with numbers needed to treat (NNT) and numbers needed to harm (NNH).[38-40]
{mospagebreak}

 Materials and Methods

The present algorithm is based on a recent review of 105 randomized, placebo-controlled clinical trials,[37] with the inclusion of 5 more recent clinical trials. The NNT is defined in the present context as the number of patients needed to treat with a specific drug to obtain 1 patient with a defined degree of pain relief. As such, pain relief was defined as 50% pain relief and was calculated as the reciprocal of the absolute risk difference.[37,41,42] Only when the relative risk was statistically significant was the NNT calculated. The NNH was defined as the number of patients needed to be treated for 1 patient to withdraw from the trial due to adverse effects. Both the NNT and NNH were calculated as the reciprocal of the 95% confidence interval for the absolute risk difference on the basis of a normal approximation. Combined measures of NNTs were generated by pooling raw data, when available, from clinically homogeneous trials.[43]

Studies used in the construction of the treatment algorithm consisted of randomized, placebo-controlled, double-blind trials with at least 10 patients, published in English in peer-reviewed journals, and evaluated for clinical heterogeneity with L'Abbé plots.[43,44] A complete description of the search and selection criteria of appropriate clinical trials can be found elsewhere.[37]

Results

The treatment algorithm for neuropathic pain is based on the results from 110 randomized, double-blind, placebo-controlled clinical trials that met the inclusion (and exclusion) criteria for this analysis. An active placebo was used in 5 of the trials; 59 trials employed a crossover design; and 51 used a parallel design. A complete bibliography can be found elsewhere.[37] The drugs used to treat the neuropathic pain were divided into 5 categories and are shown in the Table .

Twenty-six trials were conducted with antidepressants composed of TCAs, including amitriptyline; selective serotonin reuptake inhibitors (SSRIs), including fluoxetine; serotonin noradrenaline reuptake inhibitors (SNRIs), including venlafaxine; and dopamine/noradrenaline reuptake inhibitors (DNRIs), including bupropion. In addition, 2 trials with duloxetine have been published.[45,46] TCAs were found to be efficacious in treating central poststroke pain, PHN, both painful diabetic and nondiabetic polyneuropathy, and postmastectomy pain syndrome, but ineffective in relieving the pain associated with spinal cord injury, phantom limb pain, and the pain associated with HIV neuropathy. The negative results may be due to methodological considerations (eg, dose range, inclusion criteria[47,48]). Across the various neuropathic pain conditions in which TCAs were used, NNT (95% confidence interval) ranged from 2.1 (1.8-2.6) to 3.1 (2.2-5.5); the combined NNT from 4 trials of SNRIs in painful polyneuropathy was 5.1 (3.9-7.4); and the NNT for SSRIs was high at around 7. The balanced serotonin and noradrenaline reuptake inhibitors trended for greater relief than that produced following the mainly noradrenergic drugs (NNT: 2.1 [1.8-2.6] vs 2.5 [1.9-3.6], respectively).[37,49] The NNH for TCAs was 14.7 (10.2-25.2) and for SNRIs 16.0 (10.9-29.5). The NNH for SSRIs was not calculated because the relative risk for withdrawal was not statistically significant.

Antiepileptic drugs, including gabapentin, pregabalin, and topiramate, were assessed in 41 clinical trials, including 2 recently published pregabalin trials.[50,51] The NNT for gabapentin across all types of neuropathic pain and incorporating all doses reported was 5.1 (4.1-6.8). However, the NNT became more favorable, dropping down to 3.8 (3.1-5.1), when the study by Gorson and colleagues,[52] which used only 1 dose of 900 mg/day, was omitted along with the doses of gabapentin other than the 2400-mg dose in a study by Rice and Maton[53] and a study by Serpell[54] on mixed neuropathic pain. The combined gabapentin NNH for withdrawal, across all neuropathic pain conditions, was 26.1 (14.1-170).[37]

Pregabalin was approved in 2004 for the treatment of pain associated with peripheral diabetic neuropathy (PDN) and PHN. In doses ranging from 300 to 600 mg from 7 trials, the combined NNT was 3.7 (3.2-4.4), similar to that of gabapentin in the higher doses. The NNH was 7.4 (6.0-9.5).

Three trials that assessed the effects of topiramate in a total of 1259 patients with PDN receiving doses up to 400 mg did not demonstrate significant pain relief.[55] In contrast, a study reported by Raskin and colleagues[56] of 323 patients with PDN found a significant reduction in pain but yielded a high NNT of 7.4 (4.3-28.5). The difference between these studies may be attributable to the high placebo response observed in the 3 trials and a modification of the study design in the fourth trial to reduce the placebo response.[56] The combined NNH for withdrawal was 6.3 (5.1-8.1).

Opioids, including morphine, oxycodone, and tramadol, were assessed in 11 studies. Morphine significantly reduced pain in patients with PHN, PDN, or phantom limb pain, yielding an NNT of 2.5 (1.9-3.4). The effect of oxycodone on PHN and PDN was tested in 3 studies with doses ranging from 20 to 80 mg, and generated an NNT of 2.6 (1.9-4.1), similar to morphine. The combined NNH for morphine and oxycodone was 17.1 (10-66).

A combined NNT of 3.9 (2.7-6.7) was calculated from 3 studies of tramadol in doses ranging from 200 to 400 mg in patients with painful polyneuropathy or PHN. The combined NNH was 9.0 (6.0-17.5).

NMDA (N-methyl-D-aspartic acid) antagonists included dextromethorphan, memantine, and riluzole, and the results from 7 clinical trials are mixed. For example, 2 studies of high-dose dextromethorphan (400 mg) in patients with painful diabetic polyneuropathy showed significant pain relief and an NNT of 2.5 (1.6-5.4),[57,58] but showed that it was ineffective in treating PHN.[57,58] The NNH was 8.8 (5.6-21.1). Memantine (20-55 mg) was also ineffective in treating PHN,[58,59] PDN,[57,58] or phantom limb pain.[60,61] As the relative risk was nonsignificant in these trials, an NNH for withdrawal was not calculated.
{mospagebreak}

 Discussion

Despite a number of obvious limitations to using NNTs derived from different clinical trials, including issues with regard to the inclusion criteria for clinical trials used in this analysis and the study designs, the retrospective nature of the calculation given the different cutoff points for the definition of pain relief, and the risk of overestimating the efficacy of a drug if negative trials are not published, the NNT provided a clinically meaningful measure of effect and risk of each drug.

Formulating an evidence-based treatment algorithm for neuropathic pain should be based on a number of factors, including a consistent outcome in randomized controlled studies of high quality, a high degree of pain relief and superiority to existing pharmacotherapies, a persistence of the analgesic effect with few and only mild side effects, a significant effect on the quality of life, and — finally — low cost. As there are no head-to-head comparison studies of existing and new compounds with the same set of primary and secondary endpoints, a treatment algorithm that is based on a combination of all of the above factors is not possible.

The flow of such an algorithm is dependent on the particular set of criteria used, such as pain relief, measures of quality of life, and study design. Including long- and short-term side-effect and risk profiles into the acceptable criteria for the algorithm necessitates consideration of the important and occasionally dangerous side effects of TCAs and opioids. Given these considerations, the treatment algorithm for neuropathic pain may be as shown in Figure 2.

Proposed treatment algorithm for neuropathic pain. Adapted with permission Finnerup and colleagues.[37]
TCA, tricyclic antidepressants; SNRI, serotonin/noradrenaline reuptake inhibitors.
     

This algorithm does not include trigeminal neuralgia, for which carbamazepine and oxcarbazepine are the first drug choices.[33,37] For a patient presenting with PHN or a peripheral focal neuropathy, a first-line pharmacotherapy would consist of a topical analgesic, such as the lidocaine patch 5%. Other neuropathies should be treated with TCAs, gabapentin, or pregabalin, or if TCAs are contraindicated, SNRIs. Although TCAs have a lower NNT compared with gabapentin and pregabalin, these differences may partly be attributable to differences in study design (crossover vs parallel-group design) and the way that NNTs are calculated (with 50% pain reduction, which is the case for all pregabalin studies, or improvement on a pain relief scale, which is the case for many TCA studies) and not directly to issues of efficacy. Additionally, these AEDs (gabapentin and pregabalin) lack serious adverse effects compared with TCAs. Newer antidepressants, such as SNRIs (eg, duloxetine), which have fewer side effects compared with TCAs, are another alternative. Opioids (including tramadol) can be considered as second- or third-line drugs in the algorithm. These drugs do have an effect on neuropathic pain,[37,74] but are not considered first-line drugs due to issues about dependence, cognitive impairment, tolerance issues, and possible hormonal problems.[75] Although the combination of drugs targeting separate pain mechanisms has the theoretical appeal for improved pain relief, the evidence for this is still lacking, except for the combination of gabapentin with venlafaxine.[76] Although a general treatment algorithm may be proposed, each treatment has to be individualized to the single patient, taking into account all comorbidities and drug interactions. Antidepressants may thus be the first drug choice in patients with depression, and some TCAs and calcium channel alpha2delta agonists may be considered in patients with sleep disturbances; pregabalin may be the first choice in patients with anxiety; and opioids, including tramadol, may be first-line in episodic pain, or in patients with concomitant cancer-related nonneuropathic pain.

Clearly, improvements can be made to this treatment algorithm, and as more evidence is generated from high-quality, randomized, controlled, head-to-head comparative clinical trials, this treatment algorithm can be further refined to ultimately benefit the patient with neuropathic pain.