Pharmacologic treatment of pain in polyneuropathy

Antidepressants.

From 10 studies, including approximately 300 patients on active treatment, the NNT for tricyclic antidepressants (TCA) is 2.6 (95% CI, 2.2 to 3.3). In one study, the TCA is used in combination with fluphenazine.8 In many of the trials, the antidepressants were dosed according to effect and side effects. This may result in an underestimation of the potential effect of the tricyclics, because with these drugs, side effects are often bothersome and dose–effect or plasma drug-concentration–effect relations have been found in some studies.9-10⇓ In two studies on imipramine in diabetic neuropathy,11-12⇓ the dose was adjusted to obtain the optimal plasma concentration of imipramine plus its active metabolite desipramine around 400 nM.13 The target concentration was obtained in 16 of 19 patients in the first study with a median dose of 200 mg/day (range 25 to 350 mg/day). From the original data of that study, a NNT of 1.4 (95% CI, 1.1 to 1.9) is calculated (i.e., a value below the lower 95% confidence limit for the entire group of studies). This strongly indicates that the effect can be increased by this dosage policy.

There is no difference in NNT between TCA with balanced reuptake of serotonin and noradrenaline (imipramine, amitriptyline, clomipramine) with NNT 2.7, and TCA with relatively selective noradrenaline reuptake (desipramine, nortriptyline, maprotiline) with NNT 2.5, which is in accordance with one study with a face-to-face comparison,14 but in contrast to others.15-16⇓ However, the dosage policy may be particularly important in this comparison, as relatively selective drugs are better tolerated than the balanced drugs, and the potential effect may therefore be more fully achieved for these drugs than for the balanced compounds.

The data from several of the controlled studies indicate that tricyclics are effective for both steady and lancinating or brief pains,11,12,15,17⇓⇓⇓ whereas it is more difficult to judge if these drugs also relieve touch-evoked pain. It is an inherited problem with these studies that none of them addressed the issue of an effect on different pain types, but only showed that patients with the different types of pain were relieved of pain in general.11

The selective serotonin reuptake inhibitors (SSRI) are a relatively new class of antidepressants. They differ from classic tricyclic antidepressants in their specific inhibition of presynaptic reuptake of serotonin, but not of noradrenaline, and their lack of postsynaptic–receptor-blocking effects and quinidinelike membrane stabilization. In two out of three studies on SSRI in painful diabetic neuropathy, there was a significantly better effect of the SSRI than of placebo (see the table on the Neurology Web site).11,14,18⇓⇓ There is no obvious explanation for the difference in effect between the different SSRI. Hitherto unknown differences in basic pharmacology may be responsible for some of the discrepancies. With respect to fluoxetine, the failure to find an effect may relate to the pharmacokinetics of fluoxetine hampering the crossover design of the study.14 Fluoxetine has an active metabolite with a very long half-life. The individual NNT from the studies showing a significant effect is 2.9 (paroxetine) and 7.7 (citalopram), and the combined NNT for all three studies is 6.7 (95% CI, 3.4 to 435). Paroxetine seems to relieve both steady and lancinating pain.11

From the results of these and other studies in diabetic neuropathy, it may be suggested that drugs with a balanced inhibition of serotonin and noradrenaline but without the postsynaptic and quinidinelike effects of the TCA could have similar effect as the tricyclics and, at the same time, be better tolerated. One such drug, venlafaxine, has been marketed for the treatment of depression, but there are still no controlled trials on this drug in painful polyneuropathy.

Ion channel blockers.

Lidocaine is the prototype of an unspecific sodium channel blocker. More than 10 years ago, it was reported that infusions of lidocaine relieved painful diabetic neuropathy, and the data from that study show a favorable NNT of 3 (95% CI, 1.5 to 10).19 Lidocaine is not convenient for chronic treatment, however, because it cannot be dosed orally. Accordingly, studies on its oral analogue mexiletine have been performed. Two of the studies found a better effect of mexiletine than of placebo,20-21⇓ whereas four other studies either reported no effect22-25⇓⇓⇓ or only an effect on secondary but not primary effect variables (see the table on the Neurology Web site).22 Dichotomous data are reported in one negative study and in one study favoring mexiletine.23,21⇓ It is important to note that only the highest of three dose levels tested in that study was better than placebo. The NNT for this dose level (675 mg/day) is 10 (95% CI, 3.0 to ∞). Data on the effect on different pain categories are not available. The combined NNT for the two studies with dichotomous data were 38 (95% CI, 5.5 to ∞).

In a review on mexiletine in diabetic neuropathy, reference is given to three Japanese studies reporting a significant effect of mexiletine. From the data in the review, a combined NNT for all three studies of 4.1 (95% CI, 2.9 to 6.9) is calculated.26 These data are not included in the current analysis because we are unable to review them.

Phenytoin and carbamazepine are widely used as anticonvulsants. These drugs exert their membrane-stabilizing properties by blocking sodium channels unspecifically; therefore, they reduce neuronal excitability in sensitized C-nociceptors. In painful diabetic neuropathy, carbamazepine has an NNT of 3.3 (95% CI, 2.0 to 9.4) and phenytoin an NNT of 2.1 (95% CI, 1.5 to 3.6).5 However, there is only one trial on each drug.27-28⇓ A second placebo-controlled study with phenytoin failed to demonstrate a significant effect of this drug,29 and it is not possible to derive dichotomous data from that study. The carbamazepine dose used (titration from 200 to 600 mg/day) is rather low and it is possible that the efficacy could be increased by using higher doses.

Lamotrigine is a new anticonvulsant, which acts by stabilizing the slow inactivated conformation of a subtype of sodium channels. It is possible that this mechanism suppresses the neuronal release of glutamate. An uncontrolled study in painful diabetic neuropathy30 indicates that lamotrigine will in fact relieve neuropathic pain. Positive preliminary results from the subsequent double-blind controlled trial were reported in abstract form.31

Gabapentin is a novel anticonvulsant with an unknown mechanism of action. It does not interact with GABA receptors or GABA metabolism. It has been suggested that it binds to an αδ₂ subunit of N-type calcium channels on neurons. In an adequately designed study with 165 patients with painful diabetic neuropathy, gabapentin at a dose of 3600 mg/day was compared with placebo and the NNT was 3.7 (95% CI, 2.4 to 8.3).32 There are no data on the effect on individual pain types in that study. Another study with a much lower dose (900 mg) showed no effect.33 The combined NNT is 4.1 (95% CI, 2.7 to 8.2).

NMDA antagonists.

The evidence for involvement of excitatory amino acids in neuropathic pain has also prompted studies on drugs with an NMDA-antagonistic effect. The low affinity NMDA channel blocker dextromethorphan relieved diabetic neuropathy with an NNT of 1.9 (95% CI, 1.1 to 3.7), but the number of patients is small (see the table on the Neurology Web site).34 It was reported that in diabetic patients, there were no differences between good and poor responders with respect to pain quality.

Opioids.

There are no studies on pure opioids in painful polyneuropathy. Tramadol is an analgesic drug, assumed to act through monoaminergic and opioid mechanisms. The monoaminergic effect is shared with the tricyclic antidepressants. Furthermore, development of tolerance and dependence during long-term tramadol treatment appears to be uncommon, and tramadol seems to have a low abuse liability. Therefore, tramadol may be an alternative to strong opioids. Two studies tested the drug in painful diabetic polyneuropathy35 and painful polyneuropathy of different etiologies.36 Tramadol was superior to placebo in both studies with NNTs of 3.1 and 4.3, and the combined NNT was 3.4 (95% CI, 2.3 to 6.4). In the latter study, nearly 80% preferred tramadol to placebo. It was shown that tramadol, in addition to its relief of ongoing pain, reduced touch-evoked pain and experimentally induced mechanical hyperalgesia.

Levodopa.

Dopamine agonists inhibit noxious input to the spinal cord. Levodopa, a dopamine precursor, may have a similar effect or may act as a precursor to noradrenaline and modulate pain via noradrenergic mechanisms. One study showed that levodopa reduced pain in diabetic neuropathy.37 In diabetic neuropathy, NNT was 3.4 (95% CI, 1.5 to ∞). This study did not differentiate the type of pain that responded to the treatment.

Capsaicin.

Capsaicin, an alkaloid derived from chilies, depletes the neurotransmitter substance P from sensory nerves. A further possible mechanism of action of capsaicin in peripheral neuropathic pain is degeneration of epidermal nerve fibers.38

Topically applied capsaicin cream showed a significant effect in three out of five studies in diabetic neuropathy,39-43⇓⇓⇓⇓ with NNT in the positive studies from 2.5 to 4.9 and a combined NNT for all studies of 5.9 (95% CI, 3.8 to 13). One caveat of all but one of these studies is an inadequate blinding owing to the burning skin sensation induced by capsaicin. One study used burning nicotinamide cream as a control and did not find any effect.43 NNT calculation on this study alone is not meaningful, because a higher fraction of patients responded to placebo than to capsaicin (see the table on the Neurology Web site).

Although topical capsaicin may have few side effects, it may be less convenient in many patients, owing to the frequent application time on the entire painful area.

α-Lipoic acid.

There is some evidence that free-radical mediated stress is implicated in the pathogenesis of diabetic neuropathy because it induces hypoxia,44 and that this can be treated by free-radical scavengers such as α-lipoic acid. A trial of IV infusions of α-lipoic acid in daily doses of 1200 mg, 600 mg, and 100 mg, pain was reduced by this drug.45 For all doses together, the NNT was 5.6 (3.2 to 24) and for the two highest doses alone 4.3 (2.6 to 11.3). There are still no trials on chronic oral dosing of α-lipoic acid and therefore this treatment modality is not included in the subsequent comparisons.

Interdrug comparison.

The NNT for the different drugs or drug classes are shown in figure 2 arranged after the value of NNT. The NNT values fall into three groups, one with values around 3, including TCA, anticonvulsant drugs with sodium and calcium channel blockade, tramadol, l-dopa, and dextromethorphan; one with values around 6, including SSRI and topical capsaicin; and mexiletine, with an NNT of 38.

• Download figure
• Open in new tab
• Download powerpoint

Figure 2. Numbers needed to treat (NNT) with 95% CI for different drug classes or drugs in the treatment of pain in polyneuropathy, with treatments arranged according to NNT value. The total number of patients on active and placebo treatment from which the NNT is calculated is given (n_active/n_placebo). For phenytoin and mexiletine, only part of the current data could be used to calculate NNT. TCA = tricyclic antidepressants; SSRI = selective serotonin reuptake inhibitors.

Strength of the NNT method.

Painful neuropathies constitute a series of etiologically different conditions within the spectrum of neuropathic pains. Treatment of painful neuropathies has until recently not been based on controlled clinical trials conducted on a sufficient number of patients. The introduction of NNT, however, permits sampling of large patient populations exposed to pharmacologic treatments and accordingly, the efficacy of treatments can now be judged more precisely. The size of the trials or the total number of exposed patients in meta-analysis is of major importance in the estimation of the size of treatment effect.46

Limitations of the NNT method.

There are several shortcomings with the NNT method which need to be considered. Firstly, a prerequisite for performing meta-analysis is a grouping of different studies; such studies may not necessarily use the same criteria and threshold for efficacy. Because the calculated NNT value is dependent upon the obtained placebo response, the NNT is relatively robust with respect to differences in the threshold at which the goal for efficacy is considered achieved. In a particular study, a minor change in threshold will change the frequency of obtained goal (pain relief) equally for both placebo and active compound.46 Secondly, the analysis of NNT is carried out on dichotomized data in which some information is lost (e.g., the number of patients obtaining 25% pain relief). Nevertheless, in daily clinical practice it may be a useful guide for physicians and their patients to know the NNT for a series of drug classes. Thirdly, in analyzed studies, different diagnostic entities are lumped together (e.g., diabetic and nondiabetic neuropathies), which may obscure the results. However, in two studies in which separation of diabetes and non-diabetes was possible, there were no differences in NNT values for amitriptyline and maprotiline,16 and in NNT values for tramadol36 between diabetic and nondiabetic subjects. Finally, in painful neuropathies, it is clear that several mechanisms may be involved in generating pain, which may compromise the NNT value. However, even in existing single trials, several mechanisms may be in operation. Future studies may provide a better understanding of this matter if NNT are determined for different categories or different mechanisms of pain.

Interdrug efficacy comparison.

Based on controlled clinical trials on an adequate total number of patients, treatment of painful polyneuropathy with TCA is now based on evidence and the efficacy is quite good with NNT between 2 and 3. Even in patients with less than 50% pain relief, there is usually some pain relief. Anticonvulsants with either sodium channel blocking properties or with a presumed action at calcium channels have a similar effect. However, the documentation for an effect is more scarce for these types of drugs. Treatment with l-dopa and the NMDA-receptor antagonist dextromethorphan appears to be quite efficacious, but the NNT value obtained for these drugs must be considered as point estimates, which are not necessarily near the “true” value. Finally, it is documented that SSRI and topical capsaicin result in a minor pain relief.

Interdrug comparisons of efficacy in the current setting may be inaccurate owing to both different dosing policy in different trials and interdrug differences in tolerability. In general, flexible dosing will tend to underestimate the efficacy of most drugs, but more so for drugs with many side effects. Flexible dosing may, conversely, better reflect how a drug will perform in clinical practice. Therefore, the current comparison of efficacy may be valid for clinical purposes.

Drug action and pain mechanism.

It is well established that patients with neuropathic pain including painful polyneuropathy may experience different types of pain, such as stimulus-independent constant and paroxysmal pain, and stimulus-dependent pain, which includes touch- and thermal-evoked pain.2 In the same patient, one or several types of pain may be present, which may contribute to the total pain experienced and account for the interindividual variability. These different types of pain may be caused by distinct pathophysiologic mechanisms, such as spontaneous activity of damaged C-nociceptors, increased sensitivity to noxious and non-noxious stimulation, sympathetic hyperactivity, or a loss of central inhibition.2 The suggestion that neuropathic pain should be classified by mechanism and that drugs should be judged based on their modes of action appears to be a useful and rational approach. However, this has only been done to a limited extent.7 In a trial with tramadol, it was shown that pain relief was accompanied by a concomitant and parallel relief of touch-evoked pain (a presumed Aβ-mediated function).36 In this group of patients with painful polyneuropathy, it was suggested that central hyperexcitability sensitive to opioids was present.

It is clear that a drug with one single mechanism of action is not likely to relieve all patients or all types of pain in a single patient. One possible reason for the apparent superiority of TCA may be related to their multimodal mechanisms of action (e.g., stimulation of endogenous pain-modulating networks via inhibition of biogenic amine reuptake, NMDA-receptor antagonist effect, and blockade of ion channels). In contrast, the current search for selectively acting drugs may not lead to more efficacious drugs overall, though these drugs may be better tolerated. However, together with increased knowledge of pain mechanisms and development of methods to determine their contributions to pain in each patient, new selective drugs may enable us to tailor the pharmacologic treatment to the individual patient. In addition, understanding the mechanisms of pain, knowing NNT and the equivalent numbers needed to harm may help in formulating algorithms for pain treatment. It should be emphasized that general algorithms for neuropathic pains starting with local treatments followed by systemic treatments are unlikely to be successful for all types of neuropathic pains. For example, although this approach seems rational in patients with postherpetic neuralgia,47 a local treatment will rarely be manageable in patients with polyneuropathy exhibiting an extensive glove and stocking-like pain distribution.

A mechanism- versus etiology-based classification of pain2 is supported by the current and some previous results.7 In terms of NNT, the efficacy for TCA is similar regardless of causative agent for the painful condition—diabetes, herpes zoster, traumatic nerve injury, or stroke. This finding indicates that etiology does not influence the overall endpoint pain. Alternatively, NNT may be too simple a measure for distinguishing between etiologies or mechanisms.7 It is noteworthy that in a large well designed study of polyneuropathy due to HIV infection,23 TCA failed to relieve pain. The reason for this failure is not clear. The symptoms in this group of patients may be different from that seen in other neuropathies. Alternatively, a subclinical myelopathy may be present and may account for the failure to see an effect of TCA on pain. The finding of efficacy of α-lipoic acid in painful diabetic neuropathy45 also suggests that some etiology-specific treatments may evolve.

Recommendation for drug choice in painful polyneuropathy.

Based on available data, a recommendation for drug choice in painful polyneuropathy can be suggested. If one only considers the amount of evidence and efficacy, the obvious choice is TCA. These drugs may also be less sedating than the slightly less effective but well documented gabapentin. There are, however, certain problems with the use of TCA. Firstly, TCA need therapeutic drug monitoring to achieve optimal effect and avoid toxicity unless a sufficient pain relief is obtained with a low dose, i.e., imipramine or amitriptyline 25 to 75 mg/day. Secondly, there may be increased mortality supposedly from sudden arrhythmia in patients with ischemic heart disease; in patients with recent myocardial infarction, cardiac incompensation, and arrhythmia, TCA should not be used at all. In such cases, gabapentin may be the first choice and is an obvious second choice if TCA cannot be tolerated owing to side effects. The third drug to be considered is tramadol, which is as effective as gabapentin and can be given in effective doses right away. However, it is an opioid drug and development of tolerance and dependence must be kept in mind. The efficacy of carbamazepine is probably similar to that of gabapentin and tramadol, but evidence is more sparse; it is sedating and cannot be used in some cardiac arrhythmias. Thus, the drugs of first choice are the TCA followed by gabapentin, tramadol, and carbamazepine.