A qualitative systematic review of head-to-head randomized controlled trials of oral analgesics in neuropathic pain

1University of Toronto, Toronto; 2Queen’s University, Kingston, Ontario; 3Dalhousie University, Halifax, Nova Scotia Correspondence and reprints: Dr C Peter N Watson, 1 Sir Williams Lane, Toronto, Ontario M9A 1T8. Telephone 416-239-3494, fax 416-239-6365, e-mail peter.watson@utoronto.ca “Food and Drug Administration (FDA) regulations should require that new drugs be compared not just with placebos but with old drugs for the same conditions. Approval would depend on whether the new drug adds something useful in terms of greater effects, greater safety, fewer side effects or substantially greater convenience...If I could choose only one of the reforms I am suggesting it would be this one. There is an ethical issue here too. It is wrong to compare a new drug with a placebo if there is an effective drug already on the market.” (Marcia Angell [1])

(Figure 1), are generally less responsive than others.Some peripherally generated problems can have a central component (postherpetic neuralgia and brachial plexus avulsion) and central sensitization may occur with other PNP disorders, making the distinction between peripheral and central pain states less clear.NP has different qualities, often in the same patient.Steady -often burning -pain, shock-like jabbing pain and pain from touch (allodynia) may all occur, indicating possible different pain mechanisms for such qualities and supporting a mechanism-based approach (6).
Thirty years ago, there was little science behind the few drugs used for NP conditions.Amitriptyline, phenothiazines and the combination of amitriptyline and fluphenazine were widely used based on uncontrolled and observational studies (7)(8)(9)(10).With the evolution of trial methodology and clinical research, current data suggest that the major categories of drugs shown to be useful for NP by RCTs are antidepressants, anticonvulsants, opioids and the emerging field of cannabinoids.There are many such studies in NP (11); most are RCTs that show a moderate effect of a drug compared with placebo.The utility of these drugs in clinical practice based on these RCTsthat is, the external validity or generalizability of these findings -has been a neglected issue until recently (12).Thus, most RCTs do not provide measures of clinical meaningfulness, such as effect size, number needed to treat (NNT) (13) and number needed to harm (NNH), which makes comparisons between specific agents difficult (12).NNT and NNH values have been generated for some of these RCTs by others (11).
Since the first head-to-head RCT discussed here (published 20 years ago) (14), there have been few subsequent trials of this nature in NP comparing drugs in different analgesic classes or different drugs within an analgesic class.A recent review regarding algorithm development (11) identified only 13 headto-head RCTs of 105 NP RCTs, but provided useful comparative data in their absence in the form of NNT and NNH values (Table 2).The advantage of head-to-head RCTs in NP is that they enable the clinician, struggling with the problem of the most appropriate drug selection, to know whether a new drug is equal to, noninferior to or better than a standard therapy (such as a tricyclic antidepressant [TCA]) and whether there are differences in adverse effects.The only other way to determine relative efficacy and safety is to compare measures such as NNT and NNH data from placebo-controlled trials.However, these measures usually compare studies with differences in experimental design, number of subjects, trial duration, inclusion criteria, pain intensity, outcome measures and data analyses; consequently, they can only be, at best, an approximate comparative measure.
Head-to-head trials may also reveal data about pain mechanisms, in that one drug may work in a particular subgroup.For example, it is possible that one of the drugs compared may relieve predominantly burning pain, electric shock-like pain or allodynia, revealing a mechanism-based difference in drug responsivity (6).Therapeutically, this may be valuable in indicating drugs for combination therapy targeting different pain mechanisms.Further subgroups, such as age, sex, race and diagnostic categories, may also reveal differences between drugs.
The purpose of the present article is to review the data obtained from head-to-head RCTs of oral pharmacotherapeutic agents in chronic noncancer NP to determine whether there is clear evidence for drug selection choices.Trials were evaluated according to the quality criteria of Jadad et al (15).To be included, trials were required to score at least 3/5 on this rating scale to ensure a standard of quality; this required randomization and double-blinded conditions, with a control group and accounting for withdrawals.A maximum score of 5 additionally indicated that the RCT described methods of blinding and randomization.Measures were sought in each trial, such as effect size, percentage of patients with 50% or greater improvement, NNT, NNH and number needed to quit (NNQ), as a means of determining clinical meaningfulness.The search excluded non-English publications, trials that compared topical agents, intravenous studies, acute NP conditions, NP with cancer, and the NP conditions of trigeminal neuralgia and complex regional pain syndrome.

Comparative trials of different analgesics for NP
Comparative trials of different analgesics for NP are summarized in Table 4.Most of the 17 comparative trials of different analgesics for NP compared a potential analgesic drug with a TCA, except for the comparisons of venlafaxine with gabapentin (31), gabapentin with morphine (33), nabilone with dihydrocodeine (38), morphine with mexiletine (39) and gabapentin with oxycodone (40).
The 1999 diabetic neuropathy study by Morello et al (29) comparing gabapentin with amitriptyline found no significant difference (NSD) in either pain relief or side effects between drugs.This study has been criticized because of the low dose of gabapentin used compared with the placebo-controlled trials in postherpetic neuralgia and painful diabetic neuropathy (52), and also the lack of placebo control.However, the degree of pain relief achieved (52% with gabapentin and 67% with amitriptyline) suggests that the statistically equal effect did not mean that both drugs were ineffective.
Results from an industry-sponsored trial comparing amitriptyline with pregabalin have been posted on the FDA website (26); however, complete details of this trial are not available for review.This unpublished Pfizer Protocol 1008-040 (www.accessdata.fda.gov/scripts/cder/drugsatfda)(26) apparently studied painful diabetic neuropathy and compared placebo, amitriptyline 75 mg and pregabalin 600 mg in a parallel design.An FDA review found that both drugs produced a 2.8point reduction in pain intensity (versus 1.8 with placebo), but  the reduction with pregabalin was not significantly different (the starting pain intensity in the pregabalin group was 0.5 points higher than in the amitriptyline group).In this trial, amitriptyline, but not pregabalin, was superior to placebo.The FDA review goes on to say that the above results came from a less conservative 'last observation carried forward' analysis but, with the more conservative 'baseline observation carried forward' analysis, neither pregabalin nor amitriptyline differed statistically from placebo with respect to mean pain intensity at the end point.The 2000 placebo-controlled RCT by Graff-Radford et al (30) for postherpetic neuralgia compared amitriptyline with the phenothiazine fluphenazine, and the combination of the two drugs, and found NSD in pain with fluphenazine used  alone or in combination with amitriptyline; however, it found a statistically significant difference in pain with amitriptyline.This was an important trial because this drug combination was one of the original treatments for NP (9,10) and the combination of fluphenazine and nortriptyline was previously shown to be more effective than placebo (17).It is apparent from this RCT that amitriptyline is the analgesic component and, furthermore, that adding fluphenazine increased sedation without adding any analgesic benefit.In 2007, Khoromi et al (35) studied chronic sciatic pain with placebo, morphine, nortriptyline and a combination of these two drugs, and found NSD with any of the treatments.This trial used between 15 mg and 90 mg of morphine per day with flexible dosing to maximal tolerable dose.Our own experience (53,54) indicates that this may be an inadequate dose of morphine to alleviate this type of refractory NP, which may explain the lack of efficacy of opioids in sciatic pain in this RCT.
Gilron et al compared gabapentin, morphine and the two drugs in combination in one trial (33) and then compared nortriptyline, gabapentin and the two drugs in combination in a subsequent trial (41).In both of these trials, the two-drug combination demonstrated superior efficacy to each single agent.However, NSDs were observed between either of the single agents.

Comparative trials of different drugs within an analgesic class used for NP
Table 5 summarizes comparative RCTs of different drugs within an analgesic class.The 10 studies (42-51) listed in Table 5 are all head-to-head RCTs of different antidepressants conducted usually with the aims of discovering the most effective drugs in this class, and potentially to inform their mechanism of action.No trials of different drugs within other classes of analgesics for NP such as opioids, gabapentinoids (gabapentin and pregabalin) or cannabinoids met our criteria.

General summary
Several general observations can be made on the basis of these head-to-head studies.Of the RCTs comparing different potential analgesics, there is no evidence supporting the efficacy of the benzodiazepine lorazepam ( 14), the phenothiazine fluphenazine (30) or the sodium channel-blocking agents mexiletine (39) and carbamazepine (27) for NP (trigeminal neuralgia was excluded).There is no evidence for the superiority of gabapentinoids over TCAs either regarding pain or adverse effects, although the nature of the latter differs with the two agents (26,29,36).There are nonsignificant trends suggesting the superiority of opioids over TCAs and gabapentinoids (32,33).
With regard to antidepressant comparisons, the TCAs amitriptyline, nortriptyline, desipramine and imipramine do not appear to differ in analgesic efficacy.The noradrenergic TCA desipramine is at least as effective as amitriptyline in treating diabetic neuropathy and postherpetic neuralgia (46,51).There is some indication that the noradrenergic antidepressant maprotiline, which has a tetracyclic structure, is less effective than amitriptyline.This has been shown by two studies (47,48) that were similar in design and covered two different conditions -postherpetic neuralgia and painful neuropathy.Unlike imipramine, the nontricyclic agent mianserin was not efficacious (45).There is some evidence that the serotonergic drug clomipramine may be more effective than imipramine and desipramine (42,44) and that the selective serotonin reuptake inhibitor (SSRI) paroxetine may be effective (43); however, these studies were small, may have been inadequately powered, did not state clinical meaningfulness and could have been vulnerable to type 1 error.There is no good evidence for the effectiveness of the SSRI fluoxetine in treating NP (46,51).The preponderance of evidence from these RCTs is supportive of the greater efficacy of noradrenergic and noradrenergic/serotonergic TCAs in NP.The selective serotonin/norepinephrine reuptake inhibitor venlafaxine appears inferior to imipramine for meaningful relief (50).There is some evidence of individual variability in responsivity when individual antidepressants are compared using head-to-head RCTs (47)(48)(49), arguing for sequential trials of different antidepressants in individuals.

DISCUSSION
There are a limited number of head-to-head comparative RCTs of analgesics in NP published between 1988 and 2009.Some of those reviewed here were quite small and may have inadequate statistical power.Within our search criteria there were 17 studies of drugs in different analgesic categories and 10 of analgesics within a category, which were comparisons of different antidepressant drugs.There were seven RCTs using a gabapentin comparator, one using pregabalin, two using a selective serotonin/norepinephrine reuptake inhibitor antidepressant (venlafaxine) and one using the anticonvulsant lamotrigine.No consistent superiority of the latter drugs was found.There were no studies comparing different opioids, gabapentinoids or cannabinoids.All published RCTs were nonindustry funded.One industry-funded study submitted to the FDA (26) has not been published.Most head-to-head RCTs used a crossover design.Most NP RCTs were performed in postherpetic neuralgia patients, painful diabetic neuropathy patients or both, and this may limit their external validity (generalizability) to other NP conditions (12).There are limited data regarding clinical meaningfulness in many of these comparative trials, which is also problematic for determining their external validity (12).

Comparative RCTs of different analgesics
From the limited data in these head-to-head RCTs of different analgesics in NP, there is no evidence supporting the utility of the benzodiazepine lorazepam ( 14), the anticonvulsants carbamazepine (27) and lamotrigine (37), or the phenothiazine fluphenazine (in contrast to the favourable effect of the TCA amitriptyline [30]).Also, the sodium channel blocker mexiletine was ineffective compared with morphine for chronic postamputation pain (39).Two studies showed no effect of mexiletine or the comparator drug amitriptyline in HIV neuropathy (28), and no effect of nortriptyline or morphine in sciatica (35).These latter two results raise questions about the generalizability of research in NP, in which 80% of the studies are performed in postherpetic neuralgia and diabetic neuropathy (12).There appears to be no good evidence supporting the analgesic superiority of the gabapentinoids (gabapentin and pregabalin) over TCAs (29,34) in these RCTs; amitriptyline was found to be superior in two studies (26,36).Combinations of gabapentin added to morphine (33) and oxycodone added to gabapentin (40) resulted in a 20% and 33% increase in analgesia, respectively.
These comparative RCTs of different analgesics indicate efficacy within three broad classes of agents for NP -antidepressants, opioids and gabapentinoids.The structures of these agents are diverse.The receptors for opioids, and their location and mechanisms of action, have received considerable attention and are reasonably well characterized (55).The pharmacodynamics of gabapentinoids (gabapentin and pregabalin), a more recently available class of drugs, focus on the alpha-2-delta subunit of Ca 2+ channels and interactions with descending noradrenergic monoamine systems (56,57).Ironically, despite being the oldest and most-studied drugs for NP, the analgesic antidepressants are incompletely understood from a pharmacodynamic perspective.Antidepressants have been well characterized in terms of their ability to inhibit noradrenaline and serotonin reuptake (58), but antidepressants as a class of drugs have a multiplicity of pharmacological actions that potentially contribute to pain modulation (Table 6) (59).The focus on amine mechanisms for explaining analgesia by antidepressants reflects their actions in relieving depression (the biogenic amine hypothesis of depression), as well as the involvement of noradrenergic and serotonergic systems descending from the brain stem into the dorsal horn of the spinal cord, in regulating pain signals as they enter the central nervous system.It is important to recognize that, while depression is common in patients with chronic pain (60), and while depression and chronic pain share mechanisms (61), pain relief with antidepressants is independent of antidepressant actions (14,62); this dissociation has been recognized since the efficacy of antidepressants for NP pain was clearly established (63).Several lines of investigation, including the use of amine receptor antagonists and gene deletion approaches, indicate the requirement of noradrenergic and serotonergic systems for antinociception by antidepressants in several preclinical models (59,64).However, several experimental lines of evidence indicate that antidepressants also interact with other mechanisms that contribute to their ability to modulate pain signalling (59).

Comparative RCTs of antidepressants
Comparative RCTs of different antidepressant agents generally indicate that drugs that potentiate noradrenaline and serotonin, or predominantly noradrenergic mechanisms, are more effective than serotonergic agents (such as the SSRIs), which are less or even not effective.The presence of both facilitatory and inhibitory descending serotonergic systems could be an important reason for this (64).There is some evidence that the more noradrenergic, weaker serotonergic TCAs nortriptyline (49) and desipramine (46,51) are equal to or more effective than the dual noradrenergic/serotonergic agent amitriptyline.Other data from two RCTs using the tetracyclic comparator antidepressant maprotiline, which is more noradrenergic than nortriptyline and desipramine, showed it to be less effective than amitriptyline (47,48).Thus, some serotonin augmentation appears to be important but the optimal proportion may not have been found yet.In keeping with this, the strongly serotonergic/weakly noradrenergic reuptake inhibitor venlafaxine appears to provide less clinically meaningful relief than the noradrenergic/serotonergic TCA imipramine (50).However, some comparative trials have also produced data that differs from these general observations because the serotonergic TCA clomipramine and the SSRI paroxetine have been shown to be effective in some instances (42)(43)(44).One trial showing clomipramine to be superior to nortriptyline in central pain (42) actually investigated the peripherally generated pain disorders of postherpetic neuralgia, phantom pain and causalgia; this profile of indications differs from many other trials.These RCTs involved small numbers of patients and the clinical meaningfulness of the results was not clear.
There may be more than one pain mechanism in an individual and between individuals, and multiple mechanisms may contribute to analgesia.Both TCAs (and opioids) are indiscriminate in relieving steady pain, jabbing pain and allodynia (47,49).To date, clear mechanism-based treatments have been elusive, except perhaps for carbamazepine and the shock-like pain of trigeminal neuralgia.There is some evidence, in three crossover studies (47)(48)(49), that different subjects may respond to one antidepressant and not the other, making the trial-and-error use of different drugs within this class a reasonable clinical practice.
It appears that a tricyclic structure is important for antidepressant analgesia, given the lack of effect of the non-TCA mianserin and the lesser effect of the tetracyclic noradrenergic antidepressant maprotiline, compared with the noradrenergic TCAs desipramine and nortriptyline, which appear equal to amitriptyline in efficacy for pain relief.Structure-activity considerations may not necessarily assist in identifying more effective agents.Demethylation of the side chain seems to be important in reducing side effects, as seen with amitriptyline and its demethylated metabolite nortriptyline (49).It is of interest that the tricyclic-based structure carbamazepine, although very effective for the shock-like NP of trigeminal neuralgia, is not very effective or not useful at all for other types of NP.From a structural perspective at the molecular level, the planar structure of TCAs may have some flexibility, allowing the molecule to assume multiple conformations and thereby engage several different pharmacological actions (ie, interact with several transporters, receptors and ion channels).Tetracyclic structures may have more structural rigidity and a lesser ability to interact with several molecular sites.While this may be beneficial for the side effect profile, it might sacrifice actions that actually contribute to efficacy.Furthermore, it may be that a drug that exhibits a multiplicity of actions that affect several aspects of perturbed pain pathways can result in

Prostaglandin reduction
Data from reference 59.GABA Gamma-aminobutyric acid synergistic or additive effects, and these are lost with the more specific agents.The concept of polypharmacy or combination therapy using different agents for NP has been addressed directly (65).It may be that the more one strives for a highly selective agent, the more one may sacrifice some of the actions that are contributing to efficacy.We have no way of knowing how individual opioids and anticonvulsants compare because there are no head-to-head trials that compare them.There are some long-term observational data that suggest oxycodone is 'most used' and 'most preferred' over other opioids (53,54).
Why is there partial efficacy with all drugs to date?It is not clear why all drugs proven by RCT to exhibit significant pain-relieving effects on NP are only moderately effective, but there are several relevant considerations.One potential reason for limited responses is that damage to or perturbation of the nervous system (peripheral or central) may be so well established in these conditions that the system is only amenable to partial restoration of function (66).A second potential reason is that not enough aspects of the mechanisms that drive the pain are being affected by the drugs used.Multiple mechanisms (both peripheral and central) are implicated in chronic NP, both within an individual and between individuals, and this complexity may be prominently involved in generating and maintaining the pain.Perhaps more attention should be paid to quieting the peripheral input to settle the resultant secondary sensitization that occurs in the central nervous system, as recent cases suggest (67)(68)(69).In this regard, existing and improved topical analgesics (70) might be considered as adjuvant therapies along with systemically active agents.A third reason to consider is the multidimensional nature of the chronic pain experience.Thus, pain is a complex experience and has well-recognized sensory, affective and cognitive aspects.In this context, any approach using drugs that target only sensory aspects of chronic pain may only be capable of eliciting a partial response.The developing literature on fibromyalgia may be useful to consider in this regard.While this condition is not necessarily regarded as NP (it lacks an explanation of how 'damage' to the peripheral or central nervous system occurs), it functionally exhibits peripheral and central sensitization in common with other forms of NP and, importantly, responds to drug classes that are used to treat NP (71).Current treatment guidelines recommend multimodal approaches to treatment such as pharmacotherapy, exercise, education, counselling and other psychological approaches, thus combining medical management as well as self-management strategies (72,73).Perhaps the therapy of more classic NP conditions, in which damage to the central and peripheral nervous system is known to be involved, needs to consider multimodal treatment strategies as well as combinations of drugs with different pharmacodynamics for more effective management.Because of the inadequacies of monotherapy for many patients with severe NP, and because no magic bullet seems imminent, head-to-head RCTs of combinations of drugs useful for NP appear to be important.Head-to-head trials have limitations regarding the evaluation of efficacy if a placebo is not used and the drugs are equal; this type of trial may require more statistical power to show a difference.Limitations of the present review are listed in trial exclusion criteria (see the Methods section) in that the focus is on oral drugs and some NP conditions were excluded.Because of the dearth of head-to-head RCTs in NP and the inadequacies of some of those described here, we clearly need more and improved studies of this nature so that clinicians, grappling with these difficult problems, will know whether newer and more expensive drugs are superior to or comparable with a standard therapy in terms of efficacy, and whether they have fewer significant and intolerable side effects.Investigators designing RCTs of any sort in NP would be wise to follow the Initiative on Methods, Measurement, and Pain Assessment in Clinical Trials guidelines for chronic pain trials (73).Regulatory bodies such as the FDA and Health Canada should require comparative data to be part of the drug approval process.In any trial, principal investigators should have a major role in study design, data analysis and writing the article, and should insist on submission for publication.In the future, clinical trial registries will help to track the course of an RCT and whether it is published.Access to trials submitted for new drug approval are available on the FDA website (www.fda.gov) if submitted for drug approval and unpublished.
DISCLOSURES: Dr Watson has collaborated on clinical trials with Purdue Pharma (Canada) and received reimbursement for presentations by Pfizer Canada Inc and Merck-Frosst Canada Ltd. Dr Gilron has received research support from Pfizer, sanofi-aventis Canada Inc, Novopharm Ltd (Canada), PharmaScience Inc (Canada) and Apotex Inc (Canada), and he has received honoraria for consultations and speaker fees for educational presentations from Pfizer, Merck-Frosst, Johnson & Johnson Inc (Canada), Ortho-McNeil (USA) and Janssen-Ortho (Canada).Dr Sawynok holds intellectual property licensed to EpiCept Corp (USA) and has received a research grant from this source.She has also had small-scale pharmaceutical industry grant support from Gensia Sicor Pharmaceuticals Inc (USA), Proctor & Gamble (USA) and Abbott Laboratories (USA).

Figure 1 )
Figure1) Prevalence of some forms of peripheral neuropathic pain.DDD Degenerative disc disease.Adapted from Irving(5)

Table 1 Categories of peripheral and central neuropathic pain
*Trigeminal neuralgia is a unique and common form of neuropathic pain with a differing medical and surgical approach from other forms of neuropathic pain (except the rare condition of glossopharyngeal neuralgia)

Table 3 excluded comparative head-to-head randomized controlled trials Study Reason for exclusion
Rowbotham et al, 1991 (24); lidocaine versus morphine in postherpetic neuralgia Intravenous study, short term Kok-Yuen et al, 2008 (25); topical amitriptyline versus lidocaine in NP Topical preparations NP Neuropathic pain

Table 4 Comparative trials of different potential analgesics in neuropathic pain (n=17) Study, condition, duration of pain Jadad score Drugs, daily dose
for PPI of SF-MPQ, PGIC and CGIC; SF-36 results for both groups better than placebo

Table 5 Comparative trials of different antidepressants in neuropathic pain (n=10) Study, condition, duration of pain Jadad score Drugs, daily dose
AT Amitriptyline; CBZ Carbamazepine; DES Desipramine; M Maprotiline; N Noradrenergic; NNT Number needed to treat; NSD No significant difference; NST Nonsignificant trend; NT Nortriptyline; RCT Randomized controlled trial; S Serotonergic; SNRI Selective serotonin/norepinephrine reuptake inhibitor; SSD Statistically significant difference; TCA Tricyclic antidepressant