The utility/futility of medications for neuropathic pain – an observational study

2.1 Ethical issues

One hundred and twenty three PDN subjects and 87 PTN subjects were eligible for statistical analysis in the RELIEF study. The protocol was approved by local Ethics Committees. Included were male and female English-speaking subjects of 18 years and older with a minimum of 3 months duration of PDN or PTN.

The demographics of the groups are seen in Table 1.

At inclusion, a physical examination was done to document the presence of neuropathic pain, and informed the patients about this observational study.

2.2 Diagnostic criteria for neuropathic pain

2.3 Statistical analysis

Statistics were calculated on input data and at each visit interval. Descriptive statistics included listings, summary tables (mean, standard deviation, median, minimum, maximum, q1 and q3) and were presented by subject group, gender and age, among others (see Table 1).

The primary outcome variable for assessing improvement was Patient’s Global Impression of Change – PGIC that is a global measure indicating the degree of change in the overall status of the subject as noted by the subjects themselves [21].

3.1 Changes in pain-intensity

Although the RELIEF study did allow for appropriate statistical evaluation of the PROs, including sensitivity to change of PGIC, surprisingly few of the subjects were actually improved by treatment over the course of the study. A high number of subjects included had a NRS/VAS above 4/10, the arbitrary level that indicates at least moderate pain and inadequate pain control. This was an indication that any treatment offered before the initiation of the RELIEF study had not had much effect. The median NRS score was 6.0 for both PDN and PTN at intake. These scores did not change significantly throughout the 2 years of the study despite efforts through medication change in many subjects to control the pain.

3.2 Patient’s Global Impression of Change

The PGIC data show that very few subjects were improved between visits (see Table 2). PGIC used in the study had a seven-point Likert scale: “Very much improved”, “Much improved”, “Minimally improved”, “No change”, “Minimally worse”, “Much worse”, “Very much worse”. The frequency of subjects who reported “Much improved” or “Very much improved” from Visit 1–3 was 14.3% in the PDN group and 16.0% in the PTN group. From visit 1–4, the numbers were 14.3% in the PDN group and 11.5% in the PTN group.

The PGIC data show also that some subjects were worse over the course of the study (see Table 2). The frequency of subjects who reported “Much worse” or “Very much worse” from Visit 1–3, was 8.0% in the PDN group and 3.7% in the PTN group. From Visit 1–4, the numbers were 8.6% in the PDN group and 7.7% in the PTN group.

There was a statistically moderate correlation between PGIC and CGIC (Clinical Global Impression of Change – the clinicians’ impression of the treatment effect) at all visits with the CGIC indicating less improvement than the subjects reported themselves.

3.3 Effects of medications

Despite the above grouping of medications to simplify statistics, it was impossible to correlate any improvement with specific medication changes. A part of the problem is that there were often multiple changes in medications at a single visit.

A surprising finding was the high use of COX inhibitors (43.3% overall and 80.6% in the general practice group at visit 3), and opioids combined with COX inhibitors (29.5% overall and 43.1% in the general practice group). Also surprising, was the rather low use of antidepressants and antiepileptics – see Table 3.

There was a further difference in medications used in general practice as compared to the specialty clinics. For the total number of participants in the study, the percentage NOT receiving first or second line drugs (excluding opioids) for neuropathic pain (TCAs, other SSRIs and NSRIs and anticonvulsants) was 80.3% in the general practice group while in the specialty clinics it was 33.8%. Of those subjects in specialty clinics (138 subjects), there were 80 new trials of first line therapies during the course of RELIEF Part 1 (the first 6 months). There were only six new trials of these medications in the general practice group (72 subjects).

A proportion of the subjects appeared to have appropriate trials of medications recognized as first, second- and third-line medications for the treatment of neuropathic pain [6], [7], [8]. Sixty-six subjects had combined therapy of medications from at least two classes of those recommended.

3.4 Low health-related quality of life

Another perspective on the lack of treatment effect was the low quality of life of the subjects as reflected in the evaluations using the EQ-5D weighted index and overall health related quality of life at the conclusion of the study. Subjects were divided into three groups: mild pain (NRS 0–4), moderate pain (NRS 4–7) and Severe pain (NRS>7). There were 38 PDN subjects with severe pain and they had a median EQ-5D weighted index of 0.4/1.0 and an overall quality of life median score of 0.3/1.0. There were 29 PTN subjects with severe pain and the median score for this group with the EQ-5D was 0.6/1.0 for the Weighted Index and with 0.5/1.0 for the Overall Quality of Life Index.

4.1 Why did so few patients experience global improvement?

Was this due to inadequate treatment or are there other explanations? Treatment guidelines for neuropathic pain in Canada, published for all medical practitioners, exist but they were first published during this study [7], [8]. Some physicians involved in RELIEF were also practice guideline authors [8]. It is assumed that they were using the guidelines but this is unknown as there was neither an expectation nor a stipulation that specific guidelines should be followed in RELIEF. RELIEF was purely a 2-years observational study of treatment and outcome of patients with neuropathic pain treated mostly in pain clinics, diabetes clinics but also in one large general practice clinic in Canada.

Other guidelines have been published and were also available; a comprehensive one was published at the start of this study [12] and newer reviews have not modified guidelines to any significant extent nor have they added more effective medications [14], [22], [23]. This is also true for the SNRIs venlafaxine and duloxetine that have rather high “numbers needed to treat” (NNT), and they do increase risks of serious serotonergic symptoms, especially when co-administered with other serotonergic drugs, tramadol in particular.

The RELIEF subjects attending pain specialty clinics were taking a variety of medications indicated as first, second- and third-line therapies recommended by the then existing and current guidelines without effect. Many were on combination therapy which is currently recommended, specifically, antiepileptic and antidepressant drugs plus or minus opioids [24], [25], [26], [27], but this was also without effect. Data from the study indicate that 44 subjects were on two first line therapy drugs at study start and 12 were on three. In all, 77 subjects continued some form of combined therapy for at least a month. Those subjects in the general practice group not taking recommended drugs, i.e. not on combination therapy, did no worse, which is surprising.

4.2 Other negative studies?

Is this failure to respond to treatment despite the high use of recommended pharmacotherapy new or unusual? Toelle et al. [28] found a similar problem in looking at PDN in a survey across six European countries including patients from general practitioners and non-pain specialists. Despite a higher use of antiepileptic drugs than in the RELIEF study, 57% of the European subjects had a Pain Severity Index in the 4–6/10 range and 25% in the 7–10/10 range. This represents a substantial percentage which had inadequate pain relief.

Toelle et al. reported a lower use of opioids/“opioid compounds” (16.4%) than seen in RELIEF (51.9% at visit 1, 47.6% at visit 4), but as was also found in RELIEF, a high use of COX inhibitors (41.4%) and “standard analgesics” (62%), although these “standard analgesics” were not described and it could be assumed that they were over-the-counter preparations such as paracetamol and ibuprofen. Forty four percent were also taking “over the counter” medications in Toelle’s group. In the RELIEF study, Toelle’s two medication classes (“NSAIDS or COX-2 inhibitors” plus “over the counter medications”) were grouped together and in RELIEF, 47.1% at visit 1, 43.3% at visit 4 were taking these medications. Nine percent of the Toelle subjects were not treated with prescription medications. In the RELIEF study, 11.4% of the subjects were not treated with prescription medications and all of the non-treated were in the general practice group. This is similar information to that in two previous English surveys where only 44%–72% of the patients from a general practice population had pharmacologic treatment initiated when they presented with different neuropathic pain diagnoses [29], [30].

4.3 Satisfaction with treatment versus pain relief

In assessing subject satisfaction with medications in the RELIEF study, the numbers were 12.5% completely satisfied (“much improved” on the PGIC), 34.0% moderately satisfied (“improved” Table 2), 31.6% neutral and 16.7% moderately/completely dissatisfied (“worse”/”much worse” on the PGIC – Table 2). This is similar to the Toelle study although a higher percentage in that study was “satisfied”. The Toelle study reported 15% completely satisfied, 47% moderately satisfied, 10% neutral and 22% moderately or completely dissatisfied. It is surprising that so many of the European subjects were satisfied despite not reporting any pain relief from treatment. This could represent an overall assessment of treatment, not just the effect of the medication. Patients are often “satisfied” if they have been well taken care of even though they may not have less pain. The Patient Global Impression of Satisfaction (PGIS) is not as good an outcome measure; Patient Global Impression of Change, the PGIC, is more precise.

This question of poor response to treatment was also raised at the Investigators’ Meeting held after completion of the statistical analysis of data from RELIEF, Part I. The investigators suggested that the high use of “COX inhibitors” and “combination/compounds” was due to these medications being available without prescription in Canada or without the need to write a triplicate prescription as was necessary for strong opioids. Investigators from the specialty clinics also suggested that many subjects received simultaneous care from general practitioners for the pain problems. The use of “COX inhibitors” and “combination/compound” medications was not standard practice for the investigators who are specialists.

4.4 Low effectiveness of guidelines-recommended drug treatments

The information that the PDN/PTN subjects were self medicating and also seeking treatment from general practitioners is another indication that more specialized treatment in pain and diabetes clinics was not effective, not because expertise was not available but that the best pharmacological treatments presently available are not very effective.

From a critical viewpoint of the literature on pharmacological therapy for neuropathic pain, there are several points to consider. A source of bias on the publicized effectiveness of medications for a variety of medical diagnoses is the report of better efficacy in clinical studies sponsored by “for profit organizations” than in those from independent sources [31], [32]. This can, in some studies, be due to selection-bias, when trials are focused on specific patient groups chosen specifically so that the trial drugs may be more effective. This means that the data from studies reporting the effectiveness of medications for treating neuropathic pain are optimistic and in reality, these medications are less effective than the pharmaceutical industry reports. This is also in part due to publication-bias: positive studies are more often published, whereas “negative” studies (like the present study) are often not published, especially if the study was funded by industry.

There is also the problem of unmasking effects in drug studies using inert placebos [33]. Lack of effect and lack of side effects of the placebo treatment mean that subjects are able to correctly identify being on the placebo arm. Although this would tend to decrease the placebo effect of treatment, it could also increase the nonspecific (“placebo”) positive effect for the investigational drug which does have side effects.

The publication-bias problem is that all drug trial studies, both from the pharmacological industry and independent of industry, are not reported in the pain literature. The negative studies which are not reported do not enter into the cumulative statistical analyses of effect in most reviews of the Cochrane type [34]. Recently, some Cochrane reviews have been modified to include them [35]. And then there is the discussion of statistical bias in how data are manipulated [36].

4.5 Low health-related quality of life predicts poor outcome of drug treatment

An alternative explanation for the failure of treatment in RELIEF may relate to a finding by Otto et al. [37]. They analyzed three studies and compared treatment response to SF-36 scales and found a correlation between responders and higher scale scores (better mental and physical health generally) on SF-36. The RELIEF subjects had similar SF-36 scores to the pooled data in the Otto et al. study and RELIEF had a high number of non-responders with low scores as predicted by the SF-36 results in the Otto et al. paper. Similarly, two recent studies showed that patients with peripheral neuropathy who are “catastrophizers” are less responsive to first line pharmacotherapy [38], [39]. There was no evaluation of catastrophizing in the RELIEF population but it is possible that a part of the population studied were catastrophizers.

4.6 Conclusions and implications

In conclusion, the current study supports that patients with neuropathic pain continue to receive inadequate treatment for their pain. The reasons for this may include inadequate use of appropriate first line agents for neuropathic pain as well as a poor response to appropriate agents when they are used.

There is a need for more effective medications for neuropathic pain as well as non-pharmacological treatments [40], [41].

There is a need for broader education regarding approved guidelines so that primary care can move from the use of medications not included in treatment guidelines, i.e. COX inhibitors, to more effective medications. The COX inhibitors, besides being ineffective, may also have serious side effects from the cardiovascular (myocardial infarction and stroke) and gastrointestinal (painless, bleeding ulcers) systems [42].

There is also a need for general access to ALL data on new medications, both positive and negative, so that those who read the pain literature have an unbiased perspective on the effectiveness and safety of new drugs for neuropathic pain. Both patients and doctors are disappointed when the proclaimed highly effective therapies do not measure up in the real world. Clearly, there is a very real need for better medications for neuropathic pain [43].

And, lastly, there is a need for improved clinical trial protocols that can demonstrate more accurately the clinical effects of existing and new medications proposed for the treatment of neuropathic pain. This problem has been addressed by the US Food and Drug Administration who have been instrumental in organizing the Analgesic Clinical Trial Innovations, Opportunities, and Networks (ACTION) group to increase the assay sensitivity and efficiency of analgesic clinical trials [44], [45]. Hopefully, in the future, better clinical studies can indicate appropriate therapies for patients of different phenotypes that present with various neuropathic pain diagnoses.