Post-herpetic neuralgia and phantom limb pain
This study found an annual incidence per 10,000 person-years adjusted to the 2010 UK age and sex distribution of 3.0 for post-herpetic neuralgia and 0.1 for phantom limb pain, with rates increasing for post-herpetic neuralgia during the study period (Figure 1). This study updates two previous similar incidence estimates based on UK primary care electronic records between 1992 and 2005 [6, 7]. Post-herpetic neuralgia decreased between the two earlier study periods from 4.0 (1992–2002) to 2.8 (2002–2005), compared to 3.4 per 10,000 in the current study. When the two most recent studies were age-sex standardised to 2010 levels, a small rise from 2.7 to 3.0 per 10,000 was seen between 2002–2005 and the current study. There were few cases of phantom limb pain with 2010 age-sex adjusted incidences per 10,000 person-years of 0.11 (95%CI 0.07, 0.17) in the current study compared to 0.08 (95%CI 0.04, 0.17) in 2002–2005 suggesting little change.
An annual age-sex adjusted incidence of painful diabetic neuropathy of 3.1 per 10,000 was found. However, the results of the GP questionnaire indicates that this is an over estimate as only 56% of those identified had a final diagnosis of painful diabetic neuropathy. A review of the codes and treatment patterns of those with and without a final diagnosis of painful diabetic neuropathy showed that recording was similar in both groups with none using a specific Read code. Consequently it was not possible to alter the case definition to improve sensitivity. It is feasible that the introduction of the UK Quality and Outcomes Framework (QoF) in 2004 resulted in GPs increasingly making a preliminary diagnosis of painful diabetic neuropathy which is then not confirmed at secondary care. Within this government initiative, primary care physicians are paid to review chronically ill patients, including diabetic patients, with some payment dependent on evidence of checking for complications such as neuropathy. A record of neuropathy testing increased sharply around the time of QoF introduction with one study reporting a rise from 8% of diabetic patient records in the 15 months to April 2003 to 66% in the 15 months to April 2005 with rates of recording then levelling . There has also been increased awareness of neuropathic pain with the introduction and marketing of new licensed treatments and publication of treatment guidelines. This may have resulted in an increase in both the diagnosis of neuropathic pain and the use of treatments such as amitriptyline and gabapentin for less well defined conditions such as neuropathic back pain. The rate of ‘painful diabetic neuropathy’ (as per our definition) increased over the study period from an annual incidence of 2.7 to 3.8 per 100,000 population (Figure 1).
Most other studies assessing the incidence of these neuropathic pain conditions have been within specific sub-populations, such as people with acute herpes zoster, diabetes or post-amputation. One earlier UK general population study reported incidence per 10,000 population of 5.4 for diabetic polyneuropathy and 1.1 for post-herpetic neuralgia . A study of shingles in patients aged 50 years or more (2000–2006) found that 19.5% had post-herpetic neuralgia at one month and 13.7% at three months which would give incidence rates per 10,000 population of 10.2 at one month and 7.2 at three months, which is in-line with this study’s findings given the older population .
Amitriptyline was the most commonly prescribed treatment in a first-line therapy for the post-herpetic neuralgia and painful diabetic neuropathy cohorts and the second most frequent after gabapentin in phantom limb pain. Gabapentin was also frequently prescribed as a first-line treatment in the painful diabetic neuropathy cohort and in post-herpetic neuralgia, with the use of both gabapentin and pregabalin increasing over the study period. This demonstrates a maintained shift from predominant use of opioid and non-opioid analgesics in the late 1990’s to use of tricyclic antidepressants and antiepileptics, with initiation of these therapies now evident in primary, rather than secondary, care. This shift appears to be the case across conditions as our painful diabetic neuropathy cohort will have included other neuropathic pain. Compared to the 2002–2005 data, amitriptyline remains a common first-line treatment although use has decreased in post-herpetic neuralgia and phantom limb pain (2002–2005 : 2006–2010, 50% : 39% in post-herpetic neuropathy and 41% : 21% in phantom limb pain). Antiepileptic prescribing has moved from carbamazepine to gabapentin and pregabalin, both of which are now used more frequently in phantom limb pain (2002–2005 versus 2006–2010: 17% versus 38%). Opioid use has increased slightly with tramadol now prescribed more often than co-codamol (codeine phosphate and paracetamol) as a first-line treatment in phantom limb pain. The changes may be due to a combination of an increasing evidence of efficacy in both placebo-controlled and head-to-head trials, education and marketing.
The prescribing analysis period of 2006–2010 was, for the most part, prior to the publication of the current UK and European recommendations on neuropathic pain treatment, although after the publication of key randomised controlled trials for the newer antiepileptics, antidepressants and opioids in neuropathic pain [10, 20]. The prescribing of antiepileptics and antidepressants as first-line treatment is consistent with the current recommendations and the shift from carbamazepine to gabapentin, or pregabalin is supported by evidence from clinical trials. The European Federation of Neurological Sciences (EFNS) 2010 guidelines recommends gabapentin, pregabalin, lidocaine plasters and tricyclic antidepressants in post-herpetic neuralgia and duloxetine, gabapentin, pregabalin, tricyclic antidepressants and venlafaxine in painful diabetic neuropathy . The 2010 UK National Institute for Clinical Excellence (NICE) guidelines recommends first-line treatment with duloxetine or, if contraindicated, amitriptyline for painful diabetic neuropathy and amitriptyline or pregabalin as first-line treatment options for other neuropathic pain . Duloxetine, venlafaxine (painful diabetic neuropathy) and topical lidocaine (post-herpetic neuralgia) use was uncommon in this study although there was evidence of efficacy before the study period  and all three treatments were included in guidelines as second-line interventions in 2006 . They were not recommended as first-line agents until 2010 [10, 20]. The common use of paracetamol first-line in the painful diabetic neuropathy cohort is consistent with guidance in type 1 diabetes from 2004 to 2010  although this recommendation has been superseded. GPs may have followed type 1 guidance until specific guidance for type 2 diabetes was available from NICE in 2008 . Paracetamol is currently recommended for use as a first-line treatment in combination with tramadol in neuropathic pain with acute exacerbations and this may partially explain the common use of both drugs in phantom limb pain .
Co-codamol was still one of the five most commonly prescribed first-line treatments in post-herpetic neuralgia and painful diabetic neuropathy and co-dydramol (paracetamol and dihydrocodeine) in post-herpetic neuralgia. While neither of these options is generally recommended as a first line treatment  it is not unusual to try conventional analgesics before moving to antidepressants and antiepileptics. The frequent prescription of capsaicin as a first-line treatment for post-herpetic neuropathy (unchanged since 2002–5) is not consistent with current evidence although it is recommended as a second line therapy in the EFNS 2010 guidelines .
The analysis of treatments in more than 100 patients showed that the most commonly prescribed daily doses for amitriptyline and pregabalin were those recommended by NICE (10 mg and 150 mg per day respectively) while higher dosage regimens were not uncommon. EFNS recommends a higher daily dose of tricyclic antidepressants (25–150 mg/day) and their suggested dose of gabapentin (1200–3600 mg/day) was prescribed first-line to a minority of the post-herpetic neuralgia and painful diabetic neuropathy cohorts. The analysis of pregabalin dosage regimen did not show a shift from a three to twice a day dosing although this has been suggested as a cost saving exercise. The majority of patients were prescribed one item as a first prescription and one therapeutic category. This is in-line with current guidance although there is some evidence that less than half of patients achieve significant benefit with any single treatment and that combined treatment may be more effective . Patients who received more than one therapeutic category usually received a mixture of an opioid and a non-opioid analgesic rather than combinations with demonstrated efficacy in neuropathic pain, such as gabapentin and opioids or nortriptyline . The proportion of patients who received first-line combination therapy had decreased since the last analysis for post-herpetic neuralgia (44% to 35%), painful diabetic neuropathy cohort (36% to 18%) and phantom limb pain (47% to 27%). Most treatment regimens (monotherapy and combination therapy) were prescribed at first-line (Additional file 2). This may simply be because many patients did not progress to subsequent/later-line therapy.
Neuropathic back pain and post-operative pain
The inclusion of neuropathic back pain and post-operative pain was considered exploratory, because these are less well defined than the other study conditions with no specific Read code to simplify recording by the GP. The initial strategy for identifying post-operative pain involved identifying records with a code for surgery followed by a record of post-operative pain, plus any marker that this was neuropathic. When few cases were identified, the search focused on breast and hernia surgery, which are associated with post-operative pain [13, 24]. The case definition was broadened to include those with a code for surgery followed by a record of neuropathic pain, or a combination of medical and therapy entries indicative of neuropathic pain. Only 49 patients were identified despite the use of a wide case definition. The low patient numbers might partially be due to our method of case selection, for example we would have missed cases if the GP recorded ‘pain’ and used free text to indicate the post-operative or neuropathic nature of that pain. It is also possible that there is under-diagnosis of post-operative pain in primary care or that previous studies have over-ascertained significant long-term post-operative pain. This area therefore warrants additional research.
Neuropathic back pain within the study definition increased over the study period from 49 to 62 per 10,000 patient-years. This cannot be considered to be a true incidence because, without a specific code for neuropathic back pain, the study definition lacked specificity. For example, patients with back pain or radiculopathy treated with tramadol would have been included in the analysis although tramadol is often also used in nociceptive, rather than neuropathic pain. Conversely patients with neuropathic back pain but without a neuropathic code and treated with opioid/non-opioid combinations would not have been identified. Additionally, radicular back pain is not always considered to be neuropathic. The prescribing patterns in the neuropathic back pain cohort differed from the other pain conditions, with tramadol the most frequent first-line therapy and other opioid/non-opioid analgesics commonly prescribed. This may be because of co-existing nociceptive pain for which tramadol efficacy is established [10, 25] or indicate that cases of purely nociceptive pain were included in the cohort.
Strengths and limitations
This study was based solely on primary care records, which provides a picture of routine treatment but has a number of resultant limitations. In particular, not all cases of post-herpetic neuralgia, painful diabetic neuropathy and phantom limb pain were identified using specific disease codes. The incidence of painful diabetic neuropathy is considered a maximum rate as the study definition will include diabetic patients with neuropathic pain which is not related to diabetes or non-painful neuropathy if this was diagnosed at the same time as the patient started a neuropathic pain therapy but for a different indication. The rates of disease we report are those recorded by the GP so any pain not reported to the GP will not be included. Additionally, treatments for pain other than neuropathic pain will have been included in the treatment analysis if they were started at the time of the first neuropathic pain record. More than a third of patients with neuropathic pain are known to have other chronic pain-related disease . We could not identify a first-line therapy for a number of patients possibly because treatment started before a diagnosis was made (phantom limb pain case definition included a record of neuropathic pain at least 3 months after amputation), over the counter medications were used, anti-epileptic treatments used in patients with epilepsy, treatments such as non-steroidal anti-inflammatory drugs were prescribed or when the patients did not require or want a prescription treatment. The validation questionnaire was sent to a small sample and, while the response rate was good, we have no data on the quality of the response and some bias is possible. For example, a record of a working rather than final diagnoses, or inconsistencies between the questionnaire and notes, may affect the likelihood of a reply.