Design issues for studies into prevention of chronic pain: lessons from post-herpetic neuralgia.

Abstract

Chronic pain can cause major problems. Patients may experience physical, psychological and social disabilities as a consequence of their unceasing pain. To relieve pain physicians often prescribe treatments that can be burdensome and costly [1,2]. Because the results of treating chronic pain syndromes are often disappointing, attention should be focused on prevention. When developing and quantifying the effect of a preventive strategy for chronic pain, several issues must be addressed such as: (1) who is at risk of developing chronic pain and can they be easily identified? (2) can the pathophysiological pathway leading to chronic pain be successfully interrupted? (3) what is the efficacy of the prevention? (4) what is its effectiveness when applied in daily care? and finally (5) is it cost-effective? To answer these, it is necessary to obtain empirical data regarding not only the intensity and duration of pain, but also the quality-of-life and costs. Research on prevention of chronic pain is faced with specific problems. It is often impossible or unethical to apply a placebo and difficult to determine the primary endpoint. Moreover, as only a minority of patients will develop chronic pain, the size of a preventive trial must be large in order to secure adequate power. Post-herpetic neuralgia (PHN) is one example of a chronic pain syndrome that might benefit from preventive strategies. Depending on the definition applied, 9–34% of all patients with herpes zoster (HZ) develop PHN [3]. The risk is low (2%) in HZ patients ≤ 50 years of age, approximately 20% in those > 50 years of age, and approaches 35% in those > 80 years of age [4–7]. Severity of pain and inflammation during the acute phase of HZ are also risk indicators for PHN [3]. Inflammation of the peripheral nerve is supposed to be responsible for acute pain in HZ patients. The development of chronic pain (PHN) can be explained by two pathophysiological mechanisms: sensitisation and de-afferentation [8]. Briefly, sensitisation originates from the ongoing input of pain stimuli and results in hyperexcitability and allodynia. Viral activity and inflammation cause neural destruction; the subsequent loss of afferent neurones leads to spontaneous activity in de-afferented central neurones, which generates chronic pain. A promising strategy to prevent PHN is epidural administration of steroids with or without local anaesthetics at the level of the affected dorsal root ganglion [9]. This intervention supposedly interrupts the mechanisms leading to PHN in two ways: steroids forestall neural damage and thus de-afferentation because of their anti-inflammatory properties, and local anaesthetics block the ongoing spinal input from pain stimuli and thus prevent sensitisation. Moreover, epidural injection of local anaesthetics also achieves sympathetic blockade, which may play an important role in neuropathic pain modulation [10]. Although anaesthetists in many countries already advocate this epidural treatment and have suggested that its effects should be evaluated in a randomised study, studies on its (cost-) effectiveness are almost lacking. This dearth has most likely been caused by the above-mentioned problems. The PINE (Prevention by epidural Injection of postherpetic Neuralgia in the Elderly; ISRCTN32866390) study was designed recently to evaluate the preventive epidural strategy. This ongoing large, pragmatic, randomised, multicentre study, conducted in the Netherlands, will include 550 elderly HZ patients. Its aim is to quantify the effectiveness and cost-effectiveness of the addition of a single (non-image-guided) epidural injection of local anaesthetics and steroids to care-as-usual (i.e. administration of antivirals and analgesics) during the acute phase of HZ to prevent PHN. Patients will be randomised to either the care-as-usual group or to the index group receiving care-as-usual plus epidural injection. The primary endpoint is presence of HZ-related pain one month after onset of rash. Various dilemmas were encountered during the design and organisation of the PINE-trial. Below we will discuss these dilemmas and propose solutions, which may help researchers involved in the prevention of chronic pain syndromes. Patient selection is an important aspect of a clinical trial design. A preventive strategy should focus on patients who are at high risk of developing chronic pain. Furthermore, early interruption of the cascade of events leading to chronic pain is most effective. In the case of PHN, a trial should include patients who are ≥ 50 years of age with symptoms of HZ who consult their general practitioner (GP). To ensure (early) prevention of chronic pain, one should include only HZ patients with a rash existing no longer than 7 days. Since clotting defects may cause epidural haematomas after epidural injection, patients with haemostatic disorders should be excluded. If one opts for quantifying the efficacy of an intervention (does the intervention actually do what it is claimed to do?), a so-called explanatory or placebo-controlled trial design is essential. The intervention under study is compared to a placebo treatment. To rule out any placebo effect one should blind the patient for the administered treatment [11]. If one opts to assess the effectiveness of an intervention (given an efficacious intervention, does it actually work in the real world with all the non-compliance, refusals and exclusions?) a so-called pragmatic design is required. Here, the intervention is tested in a daily care setting and compared to another (existing or usual) treatment. As it is often unethical to deprive patients of the best available therapy, blinding patients is not feasible but also not desired. Any possible placebo effect in each group will be considered as part of the overall treatment effect, which would also be observed in future patients in daily care. Regarding PHN prevention, an epidural injection is quite burdensome for the patient and not without risks. Epidural haematomas, for example, can occur. Although rare, the risk of this severe complication makes it unethical to perform placebo epidural injections with saline. Furthermore, the addition of local anaesthetics is an essential part of the preventive strategy because the concomitant immediate pain relief they cause can never be realised using a placebo injection as control. Finally, it is preferable to quantify the effect of an epidural injection as might be expected in future practice. All these issues therefore demand a pragmatic (non-blinded) trial design in which an epidural injection is compared to care-as-usual. Furthermore, with respect to PHN prevention, considerable attention has been given to antiviral medicines. Meta-analyses, for example, have shown that these drugs, when given within 72 h from rash onset, have some effect on the duration of HZ-associated pain [12]. For this reason, physicians should prescribe antiviral medicines to all patients who present within 72 h. Currently, there is no clear evidence of a difference in efficacy between three antiviral drugs (acyclovir, famcyclovir, valacyclovir) on the development or duration of PHN. Hence, the choice of drug may be left to the GP. The antiviral drug prescribed to each patient should, however, be carefully monitored for future analyses. When assessing the effectiveness of a (preventive) intervention in a pragmatic study, the intervention should be executed under daily care conditions. Execution under ideal circumstances may lead to an overestimation of the effect that might be observed in future patients. In the case of PHN, the epidural injection might better be administered using image guidance to guarantee deposition of the injected drugs around the affected ganglion. However, fluoroscopy is not routinely available in most clinics. Epidural injection without image guidance merely reflects the procedure in most hospitals. Therefore, to reflect care-as-usual as much as possible, the non-image-guided approach to the affected ganglion may be chosen over image-guided targeting. Moreover, the analgesia in the corresponding dermatome, induced by the local anaesthetics, indicates proper deposition of the drugs at the targeted ganglion. If the injection does not result in local anaesthesia, the procedure can and should be repeated. The consequence of using a non-image-guided approach is that the study is confined to patients who present with HZ below the sixth cervical dermatome since an epidural injection can only be performed safely without fluoroscopy from that level downwards. The risk of false-positive errors (i.e. demonstrating a preventive effect which is absent in reality) and the difficulty in interpreting study results increase with the number of primary endpoints used. Therefore, trials should adopt a minimum of (primary) endpoints related to the clinical problem [13]. In the example of prevention of PHN, the presence of HZ-related pain 4 weeks after enrolment in the trial was chosen as the primary outcome in the PINE study, because it is the most widely advocated definition of PHN [13,14]. There is, however, some discussion about the definition of PHN with respect to the severity and the duration of pain after rash onset [13–19]. Some researchers, for example, define PHN as HZ-related pain after the rash has healed, while others describe it as HZ-related pain 1 month [13,14] or 4 months [19] after rash onset. Hence, the severity of pain at 1 month, the presence of HZ-related pain after 3 and 6 months, and the duration of pain should be studied as secondary outcomes. Also, for all endpoints, the presence of pain is to be assessed preferably by self-reporting [13]. To minimise bias, which is inherent to a subjective endpoint in a pragmatic trial, one should systematically document endpoints using structured written questionnaires. The size of a trial is directly related to the number of patients needed to detect a clinically relevant reduction of chronic pain. This number depends on the expected absolute risk in the control group. For example, based on a power of 0.8 and a two-sided alpha of 0.05, a 50% risk reduction from 30% to 15% requires 120 patients in each treatment group, while the same risk reduction from 12% to 6% necessitates 356 patients in each group. Hence, the lower the expected absolute risks the higher the number of patients needed. Clinicians invariably overestimate the number of patients available [20]. The reasons are several: (1) overly strict criteria for the inclusion and exclusion of potential subjects; (2) refusal of patients (or their physicians) to participate; (3) drop-outs once enrolled; and (4) erroneous memory of how many patients one actually sees with the condition in question. Researchers therefore should first perform a pilot study in order to discover the rate of patient inclusion. The recruitment rate realised in that study can then be used to estimate the number of physicians required to complete patient enrolment within an acceptable time period. Depending on the disease studied, this number can be quite large. To ensure the commitment of large numbers of physicians and thus accomplish a comprehensive study in daily care, it is desirable that the study be a co-operation between researchers from both pain and GP departments, where commonly patients at risk of developing chronic pain are first encountered. In summary, chronic pain is a common problem and is associated with large decrements in psychological health and daily functioning. Moreover, it can have a large financial impact on society. Efforts therefore should be made to develop and evaluate strategies to prevent chronic pain syndromes. The success of these efforts is highly dependent on the co-operation between GPs and pain specialists and on the implementation of preventive strategies in daily care. The above-described design dilemmas and suggested solutions, which emerged during a large study on PHN, might help future researchers interested in studying the prevention of chronic pain syndromes.