Accelerating the translation of research findings to clinical practice: insights from phantom limb pain clinical trials

Phantom limb pain (PLP) is defined as a painful sensation in an amputated limb or pain that follows partial or complete deafferentation. It belongs to a group of neuropathic pain syndromes, and its treatment is challenging and often refractory to many modalities, causing significant burden and suffering. The prevalence of PLP after amputation ranges between 60–80% (Limakatso et al., 2020). In some cases, the pain might decrease with time and even fade away, but studies have shown that even after two years, a majority of the patients still present pain, and only a small percentage have reported a decrease in intensity of pain (Erlenwein et al., 2021). To date, there is no guidelines for the treatment of PLP and most therapies have low levels of evidence. Possible treatments include the use of pharmacotherapy including opioids, antidepressants, neuroleptics, muscle relaxants and anticonvulsants (Flor, 2002), a variety of surgical procedures (Nikolajsen & Christensen, 2015), nerve blocks, local anesthesia, transcutaneous electrical nerve stimulation (TENS)(Kern et al., 2012), acupuncture, psychological interventions, physiotherapy, and mind-body techniques such as hypnosis(Bamford, 2006; Moura et al., 2012).

Conventional treatments used for PLP, such as pharmacologic agents, have pertinent limitations. The use of pharmacologic agents for controlling pain is associated with several side effects and can become unsustainable in the long term (Els et al., 2017). Opioids are not recommended but are widely used drugs for chronic pain being associated with constipation, fatigue, dizziness, nausea, vomiting, physical dependence, addiction, and others (Hoots et al., 2018). In addition, these drugs worsen central sensitization thus leading to longer and more intense neuropathic pain (Braulio et al., 2018; Kim et al., 2014). Beside opioids, antidepressants and antiepileptic drugs are recommended and used medications but they can lead to sedation, dizziness, nausea, and vomiting (Quintero, 2017). So far, studies that reported an effect of such interventions are inconclusive and have low quality, reporting a small effect of these drugs on pain, function, mood, sleep, quality of life, treatment satisfaction (Alviar et al., 2016).

The heterogeneous and partially effective treatment approaches for PLP can be explained by the lack of understanding of its pathophysiology. Current evidence suggests that PLP is driven by maladaptive cortical plasticity, mainly in the sensorimotor areas (Duarte et al., 2020; Gunduz et al., 2020; Pacheco-Barrios, Pinto, et al., 2020). It has been reported that this cortical reorganization is associated with higher PLP levels, thus converting it into a potential treatment target. Paradoxically, the most common treatments (opioids, antidepressants, and antiepileptic drugs) do not address PLP underlying pathophysiology directly.

Targeting maladaptive neuroplasticity

Some therapies, such as motor cortex stimulation (Pacheco-Barrios, Meng, et al., 2020) and somatosensory representation techniques of the phantom limb (Cuenca-Martínez et al., 2021; Thieme et al., 2016), are promising options since they are safer and low-cost, and can potentially revert the maladaptive plasticity associated to PLP.

Motor cortex stimulation

Considering phantom limb pain (PLP) is a hard-to-treat condition that is usually refractory to common therapies, noninvasive brain stimulation (NIBS) techniques such as transcranial magnetic stimulation (TMS) and transcranial direct current stimulation (tDCS) targeting the primary motor cortex (M1) have recently gained traction as a possible alternative therapy for this condition (Collins et al., 2018; Pacheco-Barrios, Meng, et al., 2020). The rationale behind stimulating M1 for PLP is based on the sensory-motor cortical rearrangement and lower cortical inhibition that occurs after amputation (Nardone et al., 2019; Raffin et al., 2016). So far, clinical trials using tDCS and rTMS for PLP have shown promising results, reflecting clinically significant changes in pain scales (Pacheco-Barrios, Meng, et al., 2020). Table 1 summarizes the current evidence for NIBS targeting M1 and their results.

Table 1.

Evidence on motor cortex stimulation for managing PLP.

Study (year)	Design	Sample size	NIBS type	Number of sessions	Main outcome and results
Bolognini (2015) (Bolognini et al., 2015)	cross-over	8	tDCS on M1	5	- VAS. - With respect to PLP intensity, the rmANOVA showed a significant effect of the tDCS week (F1,7= 6.64, p= 0.04, pη 2 = 0.49), showing a significant difference between Sham tDCS (−9%) and Active tDCS (−28%, p= 0.04)
Malavera (2016) (Malavera et al., 2016)	parallel	54	rTMS on M1	10	- VAS - −53.38±53.12% vs −22.93±57.16%; mean between-group difference=30.44%, 95%CI 0.30, 60.58; p=0.03
Kikkert (2019) (Kikkert et al., 2019)	cross-over	17	tDCS on M1/S1 of the missing hand	1	- Short pain questionnaire (SPQ - This resulted in a significant difference in PLP modulation between the intervention and sham conditions immediately following stimulation (Wilcoxon Z = −2.73, p = 0.006, r = −0.73)
Gunduz, Pacheco-Barrios, Pinto (2021) (Gunduz et al., 2021)	factorial	112	tDCS on M1		-VAS - analyzed the main effects separately and we found a statistically significant main effect of tDCS on PLP (beta coefficient = −0.99, P = .04) when comparing active versus sham tDCS, controlling by PLP-PLS index and PLPPLS category

Studies investigating repetitive TMS and tDCS as therapies for PLP have conveyed conflicting results regarding its effectiveness towards pain alleviation. While some rTMS studies report improvement in both active and sham groups after a 5-session protocol, but no difference between groups, others with longer protocols report clinically significant pain reduction in the active groups compared to sham (Ahmed et al., 2011; Irlbacher et al., 2006; Malavera et al., 2016). Whereas studies assessing the use of tDCS report more consistent and significant, short-term improvements in pain symptoms, which are attributable to increased excitability in M1 (Bolognini et al., 2013; Bolognini et al., 2015; Kikkert et al., 2019). This improvement suggests that individuals with PLP do have maladaptive cortical plasticity because of their pain, and therefore can benefit from brain stimulation techniques (Bolognini et al., 2013). However, the short-term effects of neuromodulation on M1 suggest that these techniques could benefit from additive therapies such as sensorimotor therapies like motor representation techniques (Gunduz et al., 2021). Currently our group is evaluating and refining the possible synergistic effects of tDCS over M1 and motor imagery for PLP alleviation.

Mirror therapy

Mirror Therapy (MT) is a type of movement representation technique that has been suggested as a therapeutical approach for treating limb pain and motor disabilities (Thieme et al., 2016). MT specifically uses the mirror’s reflection of a non-affected limb to provide visual feedback of normal pain-free movement to the affected limb, therefore the subject feels performing the movement through imagination (Wang et al., 2021). It was fist applied in amputees to relieve pain in the phantom limb and several studies have been published since then (Barbin et al., 2016; Wang et al., 2021). Pooled analyses are, however, conflicting. Thieme et al. performed a meta-analysis showing an inconclusive effect of MT in reducing pain on phantom limb (Thieme et al., 2016), while a recent meta-analysis showed reasonable evidence for its analgesic effect in the same population (Thieme et al., 2016). The unsettled evidence might be related to the different sorts of control interventions used in RCTs, leading to heterogeneous findings (Thieme et al., 2012). We currently question not only the need, but the value of mirror feedback in inducing analgesic effects on PLP – beyond the effects of components of motor imagery and limb movement provided by movement representation techniques in general.

To illustrate, Brodie et al. randomized 80 subjects to either mirror therapy or a control condition (i.e., no mirror feedback), and results depicted similar analgesic effects between both groups (Brodie et al., 2007). Similarly, Rothgangel et al. found no significant differences in pain reduction between traditional MT and sensorimotor exercises on a PLP sample of 75 subjects (Rothgangel et al., 2018). In a recent factorial trial published by our group, 112 subjects were randomized to different levels of tDCS and mirror therapy (active M1 tDCS or sham tDCS and active mirror therapy or covered mirror therapy) (Gunduz et al., 2021). MT and its control differed only by the presence of mirror feedback, but both groups performed the same sensorimotor tasks. We did not find synergistic effects of tDCS combined with MT, but active tDCS had a significant main effect in comparison to sham tDCS, while mirror therapy and its control conveyed similar results (beta coefficient: 0.71, P = 0.16). We hypothesize that this phenomenon was secondary to a real effect induced by both active and covered mirror therapy whether than due to high placebo responses, since their effect sizes (0.81 – 1.22) were notably larger than placebo responses on previous PLP trials (varying from 0.38 to 0.53)(Hsiao et al., 2012; Nikolajsen et al., 2006; Pacheco-Barrios, Meng, et al., 2020). Notably, we also performed secondary analysis on phantom limb sensations and found a statistically significant pain reduction favoring covered MT group (beta coefficient = −1.01, P = 0.03).

It might be that the visual feedback not only is negligible to induce analgesia but may jeopardize the engagement in somatosensory training. Both M1 tDCS and movement representation techniques are expected to engage similar neurophysiological circuits to treat pain, and therefore there has been an attempt to couple MT with neuromodulation techniques in clinical research. The aforementioned body of facts has led us to believe that somatosensory training without mirror visual feedback can be a better fit for M1 tDCS on PLP and a turning point to yield significant and larger effect sizes.

The process of translating research findings into clinical practice

Unfortunately, the “bench to bedside” translation of neuroplasticity-targeted therapies for PLP remains a challenge. The process in which research evidence is created, disseminated and applied in clinical practice is known as knowledge translation (Curtis et al., 2017). This process is a key part of a bigger one called knowledge-to-action cycle (Graham et al., 2006), which shows different steps, divided in three phases, for the transfer of knowledge from pre-clinical environments to the general population. In this cycle (Figure 1), the first phase consists of the evidence generation, generally starting from pre-clinical laboratories and then moving through the three phases of clinical trials. These steps assure the gain of knowledge on safety and efficacy. Moreover, those interventions with great profile under controlled conditions, need to be explored in real-world settings to test their effectiveness and impact on terms of cost in the health systems. At this stage of evidence generation, pragmatic trials and cost-effectiveness studies are needed.

Figure 1.

Clinical research translation process

The second phase is the synthesis of the available evidence, especially confirmatory trials (phase III), pragmatic trials, and economic studies. This can be accomplished using several tools such as systematic reviews and clinical practice guidelines, which allow easy access and updated information to clinicians (Curtis et al., 2017). Nonetheless, a rigorous analysis of the available data must be done. The third phase, known as the action cycle or implementation phase, consist of the steps needed for that evidence to be applied in the clinical field, such as adapting the knowledge to the local context, assessing the possible barriers for knowledge usage, selecting and implementing the interventions, and monitoring the knowledge use among others (Curtis et al., 2017; Graham et al., 2006). It is important to highlight the cyclic nature of the knowledge translation process, meaning results from advanced stages can inform prior steps and re-start the cycle, a process known as evidence refinement.

In order to move from the evidence generation and synthesis phases to the implementation stage several regulatory issues must be cleared by proper institutions in each country. In the US, the highest authority is the Food and Drug Administration (FDA), and they must give their approval.

According to the FDA, for an intervention to be able to be offered in the U.S it must have been reviewed by the FDA’s Center for Drug Evaluation and Research (CDER). It must pass through all the research phases (preclinical and clinical) and show that its benefits outweigh “it’s known and potential risks for the intended population” (Ciociola et al., 2014).

However, the fact that an intervention is approved by the FDA does not necessarily mean that it will be available for the general population. Other factors come into play: the amount of resources, the training, clinician behavior, the cost, the entity that will cover the expenses, among others (Lalu et al., 2020). Many insurances will not cover the expenses of these new treatments and therefore handicap this process (Lalu et al., 2020). That is why it is important for researchers even in the first stages of research to plan for the long term, in other words, explore interventions that may be useful, with long lasting effects and that are cost-effective (Curtis et al., 2017). In the case of PLP, one of the therapies mentioned before was the non-invasive brain stimulation, which presents itself as a safe, effective and cheap alternative for this population (Pacheco-Barrios, Meng, et al., 2020; Zaghi et al., 2009).

The next step: designing a pragmatic trial

Over 30 Randomized Clinical Trials on tDCS in chronic pain conditions have been published according to a recent evidence-based guideline reporting a level B of evidence for the use of anodal M1 tDCS stimulation for neuropathic pain (Fregni et al., 2021). These studies include strict inclusion and exclusion criteria resulting in a relatively homogeneous study population with a small sample size (from 18 to 132 participants), and the application of the treatments is restricted to a controlled laboratory setting with highly trained researchers. Therefore, most clinical trials of tDCS do not represent the complexity of patients in a general population, and may overestimate treatment efficacy and our understanding on responders to this therapy (Wexler, 2016).

The next step in expanding our knowledge on the application of neuromodulatory techniques to treat PLP and to accelerate its translation to clinical practice (Figure 1) is the design of pragmatic trials. These pragmatic studies, unlike the “classical” randomized clinical trials focus on demonstrating efficacy of a specific therapy in a hermetic and controlled scenario, are focused on assessing the effectiveness of one or more therapies in scenarios closer to clinical practice in the real world (Hohenschurz-Schmidt et al., 2021). One interesting example of a pragmatic clinical trial is the “stepped-wedge” design, in which all participants start as the control group and clusters of subjects are randomized to the order in which they will receive one or more interventions (Ellenberg, 2018). Pragmatic trials can offer the possibility to generalize results to a broader population and capture key outcomes relevant to patients and physicians including overall survival, functional status, and costs. They are also useful for policy makers as they look at comparative effectiveness of interventions and, eventually, cost-effectiveness analyses (Maclure, 2009).

Thus, as we proceed through the PLP knowledge translation and face new research and clinical practice demands, the framework summarized in this editorial can help guide us on how best to improve the translation of clinical research into clinical practice and reduce the substantial discovery-delivery gap.